1.22.2009

Another wider-than-expected loss for AMD

SAN FRANCISCO (Reuters) - PC chipmaker Advanced Micro Devices (AMD.N) posted a wider-than-expected loss in the fourth quarter as worldwide demand for PCs continued to shrink.
AMD reported on Thursday a net loss of $1.42 billion, or $2.34 a share for the quarter ending December 27, compared with a loss of $1.77 billion, or $3.06 a share, a year ago. Excluding certain items, the company posted a loss of 69 cents a share. Analysts, on average, had expected a loss of 56 cents a share, according to Reuters Estimates.
Revenue for the second-largest maker of central processing units for personal computers fell 33 percent to $1.16 billion, compared to analysts' estimates of $1.19 billion, according to Reuters Estimates. AMD said it expected revenue in the first quarter to decrease from the fourth quarter.

If AMD never made money during the best market condition, what hope is there when the economy is at its worst. For the first time the talk of bankruptcy for this company can never be more serious. Looking at the bright side of things, the Zoners could finally get a life.

354 comments:

1 – 200 of 354   Newer›   Newest»
SPARKS said...

http://www.marketwatch.com/News/Story/amd-rejects-intels-legal-argument/story.aspx?guid={5DD08467-EE60-4CA6-8664-59E1E6BBDDA1}&siteid=msn


It gets worse. AMD is rejecting INTC's licensing claim. Apparently, INTC's lawyers have been busy.

"Thursday said it is rejecting rival Intel Corp.'s ................legal argument that the creation of The Foundry Company and its acquisition of ATI Technologies breached some provisions in cross-licensing agreements between the two giants. Intel had asked for a meeting in a Jan. letter, according to an AMD filing with the Securities and Exchange Commission. AMD Spokesman Drew Prairie said "the letter is another attempt by our competitor to cause uncertainty as we approach our Asset Smart deal closing next month."

The Scrappy Little company is gonna get it's teeth kicked in, at the worst possible moment.

SPARKS

Anonymous said...

Heh, I wonder how long AMDZone will stay up after AMD bites the big one.

Sparks: The Scrappy Little company is gonna get it's teeth kicked in, at the worst possible moment.

I haven't seen any legal news in a long time, and had thought maybe AMD and Intel reached some sort of deal like was discussed here before - Intel not pressing the x86 license restrictions for AMD not pressing any antitrust lawsuits. But, I guess AMD is not so smart after all. I just wonder how much this will cost them, and if Intel asks for an injunction it could drag out for years. Unfortunately without the Abu Dhabi money, I think AMD only has a quarter or two left to live, not years...

Tonus said...

Sparks, I suspect that after some PR tough talk, the two companies will sit down and hammer out some sort of deal. That may not bode too well for AMD, but I don't think they'll spend much time in a courthouse. Of course, I could be wrong.

Nonny Moose: "I had thought AMD's revised guidance said something like 20% down - turns out it was more like 33% down."

As I recall, AMD had revised guidance down 25% from the previous quarter. Combined with the increase that would normally be expected in Q4, the guidance would have been around 35%.

I guess the good news is that they were really good at estimating their Q4 numbers. :p

SPARKS said...

"Sparks, I suspect that after some PR tough talk, the two companies will sit down and hammer out some sort of deal."

GURU, with his magical crystal ball, coupled with his keen sense of clairvoyance, predicted the possibility of such a deal (with AMD's monopoly suit on the table), a year ago.

SPARKS

SPARKS said...

"I just wonder how much this will cost them, and if Intel asks for an injunction it could drag out for years."

I think the phrase "litigate them into extinction" might be appropriate.

SPARKS

Anonymous said...

Tick Tock Tick Tock

What a joke, AMD is finished

New generations cost billions to develop, billions to design and validate new.

AMd couldn't make money in good times, wasted billions on ATI, spread their legs to the Arabs.

Now they got to pay the piper they are finished.

They got crappy products, crappy management, crappy process compared to INTEL

Finished, BK

All the fluffing from Scientia, Sharkiou and other silly folks can't change the fact they got no money, got no technology, got no products.

BK BK BK... Poor Jerry, sorry ass Hector, I feel for the engineers.

But they are finished, done, end of story

Anonymous said...

Actually the combination of Hector, poorly yielding process technology from IBM (do you expect anything better really from IBM?), and the current recession will prove to be too much for AMD to survive.

Anonymous said...

As I recall, AMD had revised guidance down 25% from the previous quarter. Combined with the increase that would normally be expected in Q4, the guidance would have been around 35%.

Again, AMD guided flat from Q3 to Q4 (and excluded the Q3 artificial 191Mil licensing revenue) - so guiding 25% down would mean ... 25% down. (Don't throw in seasonal as AMD guided flat)

Crunching some #'s the net revenue went from 1.8Bil to 1.16Bil which is the 37% people are reporting. If you take out the artificial 191Mil one time revenue in Q3 out the "real" revenue went from 1.6Bil to 1.16Bil which is down ~27% (not too far off from what AMD guided).

It is good to see the analysts punishing AMD - they tried to play games with the 191Mil in Q3 to get favorable revenue comparisons and actually prop up the gross margin (I think it added 8% to the gross margin). Now, even though AMD's guidance was excluding the Q3 license revenue, the analysts are holding them to the #'s AMD CHOSE TO REPORT IN Q3 (i.e with the Q3 191Mil license revenue thrown in as "operating" revenue).

Overall the #'s aren't too bad (I chose to exclude the 191mil fee when looking at both the Q3 and now the Q4 #'s). The concerning thing however is the balance sheet. Compared to last year (Q4'07):
- Cash on hand has gone from 1.9Bil to 1.1Bil
- Accounts receivable has gone from 640Mil to 320Mil
- That (and some other junk) has taken current Assets from 4.6Bil to 2.4Bil...
- Total assets have gone from 11.5Bil to 7.6Bil

Now the asset #'s alone wouldn't be that bad if it was due to paying down liabilities, but the concerning thing is that current liabilities have only gone down 2.6Bil to 2.2Bil in that timeframe!

http://finance.yahoo.com/news/AMD-Reports-Fourth-Quarter-bw-14131797.html

Anonymous said...

Agreed IBM process is crap and they are falling further and further behind.

Can't wait to see IBM's Ion/Ioff they will likely never publish it as it is sorry

Anonymous said...

Actually I think now is the time for the lawsuit settlement (I previously proposed). While some may say why bother if AMD is heading toward bankruptcy - well Intel still has to face the EU and other governments seeking payola to offset their growing deficits - and this will happen regardless of whether AMD bankrupts or not.

If Intel can get AMD to drop the lawsuit, that would help them fight some of the various government claims ('hey our own competitor dropped the lawsuit') , and potentially prevent the NY cronies that we all know Sparks voted for :), from looking for another pelt for their walls on the way to their run for Federal offices (read: Cuomo). In return Intel would take out the outsourcing restriction in the x86 license - this would take away any potential issues Intel may have with the foundry split, and maybe Intel throws AMD a little cash (~300-500Mil), but admits no wrongdoing.

The other benefit for AMD is that it gives them some short term flexibility as they can essentially hang the foundry out to dry and truly force them to compete with other foundries, as if they fail they would be free to outsource to Chartered or TSMC... Right now if the foundry fails, AMD is done as they can't outsource CPU production. So even if they win this current spat with Intel about whether or not the foundry violates the x86 license, they are still attached at the hip to the foundry and are dependent on the foundry's survival.

Intel would also get some PR out of this as instead of stepping on AMD's throat and going for the killshot (which it seems like they could easily do if they wanted at this point with another round of price cuts), they can say they threw a lifeline by eliminating the outsourcing restrictions.

So yes I stand by the prediction - I have to go look back but I think I said sometime mid-09?

SPARKS said...

"I think I said sometime mid-09"

Yes, you certainly did.

We don't vote here. Either they get thrown out of office, especially on Long Island, or they move on to bigger and better things.

Let's not forget photo ops in a cathouse. NYS Government gives a new meaning to 'the blind leading the blind'.

SPARKS

Anonymous said...

Why should INTEL give charity to a company run by executives who focused on how intel hurt them versus focus on executing and creating great products?

If a company has a broken business plan why should they be bailed out?

I do agree AMD is finished, regardless of whether they win or lose the lawsuite, regardles of whether INTEL loses the EU investigation and is fined nothing or a billion.

AMD simply doesn't have to tools to design a complext CPU, develop the process to manufacture it and invest in the billions to manufacture it. Why anyone thinks they can simply doesn't understand the business and gets lost in silly benchmarks

SPARKS said...

"I haven't seen any legal news in a long time."

Moose, When Abdulie the deal was first anounced a few months back, INTC's top legel bulldog Mr. Molloy said 'this is something we will have to look into'. The link escapes me, presently.(I'm too lazy to dig)

He rarely talks. What looks like a small comment on the surface to most; with this guy, he's probably under your bed.

SPARKS

SPARKS said...

"Why anyone thinks they can simply doesn't understand the business"

"Why should INTEL give charity to a company run by executives who focused on how intel hurt them."

Listen, guy, I don't the difference between strained silicon and strain relief, but I do know what the gangsters in the EU could do, and what the New York State Attorney General could do, hurt INTC.

They both have a few things in common, AMD, Intel, money, and political agendas. This is gospel.

SPARKS

Anonymous said...

If a company has a broken business plan why should they be bailed out?

Who said bail out - Intel paying some money to settle this thing out of court comes down to risk management... while some may view the chance of AMD surviving until the suit gets to court, or the actually chance of winning the case (and appeal) as small - it is about expected value.

You also dismiss the legal fees Intel is incurring/will incur to defend itself - that will easily be tens of millions if not 100+ million when all is said and done. Also this could have some influence on the governmental lawsuits and any positive impact on that also represents positive potential return for Intel in settling with AMD. (or more importantly think about the gold rush from governments around the world should AMD win this case)

You're reasoning is strikingly similar to what I think is AMD's weak point. It isn't always about winning/losing or doing what is theoretically right; it's about making a sound business decision. So while the chances of Intel losing a case and appeal may be small, when you consider legal costs, the extremely large costs if they do lose, impact to other cases and other intangibles like the distraction to Sr management, moral, etc... a 500Mil or less "investment" in settling the case may actually be in Intel's BEST interest too... it's not about charity.

Settlement value <> Legal cost + (prob of losing)*(judgement value) + any other costs

So do a simple expected value calculation. Even if the probability of AMD winning is low (let's say 10%), when you multiply that by potential damages (these are generally tripled the actual impact... 3+Bil?) that would be 300Mil.. throw in a 100Mil in legal costs and you should be able to see how 500Mil is suddenly no longer unreasonable? (maybe?)

Anonymous said...

Sparks: Moose, When Abdulie the deal was first anounced a few months back, INTC's top legel bulldog Mr. Molloy said 'this is something we will have to look into'.

Yep, that's what I read at the time too (October? November?) but nothing since then. So I thought AMD and Intel were making some sort of deal under the table, the kind where we don't see any announcement until it's dealed and sealed.

Anyway, I'm puzzled by the "total liabilities" being just 2.2B - doesn't the debt they racked up with the ATI purchase get entered into that sum?

Tonus said...
This comment has been removed by the author.
Tonus said...

"Why should INTEL give charity to a company run by executives who focused on how intel hurt them versus focus on executing and creating great products?"

There are those who believe that Intel must prop up AMD, lest they be declared a monopoly and broken up. :)

SPARKS said...

Yeah, INTC is also checking into the ATI deal, too.

http://www.theinquirer.net/inquirer/news/596/1050596/moody-amd-lashes-intel


It does get worse, The Scrappy Little Company knows it's gonna get it's teeth kicked in.

Now they're crying to the media. However, the 'flavor' of the (former AMD pimping) INQ report is less than conciliatory.

AMD's public whining appeals makes me yearn for INTC just to crush the slimey little cockroaches where they stand, today. (Very stupid, I know.)

They keep their deals confidential, yet they cry publicly when challenged.

SPARKS

Anonymous said...

Anyway, I'm puzzled by the "total liabilities" being just 2.2B - doesn't the debt they racked up with the ATI purchase get entered into that sum?

Sorry, I was referring to current liabilities (the debt would be long term liabilities). It wasn't quite clear in the post.

Anonymous said...

From the conference call (AMD's Rivet):

And then as time goes on and capital calls are required to fund that business to make the large capital investments to build out different clusters, we will, as we've said in the past, most likely not participate in those capital calls and our ownership will decline as time goes on. Through a combination of all those things, it will become just an investment by AMD and then eventually it just kind of winds its way out as this time goes on. (referring to foundry)

AMD is betting the farm on the removal of the outsourcing restriction in the cross license patent when it comes up for negotiation in 2010. If it is not removed AMD will be forced to continue to input capital as their stake now is only around 35%.

I'm still stunned as to how this quickly and easily this deal got through government approvals. I'm not saying it shouldn't be approved but we are talking about a transfer of chipmaking IP to the middle east, while we restrict BASIC semiconductor IP to other parts of the world. I would think this would get a harder look then it did. And it is not a matter of the just the middle east - if it is going to any foreign country it should be vetted more vigorously.

I'm concerned that the regulators bought hook line and sinker into thinking this is just a partnership, when the quote above from Rivet clearly shows there is no partnership intent.

SPARKS said...

"I'm still stunned as to how this quickly and easily this deal got through government approvals. I'm not saying it shouldn't be approved but we are talking about a transfer of chipmaking IP to the middle east, while we restrict BASIC semiconductor IP to other parts of the world."

Why should this surprise you? They've sold out America (and it's workers) so many times before, why should this be an exception?

We INVENTED TV! There hasn't been one built in this country for decades. Let's not even talk about steel.

Remember your political correctness! We can't have America be perceived as a dominant force globally. Hence the EU attacks on Microsoft and Intel, and the assholes in this country (fellow Americans)that side with them.

Disgusting, stupid bastards like Share-a-Kook and Dementia will unwittingly sell out America every time. They have by default, when they put a positive spin on this deal, dispicable.

Further, I am a taxpayer in a state who's funding 1B+ for the deal, with an IDIOT liberal bastard like Andrew Coumo who's greasing it in. Give it away to the ARAB's and fuck the last remaining stonghold/company we have on a world market.

You can't make this shit up!!!

RANT OFF!!!! GODDAMMIT!!

SPARKS

SPARKS said...

By the way, I can't tell you how proud I am of Intel Corporation. The greatest (American) company in the world.

Sparks

Anonymous said...

Sorry, I was referring to current liabilities (the debt would be long term liabilities). It wasn't quite clear in the post.

OK, now I see. I wonder how much of the ATI purchase still remains as debt for AMD to repay now?

Anyway I took a look at the report posted on amd.com and it seems that AMD lost about $4B in total assets while only reducing liability by about $400M, or 10% of the asset loss. If the Abu Dhabi deal gets canceled or delayed, then I'd think AMD stands a good chance of going bankrupt this fiscal year, certainly early next FY, if everything else stays the same.

Anonymous said...

end of Q3'08
long term liabilities = 4.78B

end of Q4 '08
long term liabilities = 4.70B

http://biz.yahoo.com/bw/090122/20090122006129.html?.v=1

Anonymous said...

long term debt
end of 2005 -- 1.33B
end of 2006 -- 3.67B

So lets be generous and call 1.5B of AMD's current long term debt "non-ATI"

That means >3.0B to go.

Anonymous said...

IIRC, ATI actually made AMD a profit of about $50M in Q3. So their investment of $5.4B has left them with only $5.35B left to go, before ATI becomes a profitable investment??

Anonymous said...

IIRC, ATI actually made AMD a profit of about $50M in Q3. So their investment of $5.4B has left them with only $5.35B left to go, before ATI becomes a profitable investment??

Oh.... it's worse than that as AMD is burning money just to pay interest on the debt taken on. So for ATI to become a "profitable" investment (>$0 ROI) it needs to make up the debt AND the interest on the debt.

In any event, the spinoff was a way to load all of the debt on the foundry so most of it will soon be on the foundry's books - which inevitably will cause the foundry to fail or require a lot more cash injection from the Paulas(TM). AMD is just betting that they will be out of the outsourcing restriction on the x86 license by the time the foundry takes on too much water.

It's actually not a bad racket if you can get away with it... acquire a company and take on huge debt to do it - then spin the debt off into a new company that will effectively be crippled without cash infusion. Enter in someone who wants to buy technology (the Middle East) and an unwitting partner who'll sweeten the deal without knowing the implications (the 1.2Bil transfer of NY taxpayer money to the Middle East courtesy of the NY gov't) and you have what will likely shape up to be a nice documentary 5 years from now (along the lines of Enron: The smartest guys in the room).

Anonymous said...

with an IDIOT liberal bastard like Andrew Coumo who's greasing it in.

You have the wrong idiot liberal bastard (though I can see the mistake given how many there are in NY gov't). Cuomo is Attorney general, no? He's the idiot looking to milk money out of companies in hopes that he makes a name for himself so he can become Senator (or Pres?) so he can then be "qualified" to give taxpayer money away like the 1.2Bil.

I think you mean Schumer, and to a lesser extent the carpetbagger Senator / now Secretary of State for this, and in all likelihood the NY state legislators whose corruption is probably second only to the great state of Illinois. You know Schumer, he's the one who complains about US companies shipping jobs overseas and getting away with tax loopholes, yet he has no problem giving money to a company INCORPORATED IN THE CAYMAN ISLANDS (the AMD foundry company) to help avoid some US taxes! (As he mumbles "I'm not looking", "jobs for NY state", "I'm not looking")

SPARKS said...

"You have the wrong idiot liberal bastard (though I can see the mistake given how many there are in NY gov't)."

...and how many have their fingers in this pie.

Coumo is no exception, the difference being, his reward for greasing the deal is political exposure and power, and of course, a future Governor seat (like dear old dad) or better. More importantly, he's a favored member of liberal democrat 'good ole boy club,' if you will.

Shoemaker, is a smooth, low profile, well spoken, consummate politician. He only surfaces when a mandatory comment is dictated by his staff. Other that that, I believe the only other significant contribution he lead, besides the NYS funding an ARAB held AMD, is the 4B debt the NYS taxpayers are saddled with, presently.

And, this is from a Jew where Israel's largest employer is INTC! Plus, he's indirectly selling out INTC's IP! Please, slap the shit out of me and tell me I'm not dreaming this up.

Oh, he recently bought us off with a property tax rebate check, ~ 8% or about 2 months supply (or less) of fuel oil here in the Northeast. Nice, heh?

SPARKS

Anonymous said...

Innovative companies that invest smartly deserve to reap the profits of their investments and hard work.

Companies with broken business plans, incompetent managers who produce crap products on crappy technology then ask for goverments to give breaks or throttle the competition should be allowed to fail

AMD is like to do do bird and should die

Anonymous said...

If AMD ever goes down, I just hope for "Roborat and gang" to finally buy their celerons at $300 a piece.
Let's not even talk about entry level or High performance pricing. :)

Everything for Intel, right pals? ;)

Anonymous said...

If AMD ever goes down, I just hope for "Roborat and gang" to finally buy their celerons at $300 a piece.

I see this same old argument from the AMD fanbois all the time. Look, Intel is not stupid. They are not going to kill the PC industry with ridiculous pricing, not to mention incur the government's wrath with such a stupid move. Prices might rise to some degree but probably 95% of the average public could switch to netbooks or some equally low-cost solution for most of their needs - email and web-browsing.

And let's not forget AMD charging extremely high prices for the K8s when they came out, up until the time Intel released the Core2 chips. By the way, did you write your thank-you letter to Intel for that, yet??

SPARKS said...

Hey "PAL" let me give you a little clue. I don't waste my time with garbage, be it Celery, Pheromone, Triple Cripple, none of it. I go for the big stuff the top bin stuff, the top of the pyramid, the King of the hill.

I spent $1500 on my QX9770, that's right FIFTEEN HUNDRED DOLLARS and it been pissing all over AMD's "value" chips big time. There's something less fortunate folks need to understand about having money and getting the best, you pay.

What does that mean to simple mind? Well, I'll tell you. If Intel doesn't come up with a new chip to make this one look STUPID slow, I won't buy it. Intel doesn't get the money, get it?

i965 is an excellent chip, a bargain by MY STANDARDS, but it doesn't make my 4 GIG badboy look stupid, so I didn't BUY it. Intel didn't get the money. So while you were playing pocket pool with your budget chips for God knows how long, I've had the B-E-S-T for nine months, and still going strong, and only out done by one other chip on the planet (@ 4GIG). So, if something does come along worth the purchase I WILL BUY IT. I don't care what it costs. Intel will get the money. GET IT?


If prices are too high MOST folks will just stay with what they have. What does that mean to the simple mind? It means the market will dictate price, whether AMD is dead or alive. (better off dead than to the ARABS with INTEL'S License) Intel is in the business to sell chips.

PRICES NEED TO BE ATTRACTIVE TO EVERYONE, TO LUNATICS LIKE ME, AND TO THE MAINSTREAM MOM AND POP USERS,

Like you,....... pal.

SPARKS

SPARKS said...

"Roborat and gang"

I like it!


QX9770-$1500 the last of the GREAT FSB chips.

Core i7 965- $1000 The T-REX of the CPU WORLD, and a new beginning of THE ultimate in performance.

HOO YA!


SPARKS

Anonymous said...

Noony Moose wrote: I see this same old argument from the AMD fanbois all the time. Look, Intel is not stupid. They are not going to kill the PC industry with ridiculous pricing, not to mention incur the government's wrath with such a stupid move.

So, according to you, it was a stupid move when intel charged more than $500-1000 a piece for 486 processors? I wondered what happened to the government in that time. :D

Sparks wrote: Hey "PAL" let me give you a little clue. I don't waste my time with garbage, be it Celery, Pheromone, Triple Cripple, none of it. I go for the big stuff the top bin stuff, the top of the pyramid, the King of the hill.

I spent $1500 on my QX9770, that's right FIFTEEN HUNDRED DOLLARS and it been pissing all over AMD's "value" chips big time. There's something less fortunate folks need to understand about having money and getting the best, you pay.


I feel your pain. How can people be so stupid when it comes to buying hardware?
Thankfully, 99% of the world population is not stupid or retarded at all when it comes to crucial purchasing decissions. ;)

Any how, I won't be wasting my time with teenagers.
I thought Sharikou's blog was a joke but now, a few posts in this blog changed my point of view.

Take care kids. Don't let momma find you on the internet on bed time. :D

Tonus said...

anonDroid: "Thankfully, 99% of the world population is not stupid or retarded at all when it comes to crucial purchasing decissions. ;)"

I find it interesting that you call Sparks stupid, and then you reiterate the exact point that he was making. Eg, if 99% of the world is not willing to pay a premium for CPU power, then Intel will have no choice but to offer models in their price range, AMD or no AMD.

This is not the same market that existed when Intel was selling 486s. They will be able to increase their ASPs and gross margins, but they cannot simply go back to a time when all of their chips sold for a premium, because they would suffer a sales slump that will make the current slowdown look mild.

PS- When you make such ignorant and silly comments while also attempting to insult the intelligence of others? That's kind of embarrassing. Just thought you might want to know.

PPS- I'm considering starting a pool to see how many times you tell us you're leaving before you actually leave. Over/Under is presently at three.

SPARKS said...

"I feel your pain. How can people be so stupid when it comes to buying hardware?"

Perhaps your're referring to the millions of dollars the channel lost when AMD abruptly killed socket 939 that everyone wanted, to socket 940 the channel couldn't give away?

How about all the hype surrounding Barcelona before it's release, despite the horrible performance, AMD fans still bought into a broken product, hoping to recapture the glory of the 2005 K8.

How about the poor guys who were misled into thinking the 'Quad Father' FX-74 was going to the ultimate Intel killer, only to find it was power sucking miserable failure, that couldn't even best a QX6700 @ stock speeds? (QX6700 will run on ANY X38 x48 motherboard today) Not only did they waste their money on a slower processor (s), they bought into a dead end platform that EOL'd a few months later. I'm sure AMD fans were happy about that. Sorry, in reality, the market spoke, no one bought them, and they simply died.

http://blogs.zdnet.com/Ou/?p=382

Look, junior, there's a reason why AMD has lost 6B in 2 years. Let's look at that number if your to myopic to understand its significance, $6,000,000,000. There's no conspiracy theory, AMD doesn't have competitive product in the market place.

The market has spoken, whether you spend $168 on a Q6600, $285 on a E8600, $315 on a Q9550, $300 on a i920, or $1000 on a i965. The choice is the buyers' and he/she will not be disappointed, for their individual applications.

I've made my choice, what yours? Or, are you too ashamed to say?

SPARKS

Anonymous said...

Guys - why do you bother feeding the trolls.

Intel's growth at this point is almost independent of AMD as there is not much market share left to take. For Intel to grow they need to grow volume or develop new markets (which I think was/is Atom's target).

If AMD evaporated Intel would grow by the remaining market share... let's say AMD has ~20% MS, so that would be a one time ~25 growth for Intel. Then what? The only way to grow further is to grow the market and clearly raising prices by an extreme amount and slowing innovation (and thus giving people less reason to upgrade existing computers) will not accomplish this.

These re-tread "we need AMD for the good of the industry" arguments are given by people with no background in business, economics or rationale thought. While it is nice to have a competitor, I'd argue that a weakened competitor (like AMD right now) is actually WORSE than no competitor as even in their weakened state they will crowd out any new (and potentially viable) competitors. Yes you will see a sort term price adjustment as others have mentioned; but the market would soon punish Intel if the prices were too high (through slower growth, slower corporate upgrade cycles, etc...).

I'd also argue that the people advocating breaking Intel up are also missing the boat. A split up Intel (which in theory "creates" competitive environment) would destroy the economies of scale needed to sustain development and enable manufacturing (as evident by AMD's spinoff to a fabless company).

I also enjoy the argument of Intel killing the computer industry by releasing cheap chips. INTEL DOES NOT SET THE PRICES FOR THE FINISHED COMPUTER, NOR DOES IT SET THE MARGINS PC BUILDERS ESTABLISH. If they buy a $100 chip they can price the rest of the computer (and pass on any additional margin on the chip) however they want. Would a cheap engine kill the automobile industry? Would cheap glass kill the TV industry? Lowering an input component cost does not kill the end product industry. How do other commodity businesses survive? The problem is the computer industry has been a value based industry (pricing by perceived value) and folks have not adjusted to the commodity movement in the industry.

SPARKS said...

"Guys - why do you bother feeding the trolls."


"G" You're right. Hell, you're always right. But, then again, it keeps me on my toes, I get to do a little history review (to sharpen my 80's affected memory), and finally (most importantly) I relish your excellent technical prose/analysis, anchored with unabated logic and fact that clearly and succinctly puts the entire issue to rest.

Besides, we're Robo's "gang"!!!

Thanks!

SPARKS

Tonus said...

guru: "Guys - why do you bother feeding the trolls."

Well, I've got this mallet and some time to kill... :)

"I'd also argue that the people advocating breaking Intel up are also missing the boat."

The arguments I've seen for breaking up Intel are made with little or no understanding of economic markets or US law in regards to monopolies. As you implied, the government is not going to harm the technology sector simply to break up a company who suffers from a case of poorly-managed competition. If the time comes when you can break up a company like Intel with no harm done to the industry, I wouldn't be surprised to see it.

In any case, I see some people who state that Intel is breaking the law by pricing its products lower than AMD in order to force AMD to compete at lower pricing levels. This is only illegal if Intel was selling these items at a loss or at an unsustainable margin, hoping that it could weather the losses longer than AMD. If Intel is able to undercut AMD and still make money on those products, then the problem isn't Intel... it's AMD's inability to make a competitive product.

SPARKS said...

Tonus-

You've got a point.

Additionally, I don't think Uncle Sam is going to step in and destroy one the few remaining PROFITABLE enterprises left in this country, especially given the current worldwide economic circumstances. Plus, Intel employs an awful lot of people, and pays an awful lot of taxes.

Further, the ripple effect though out the entire industry would be catastrophic worldwide. I think of the infrastructure, like small and large vendors, customers, suppliers, the channel, etc., that would be undoubtedly disrupted by such a move, especially now as the government is bailing out companies in nearly every major sector.

With today's economic calamity, I think standard Government policy is 'laissez-faire' with regard to healthy companies. If something works you simply don't fix it, especially when the walls are caving in around you. Not now, anyway.

First the world banking, now the disrupt the world leader in semi's? The whole idea from 'anonDroid' (and those like him) is ridiculous. For the government to even TALK about such nonsense is, well, I apologize for putting this way, but it would be just plain fucking stupid.

(And "G", they're stupid, I'll admit, but they ain't THAT stupid!)

SPARKS

Anonymous said...

So, according to you, it was a stupid move when intel charged more than $500-1000 a piece for 486 processors? I wondered what happened to the government in that time. :D

Umm, in case you forgot, your original post said "celerons
for $300 apiece". Get that? CELERONS! for $300 apiece! Sound familiar yet?

That was what my post was in response to. Nobody but an AMD fanbois would think we wouldn't notice you trying to shift the subject.

At the time the 486 was introduced, guess what - it was the high-end chip. Intel was still making 386s, probably some 286s and I dunno - maybe 8088s. That would be the "Celeron" value CPU and no, Intel didn't charge $300 apiece for them after the 486 was released.

Perhaps you didn't understand your own post too well. I suggest you go back and read it some more.

SPARKS said...

Ah, Moose, the venerable i486DX. It brings back such fond memories. When coupled with WIN 3.1, it changed everything.

I miss writing my own autoexec.bat and config.sys files on my 33 MHz machine. I had a nifty little file utility called X-Tree Gold. It made the job so much easier. In fact, after DOS was loaded, it was the first thing on the drive, an Seagate ST-101, 100MBs!

I remember running DOOM freeware, the first great FPS. I was like a drug someone gets addicted to. Setting DOOM for play required your drivers, both music and 'FX' sounds, be individually setup and compatible with your sound card drivers.

Man, I loved that chip. I have a special place in my curio collection for it. Ah, the good old days.

SPARKS

Anonymous said...

Sparks: I remember running DOOM freeware, the first great FPS. I was like a drug someone gets addicted to. Setting DOOM for play required your drivers, both music and 'FX' sounds, be individually setup and compatible with your sound card drivers.

Heh, I was more into the Dungeons & Dragons genre, first with Dungeon Master on the Atari 520ST and then with the Eye of the Beholder series on my 486 homebuild. But, I played both Doom and its predecessor, Wolfenstein 3D. Hated those damn Nazi bosses :).

My homebuild had a 486-33, a generic VGA card, a Proaudio Spectrum 16 sound card that came with about 20 MOD music files, a 200 or maybe 250 meg HD, 2400 baud modem later upgraded to a 28.8K, and I forget how much DRAM :).

After I replaced it with a Pentium-60 homebuild, I gave it to my Dad as his first computer. He really got into computers at that point - found a bunch of his WW2 buddies to email and ICQ, played Flight Simulator (he went through test pilot training at NAS Patuxent River), even wrote his memoirs with it, complete with scanned and fixed pix of things like his big date with the movie star Betty Grable before shipping off to his squadron in Hawaii :). I was really surprised at his interest and use of it - my mother told me it was probably the best gift anybody had ever given him.

Unfortunately, it fell to me to help him learn how to use it. One time he got some Win3.1 error message, telling him to hit the ESC key to continue. So he calls me up long distance and our conversation goes like "Dad! - just hit the damn escape key - it's located at the top left corner!" and his reply "I don't see it" and my "Just look at the top row, and then scan to the far left!". It finally degenerated into stuff like "Well, Dad - it had an ESC key when I gave it to you. Did you drop it on the floor or something?"

This went on for half an hour until finally I said "Dad - you ARE looking at the keyboard, right??". Dead silence for a minute, then I hear a "Ohhh!" :). He had been looking at the screen for the ESC key the whole time :)

Anonymous said...

This is a bit troublesome:

http://www.eetimes.com/news/semi/showArticle.jhtml;jsessionid=3VDCWL53TA1C4QSNDLPSKH0CJUNN2JVN?articleID=212902342

"Microsoft has a moral obligation to protect these American workers by putting them first during these difficult economic times," Grassley said in the letter, copies of which were provided by his office Friday.

Question 1: A moral obligation? Really - wouldn't morality suggest they keep the best workers? Now if he said a nationalistic obligation...

Question 2: "during these difficult times"? So the obligation exists because times are difficult? If times are not difficult then is there then no moral obligation?

If he wants to control who gets hired/fired he should go work for Micorsoft. If he is worried about H1 Visas... then he should introduce legislation on the # of Visas (which by the way is controlled BY HIS CONGRESS!).

Look I'm all for saving American jobs, but this shouldn't be under the guise of some "moral" code and it shouldn't be done if it is not inline with competency and performance. If it is then this is nothing other than another form of quotas and discrimination. I'd be curious to hear Mr Grassley's take on border policing in the "difficult environment" where there is so much concern about American jobs

Anonymous said...

Looks like our ol' buddy Scientia is patting himself on the back for the "accuracy" of his predictions lately:

However, in 2006 Intel released C2D and my perception became too optimistic for AMD. In other words, going into 2006, Kaa's negative view was much more accurate than mine. In late 2006 and all through 2007 I tended to overestimate AMD and underestimate Intel. It took 2 years for my perception to get back in sync and it seems to be doing better as of late.

I guess he is conveniently forgetting about this statement from his blog back in April: Although Hector Ruiz mentioned the term Asset Smart at the Q1 2008 Earnings Report he avoided explaining what it meant. There has been a lot of speculation in this vacuum but all of it that I have seen has been wrong. Most theories seem to focus on either the idea of selling all or part of a FAB to raise cash or on the idea of making graphic products at Dresden. It is actually something quite different.

Oh how soon the fanbois forget :)...

SPARKS said...

"Look I'm all for saving American jobs, but this shouldn't be under the guise of some '"moral" code and it shouldn't be done if it is not inline with competency and performance. If it is then this is nothing other than another form of quotas and discrimination."

At first glance, with a typical knee jerk reaction, there might be those who would agree with his position. Save American jobs, for Americans. (let's not forget I'm a union guy). However, when you get past the surface, it's ugly scenario, any way you slice it.

It is without a doubt immoral and discriminatory. One must ask, if a course like this is taken, what's next, Hispanics, Blacks, Jews, Irish Americans, Italian Americans? Perhaps, after they address those avenues, they can use secular beliefs as a criteria?

Horseshit, it doesn't fly.(And there may be legal issues.)

I am compelled to agree with you, despite my own deep personal nationalistic sentiment. I would add one word to your, "competency and performance," if I may, and that is 'necessity,' which should be left entirely up to Microsoft.

This is just another very UGLY example of Government evolvement in the business sector.

I'll correct myself, Politicians are that fucking stupid.


SPARKS

hyc said...

Sorry, just had to interject a few random reactions, particularly as the Atari ST was mentioned...

Sparks, your worship of MS and Intel is really incredible. Best companies in the world?

The 486 and all the 808x's prior to it were sick jokes compared to the MC680x0s of the equivalent time.

http://www.skepticfiles.org/cowtext/comput~1/486vs040.htm

Microsoft has retarded the state of the art in OS and application software by 20 years at least. Anyone familiar with the Atari ST or Commodore Amiga can vouch for that.

As financial successes, I'll grant you that MS and Intel have been stellar. But these companies aren't banks - making money isn't supposed to be Job#1 for them. Job#1 is supposed to be bringing great technology to market, and on that score they're both wretched.

re: saving American jobs - mixed opinions on that. It's the government's job to decide who to grant visas to. But after Microsoft made its case and those visas were granted, it doesn't seem right to have them arbitrarily rescinded. Dunno.

Ah, Dungeon Master. I spent so many hours on that game, building up my characters... Few games since then have held my attention the same way.

Anonymous said...

But after Microsoft made its case and those visas were granted, it doesn't seem right to have them arbitrarily rescinded.

Sooooo... it's the governments place to tell MS who to hire and fire. Sorry this ain't government via central planning. Microsoft is a PRIVATE company and has the right to hire/fire employees (within jurisdiction of the US law). For the government to say who this should be or how the distribution is arrogance and discrimination. If the government is worried about American jobs, grant fewer visas and stop taking campaign donations from companies.

Anonymous said...

oops - hyc, I'm not sure I read your comment correctly.

But I think whether MS made a case for more Visas and whether or not the gov't agreed with it is COMPLETELY IRRELEVANT; it is simply not the US gov'ts job to determine who a PRIVATE company can layoff. Suppose MS has determined the least efficient areas of their company are places where there are little/few visas - suppose it was US sales or marketing - would cuts in these areas be "disallowed" if they were coming from divisions which had few H1 visas? Perhaps they would have to cut "non-American" jobs in critical performing areas to offset it? It gets more absurd the more I think about it. Since we all know the government is the role model of efficiency and good management, it is obvious they should be arbiters of American private companies in matters involving jobs losses (or job creation for that matter). What's next - hey MS you can hire 1000 people to expand R&D, but 85% of them have to be American?

I guess we could ask the MS folks to join the UAW union and have people just not show up for work and get paid 85% of their salary for not showing up (the "job bank") - thankfully I think this has been eliminated in the latest round of negotiations with the UAW.

I'm scratching my head trying to figure out why the US auto industry is not ocmpetitive...hmmm could things like that be partly responsible or perhaps things like the ridiculous proportion of car dealers we have in the US?!? No it must be because car companies are not making enough fuel efficient cars. You know these cars are generally where you make really good margins (sarcasm) and where an imbalance in labor and assoicated overhead costs is not that important (more sarcasm). If folks think the car companies were in bad shape... just think what happens as their high margin products (SUV, trucks, luxury cars) continues to dwindle - these companies will be massive money pits. The only thing sustaining these companies over the last 10 years were the inflated margins on these types of vehicles offseting the competitive imbalance on costs with foreign automakers.

This is a case where US lawmakers are using a financial mess to push forward an alternative agenda (in this case environmental one). I have no issue with pushing environmental standards, but for the Congress (and this case specifically the Democrats) to link the carmakers issues with a lack of fuel efficient cars is an at best tenuous (and loose) correlation. It really has nothing to do with this and unfortunately we will figure this out in about 3-5 years when all these "fuel efficient" car manufacturing lines are online and the US automakers will still be a mess with their hands out in order to 'save millions of jobs' (again). Folks seem to ignore the fact that the carmakers were heading downhill and downsizing well before the economic issues and $4/gal gas... these things just highlighted and accelerated the fundamental flaws in these companies.

/end rant

Tonus said...

I was a huge fan of the Commodore Amiga. True multitasking in 512KB of memory, and graphics and sound capabilities that put just about anything else available at the time to shame. I believe that years after they had stopped manufacturing them, they were still used in many television stations, because an Amiga and a Video Toaster made for a low-cost visual effects station.

It is a reminder that you are correct, hyc-- the best technology doesn't always win out. Companies have to keep the business and marketing aspects in mind, and also perhaps be a bit lucky to prosper. Even Intel has wound up trapped by the x86 monolith. They've sold millions of Itanium processors, but it's still considered a flop by many people, and it has never come close to moving towards the mainstream.

Chuckula said...

Hey Guys!
Well it looks like the notorious Geode has been put on death row. It's hilarious that AMD is actively retreating from the low-power arena and leaving it to Intel... at a time when low-cost chips are becoming more important than ever.
AMD's "replacement" is just an underclocked K8 that uses about 10x more power than the Atom... good luck with that AMD!

hyc said...


oops - hyc, I'm not sure I read your comment correctly.

But I think whether MS made a case for more Visas and whether or not the gov't agreed with it is COMPLETELY IRRELEVANT; it is simply not the US gov'ts job to determine who a PRIVATE company can layoff.


OK, having thought about your posts, you're absolutely right.

On the other hand, the gov't could have *effectively* done the same thing. If he'd been smart, this Senator could have said "hey, we need to protect American jobs, let's revoke all H1 visas granted between thus-and-such dates." At which point, lots of companies would have had to fire people. Singling out Microsoft was stupid...

hyc said...

Note, I'm not saying that revoking visas is right either way.

But it's not a far cry from "we need to crack down on terrorism" and deciding to deport all recently arrived immigrants, etc. They hadn't done that yet, but it looked like the Bush administration was stupid enough to have that on their list..

Anonymous said...

Well it looks like the notorious Geode has been put on death row. It's hilarious that AMD is actively retreating from the low-power arena and leaving it to Intel...

I must say it will be with great amusement I watch to see what the arrogant Negroponte decides to do with the OLPC project after hammering Intel on it. All Intel wanted was an opportunity to sell competitive products (to the $100 laptop) in exchange for participation and Negroponte found that request to be unreasonable.

I guess there's always Via and Nvidia(?). By the time he gets it together their will be FOR PROFIT companies selling $150 (which is where his laptop is) and he will probably be bitter about it (bitter because he is not owning the solution). Karmas a bitch... if only he had been happy about enabling a solution it would be done by now... instead, soon there will be a solution without him...

Anonymous said...

hyc: The 486 and all the 808x's prior to it were sick jokes compared to the MC680x0s of the equivalent time.

I don't think the 486 was shabby at all compared to the 68000 the original Atari 520ST used, although since my 486 had a hard drive and the Atari only a floppy, a direct comparison was difficult esp. when going up or down the stairs since the Atari had to read the new level into memory from the floppy. A friend of mine who was really into Ataris at the time, built a 486 for himself shortly after seeing mine. As a joke, he benched the 486 and his Atari 1040ST running a DOS emulator. The 486 scored something like 25x an equivalent 8088 and the Atari 0.1x the same 8088 equivalent :). It's been so long now I forget the name of the benchmark, but it was the one commonly used at computer shows since it displayed big bar graphs on the screen - probably just measured how many NOPS it could run through a counter in 10 seconds or some such silly thing.

Ah, Dungeon Master. I spent so many hours on that game, building up my characters...

DM was the first and only game to make my hair stand up :). I used to stay late at night in my friend's office after work playing it, since he had brought in one of his Ataris rather than put up with the typing pool mistakes. Pretty spooky in a dark building at 2AM. After a week of me telling my wife that I was "working late" I decided it was cheaper to buy my own Atari than go through divorce court :).

I got a character up to level 4 and level 5 Master ratings in all 4 classes once - could go toe-to-toe with Chaos for hours at a time, with the appropriate protection spells, although Chaos was immortal of course.

SPARKS said...

"Sparks, your worship of MS and Intel is really incredible. Best companies in the world?"

"The 486 and all the 808x's prior to it were sick jokes compared to the MC680x0s of the equivalent time."

Who cares? Motorola can't even compete with cell phones, now.

Let's get one thing straight, I don't worship Microsoft, I WORSHIP INTEL, got it? All my retired chips are on display and are like family members. Some guys have A-Rod and Michael Jordan as hero's, I have Mark T. Bohr.

I see you are still fixated on criticizing the two world leaders of hardware and software that are both technologically and financially successful worldwide. It's amazing how you, with your "High Brow" analysis, render both companies nearly incompetent in design and execution. In typical fashion, your comments are always focused on the negative aspects of the way they do business, or some vague esoteric technical commentary about hardware/software inefficiencies.

When you purchase a donut, do you complain about the hole? Do us all a favor, look at the donut, not the hole, take a bite, and enjoy. Are there pea's under your mattress. Do they bother you?

These two companies hold 80% of their relative markets worldwide. They have brought information technology to the mainstream on this entire planet together on a relatively stable and cheap platform in only twenty or thirty years.

Do you think Microsoft should open source it's O.S, or INTC should open source X86 to, say, the WORLD? Maybe they should give it away to the Arabs?

No, the point I was trying to make, if it slipped past that keen analytical mind, is that this country should PROTECT INTC's and M.S. intellectual assets, vehemently.

These are great American Companies that are dominating the world market and are the envy the world over. Of course, they do have their petty, pain in the ass critics and detractors.

Or perhaps YOU with all your brilliance have a better idea on both hardware and software that will revolutionize computing on this planet as we know it. You're a capable guy, design some hardware and some code and show them how it done. Who knows maybe you could tap into this multibillion dollar market.


Oh yeah, I ABSOLUTELY ADORE INTEL CORPORATION.

SPARKS

InTheKnow said...

hyc said,

As financial successes, I'll grant you that MS and Intel have been stellar. But these companies aren't banks - making money isn't supposed to be Job#1 for them. Job#1 is supposed to be bringing great technology to market, and on that score they're both wretched.

I can't tell you how far off base I think you are here.

These companies exist for only one reason. To make money for their investors. They were not labors of love, nor were they ever intended to be. Investors don't contribute money out of the kindness of their hearts, or for the love of science. They risk their money in an expectation of making more.

"Bringing great technology to market" is a means to an end, not the end itself.

Right or wrong, good or bad that is the way the world works.

I'm also amused by a couple of other things in your post. The first is the article you linked to. It showed that several of the design decisions of the Intel 486 were superior to the Motorola chip. So who had the better technology?

Look at the benchmarks you say? Well that is the other amusing point. If I were to go to, oh I don't know, let's say AMDzone and say look at the benchmarks, I'd get beaten within an inch of my life. They are biased! They are skewed in Intel's favor! They don't represent real workloads!

But you bring that stuff here and no one screams foul. Funny that you can do that here where we are all supposed to be Intel fanbois, but you can't do it on an AMD centric site. Perhaps there is a bit more willingness to look at the data here before rushing to judgment?

I would also point to one common denominator that Intel's competitors have shared. Hector Ruiz. He was at Motorola when they melted down and now he has been at AMD and they are melting down. Your Motorola chip example only demonstrates that there are other factors at work than apparent technological superiority. Why blame Intel for someone else's work on the inside?

Perhaps you should ask the real question. Why didn't Moto's chip succeed if it was better. There were plenty of IBM clones in the early days. If these had been that much better, they would have won in the market place. Intel was not the 800 lb gorilla back then that it is now. But they have succeeded while others haven't.

I would submit that this is because they understand that the best technology is only part of the equation.

hyc said...

Been there and done that. When Space Shuttles launch and find their way home without getting lost, that's 'cause they're running my code. After my first launch I kinda got bored so I left...

I was an Atari developer 'way back when; helped write their multi-tasking kernel. We had real virtual memory, memory protection, multiple priorities (including realtime) and POSIX compatibility back in 1988 when Windows was still just an ugly shell over DOS.

I'm not looking at your donut hole, I'm looking at the cheese Danish that I had... I think the beginning of the end was the whole RISC fad, when all the workstation vendors dropped the 68020. And yes, anybody who thought seriously about it would have realized what a stupid fad it was; long sequences of large fixed-size instructions overwhelming the small on-chip I-caches of the day.

As for MS opening up their IP - I'm all for it. I'd bet that at least 40% of what they sell today was stolen anyway, and it would be nice to see them come clean.

re: Intel's IP - I frankly have no desire to see any of it. I'd prefer if they just went away and took x86 with them. The two together delayed the transition to pure 32-bit architectures by at least 10 years, and Intel kept trying to delay 64-bit until AMD forced their hand.

Hint: Information technology was *already* spreading happily and healthily across the planet without their influence, thank you very much. Remember, the Commodore Vic-20 held the record for most sales of any computer for more than 2 decades. Everything Microsoft has done has only *slowed down* the advancement and deployment of reliable technology... And where do you get the idea that Microsoft has made it cheap for anyone? You seem to be forgetting the days when the Borland Turbo compilers sold for $39 while the Microsoft compilers sold for $399, etc... People don't joke about "Micro$oft" for nothing. Every benefit you claim we've derived from them is in fact what we have clawed out for ourselves *in spite* of them.

hyc said...


"Bringing great technology to market" is a means to an end, not the end itself.

Right or wrong, good or bad that is the way the world works.


I understand that that's an accurate depiction of how things are. But that doesn't make it right, and the fact that things are that way now doesn't mean we should accept that it's the only way things can be.

Hm, several of the 486 design decisions were superior? Please enumerate them... Having written compiler backends for both, and tons of assembly language for both, I obviously think the 680x0 was superior all around, but I'd like to hear your perspective.

Why blame Intel? I mainly blame Microsoft, really. But the two have been inseparable for a long time, even going back to LIM memory standard: Lotus-Intel-Microsoft.


I would submit that this is because they understand that the best technology is only part of the equation.


Again, I'm forced to agree with you, but I don't think it's a foregone conclusion that "that's the way it has always been" means "that's the way it must always be."

As a side note, I founded my current company in 1999, as the dot-bombs were imploding everywhere. I am currently the lead developer on the OpenLDAP Project, and as a result of my work, for the past 5 years it has been the world's fastest directory server, open source or otherwise. Interestingly, when financial times get tough, my company grows faster, as prospective clients re-evaluate their expensive closed source licenses and come looking for our solution.

We don't spend any budget on marketing, we focus purely on the technology. Because of that, we're nowhere near becoming a household name. We will probably never grow to be as big or as "successful" as Micro$oft, even though our software spreads ubiquitously around the world. As long as we're comfortable and growing steadily that's OK with me, because there's more to life than making money.

Ultimately, marketing is convincing people who won't think for themselves and have no way of knowing any better, to buy your stuff. I prefer to work with customers who *do* think for themselves, and are willing to invest the time into teaching themselves to know better. Right away that means we *won't* be selling to 80% of the market...

Anonymous said...

Why blame Intel? I mainly blame Microsoft, really. But the two have been inseparable for a long time, even going back to LIM memory standard: Lotus-Intel-Microsoft.

Well that's really a fine intellectually bankrupt logic - guilt by association, blame one for the other? Makes sense.... (?)

And AGAIN with the 64bit... exactly what has it brought us thus far (spare me the theoretical arguments) what TANGIBLE performance (%improvement, or new tangible capability) has it given the average consumer? Frankly I'm like most of the rest of the consumers around the world am running 2GB memory(or some who run less) and really don't care about going over 4GB for the sake of going over 4GB. It'd be nice if my car company upgraded the engines to go over 200mph on the new models, but I really wouldn't consider that bringing technology to market because it would be pretty much useless for most of the world

It is quite interesting how your argument has evolved from Microsoft/Intel being WRETCHED in bringing technology to market to mainly Microsoft. (but again using your unassailable, clear thinking logic the two are inseparable, so anything bad MS has done is also by default Intel's fault too)

hyc said...

It is quite interesting how your argument has evolved from Microsoft/Intel being WRETCHED in bringing technology to market to mainly Microsoft. (but again using your unassailable, clear thinking logic the two are inseparable, so anything bad MS has done is also by default Intel's fault too)

Oh give me a break. The x86 is still wretched. Just because most of my ire is aimed at M$ doesn't mean the complaints against Intel don't still hold.

We all know the Silicon Valley stories. If Gary Kildall was home the day IBM knocked on his door, there would be no MS-DOS and we would all have watched CP/M evolve. A lot of dumb decisions and dumb luck are what grew the PC architecture, and M$ and Intel rose to the top on that wave, not on their technical merit. Apple made it easier for them, by being stupidly proprietary with their Mac, after having such raging success with their totally open Apple II.

So yes, Sparks, open up that IP. Technology and society advances when innovations are shared openly. The PC ISA bus succeeded because it was open. IBM MCA failed miserably because it was closed. How many more examples do you need to show that rabid IP protection is wrong?

Anonymous said...

There is an interesting article on this messy law called Moore's at EE Times.

http://www.eetimes.com/news/design/showArticle.jhtml?articleID=212902839

So if INTEL wasn't around would we have seen the investments and the fruits of these investments if intel wasn't able to reap the rewards?

Scientia, hello are you out there? What do you think of this?

SPARKS said...

Circa ~1980, thereabouts, Reagan to Carter....
"There you go again."

Aboard the Enterprise, Kirk to Khan.........
"I'm laughing at the superior intellect"

Present day Anonomous (G) to hyc.......
"Again with the 64 bit"

Priceless, I just about fell out my chair.

"The PC ISA bus succeeded because it was open."

Now this cuts into some muscle. The arguement that we would be better off today had INTC and MS opened up their IP is fundementally flawed.

On the contrary, the three components, platform component standardization, coupled with a strictly disciplined CPU design and software design are the cornerstones of why we are WHERE we are today. There was no chance that any of these three major components going off in some industry wild tangent, in some unknown (new and improved?) direction.

In fact, I would argue this lead to the stability and invocation (AND LOW PRICES) that we enjoy to this very day, worldwide. Computing went mainstream to the average Joe because of LIM and ISA standards, not inspite of them. The entire industry took of like shot, thereafter. Everyone knew what to do and where to go!

I don't understand your rational.

SPARKS

Anonymous said...

The x86 is still wretched. Just because most of my ire is aimed at M$ doesn't mean the complaints against Intel don't still hold.

Is this an MS issue, an Intel issue and/or also partly a consumer issue? Consumers expect (demand?) backward compatibility on everything and with all ancient HW. It's hard to redo x86 when you have those constraints.

Would things theoretically be better starting from scratch with no boundary conditions? Sure... but this logic also could be applied to the electrical grid, road, communications, trains, airplanes and virtually every industry.

You choose to see the worst in Intel (and yet AMD gets a pass because they are the underdog and mere victims of Intel's brute force?) and that is fine, it is your own personal opinion and theoretically you may be correct about Intel in terms of design in an ideal world. However in the real world, you just don't get it. If x86 is so wretched isn't AMD also partly to blame for pushing x86 forward onto 64bit as opposed to coming up with a more optimal architecture and using 64bit as a potential inflection point for CPU design? (oh nevermind they are victims, they get a pass for pushing a "wretched" architecture forward dfrom 32bit to 64bit)

And why would a company spend billions on R&D if the IP suddenly becomes open? In this open world who pays for the development? Open Source SW is a bit different as this is far more labor intensive and does not require huge up front investments in capital. This is clearly not the case for Intel (or for that matter AMD as we have seen by the spinoff). Perhaps Intel should also just open up their process flow for the good of mankind to utilize and improve. Then we can have Intel continue to pour billions into it only to help others - I'm surely they'll be highly incentivized to keep Moore's Law churning for everyone.

While we're at it might as well open up the design, critical design rules, modeling and mask making to further the good of technology. After all, having everyone work on a new design is the optimal solution, no? Now we just need to figure out who pays for it... maybe we can collect all the global revenues together and have a government(s) dole it out to the various companies.

A lot of dumb decisions and dumb luck are what grew the PC architecture, and M$ and Intel rose to the top on that wave, not on their technical merit.

This ia an INSANE statement... you can say the opportunity was lucky, but being able to capitilize on an opportunity is not about luck. You make it sound as if in addition to the x86 design, the whole 'manufacturing thing' and process technology development is simply paint by numbers that any idiot can do.

Yeah, Intel's ability to manufacture the chips and shrink them at what in virtually any other industry would be an ABSURD (and unsustainable) pace is just riding a wave and represents no technical merit. The decision to get out of the memory game and focus on logic way back when was also luck I guess. In fact I think you can track almost every major strategic decision and every key patent back to luck.

Do you really believe some of the statements you are spouting out or are you just exaggerating / trolling for effect? Seriously man, I thought you were more intelligent than some of what is coming out - it seems like you have no grounding in reality and just choose to say anything Intel or MS does must by default be bad and any accomplishment mere luck out of some vendetta you have against these companies.

SPARKS said...

"Do you really believe some of the statements you are spouting out or are you just exaggerating / trolling for effect?"

"G"- I'm afraid he does believe everything he says, he's not trolling:

"Ultimately I think the basic notion of corporations is flawed and should be abolished. But in the interim, another measure that would improve the parity among all parties would be to require that all corporations extend ownership to all of their employees. The way large corporations get away with abusing folks working for minimum wage while company profits soar is totally reprehensible. The phrase "wage slave" should not even exist in our language."

This just a small excerpt of the larger piece, which ---um---is an interesting read.


http://slashdot.org/~hyc/journal/

SPARKS

SPARKS said...

ITK-

"choose to say anything Intel or MS does must by default be bad and any accomplishment mere luck out of some vendetta you have against these companies."

Yes, quite right, it's personal.


"I suppose some would object that "this is America! We should be free to run our businesses any way we want!" That's true, but only as long as how you run your business doesn't infringe on my personal liberty. In today's world, Big Business is a major threat to civil liberty and it *is* the government's proper duty to remove that threat."

From the same read.

http://slashdot.org/~hyc/journal/

Hence his discust with my admiration of INTC. Mr. Chu is a communist by the truest definition.

SPARKS

hyc said...

Sparks, since it seems that you so completely disagree with the views I posted in my journal - do you really think it's OK that:

* the RIAA and MPAA are asking for huge fines for small-time copying infringement, far beyond the penalties for grand theft auto, other felonies, and even manslaughter?

* companies like Sony get away with installing trojans/rootkits on your PCs without any criminal penalties?

* the RIAA and MPAA dictate how we should be able to manipulate our own creative works, by dictating how the consumer electronics companies build their devices?

* companies like Disney push through legislation extending US copyright to 95 years past author's death?

Big Business *is* a threat to personal freedom.

re: Intel's manufacturing expertise - it's already demonstrated that there were other companies (like Motorola) that were better at designing and building processors than Intel. Intel continued to grow because they were *lucky* enough to be in the PC that IBM ramrodded as the industry standard. Yes, they rode on the wave of IBM's decision, not their own foresight.

As for Moore's Law - BFD. The number of transistors in your CPU has grown exponentially over time. Has your computing experience today improved exponentially from what it was even 3 years ago? 5 years ago?

Aren't people everywhere already saying an Atom is good enough for their day to day needs? They haven't needed a faster desktop CPU in years? Weren't people all over the place saying "why would I need a dual-core processor?" Aren't people now saying "why would I need a quad-core? I'll certainly never need octo-core..." What has anyone gained from the slavish pursuit of Moore's Law?

The fact is, Intel went off the deep end, and now we have unbalanced systems where the CPU is so many times faster than main memory that we can't get data in or out fast enough to actually make use of all that horsepower. Maybe Intel should have *stayed* in the memory business, and pushed the state of the art there.

And no, I'm not saying AMD is any better. Their whole "x86 everywhere" strategy nauseates me too, but it's obvious they're forced to play that game because Intel/x86 already won over most of the other competing architectures.

But again, the way it's been so far is not necessarily the only way it can be from here on out. Open source software means that binary compatibility is irrelevant. There's no need to keep dragging the bad parts of the x86 legacy forward into every new design. ARM is doing well, coming from the embedded world up into smartphones. It may be that it can continue growing upward, faster than x86 can scale downward. And no one cares that their smartphones aren't able to run their Windows/x86 binaries...

Anonymous said...

the RIAA and MPAA are asking for huge fines for small-time copying infringement, far beyond the penalties for grand theft auto, other felonies, and even manslaughter?

Nice equating copyright infringement/protection to patent protection... In any system there will be extremes (and I fully agree the RIAA and MPAA have gone way too far). But are you going to paint a broad stroke over all intellectual IP based on these organizations or perhaps some of the extreme examples like the patent troll shops that buy up patents simply to sue large corporations in hopes they will settle for a license rather than go to court? Sounds like a level-headed argument to me (?) Perhaps we should apply that to religious organizations and judge that on the radicals or political parties. EVERY system will have it's extremes, even open IP systems.

Their whole "x86 everywhere" strategy nauseates me too, but it's obvious they're forced to play that game because Intel/x86 already won over most of the other competing architectures.

Ahh... the victim card, if only the big bad Intel were not around, AMD would surely be giving away their chips at virtually no profit and powering computers with baby smiles and puppy's kisses. Apparently the $1000 FX series was another unfortunate example of AMD being victimized... I always love the 2 wrongs make a right argument. The evil Intel is forcing x86 on people so that makes it OK for others to do the same thing "to keep up".

The fact is, Intel went off the deep end, and now we have unbalanced systems where the CPU is so many times faster than main memory that we can't get data in or out fast enough to actually make use of all that horsepower.

Shame on them.... they should have just stopped development and waited for the rest of the industry to catch up.... or better yet given their R&D money to memory makers, MOBO ,makers, LCD flat panel designers, software firms... DAMN those crazy Intel engineers for making their product too fast! If only they would slow down like everyone else.

Aren't people everywhere already saying an Atom is good enough for their day to day needs?

Just so I'm clear this is the same Atom that is deemed a complete and utter failure at UAEZone? How could everyone being saying this if it is such a failure?

It's crazy on one hand you argue about bringing the best technology to the market and then on the other hand you whine that Intel has outpaced the rest of the IT industry. Which is it?

I'm just spitballing here... perhaps other companies should step up and improve the rest of the systems around the CPU. Or should just Intel stop developing and wait? Or perhaps Intel should take on these other areas too?

Serious question: In your world, since the CPU has outpaced the rest of the system, what would you have Intel do?
- Simply wait for the rest of the industry to catch up? (this might involve lowering R&D budgets for the short term and perhaps cutting prices accordingly?)
- Use the money to take on some of the other areas that are lagging (at the risk of becoming an even bigger company, which would only make them more evil and more intrusive to personal freedoms?)
- Consult/give away their expertise to those in need? (could they charge for this? or should it be free? or should they not bother sharing and just wait for others to figure it out?)
- Start a new non-x86 architecture that will potentially eliminate all competition if successful (and potentially give someone like MS or another large company who needs to spend tons of money to develop an OS a virtual monopoly on the new architecture?).

Seriously - given your view of Intel and the world in general, what should they do right now given the state of the industry?

What has anyone gained from the slavish pursuit of Moore's Law?

Computers that don't fill the entire size of a room. The ability to connect and access information wirelessly. Access to improvements by the use of high performance computations in numerous fields (drug development in medicine for example). Then there's the everyday convenience factors (which I'm sure some will dismiss as unnecessary, but if push came to shove would not want to give up); banking or shopping online, downloading a movie or a CD instead of driving to a store, making a reservation, reading others' reviews of product, movies, art, etc. (you know in order to make an informed decision instead of being "marketed" into a decision)

SPARKS said...

"What has anyone gained from the slavish pursuit of Moore's Law?"

Thirty-eight years ago I built an electromechanical device for my remote control aircraft called a 'servo.' The kit was designed entirely of discrete components. With nine transistors, their associated circuitry, and a tiny circuit board, it was a nightmare to put together. That was until a newer more powerful model with and op amp was introduced a few years later. It eliminated most of the discrete components by two-thirds, sans the two out put transistors. Hence the beginning of my fascination with integrated circuits

In high school, at the same time, they had this huge computer, that from my estimation, at the time, it did nothing but print out useless lines gibberish on paper with holes in the side.

Today for what I saved I good portion of my weekly salary as a dishwasher for two servos, you can now purchase and entire R/C system for 1/5 the equivalent amount. Likewise, the absolute cheapest computer I could assemble with JUNK I have laying my hobbist/lab would decidedly blow the absolute doors off that monstrosity we had in high school.

Today, I work with automated building controls and fire alarm systems that have been a boon toward energy savings and public safety.


"Has your computing experience today improved exponentially from what it was even 3 years ago? 5 years ago?"

The speed and graphic detail in my specific applications have heighten my playing experience exponentially. The AI in games like Crysis, Far Cry, F. E. A. R. and C.O.D., along with the precision and detail for MS FS X, leave any that was developed five years ago in their wake. Chips made five years ago would barely run such software today. You would be looking at slideshows, if at all.

It's very comforting to know that when my wife goes on one of her forays to central New Jersey that she has a marvel of communication called a cellphone, and her Garmen GPS to assist her and my childen to a safe destination.

Personally, I am waiting for you program boys to come up with a few lines of code to produce me a cheap HAL 9000 that will monitor my bank and investment activity, home energy usage, and friend to help eliminate this key board as a communication interface.

"They haven't needed a faster desktop CPU in years? Weren't people all over the place saying "why would I need a dual-core processor?" Aren't people now saying "why would I need a quad-core?"

Frankly, you guys, especially those who have 'the gift of talking to the machine in code' are behind the curve on utilizing this power towards the previously mentioned acheivements and goals.


"do you really think it's OK that........"

Yes, I think it is essentail the technogy and its steady march foward go on. I would love to see the US Air Force have intelligent UAV's flying over New York City (where I work). No one would even think of destorying my city, my colleagues, my friends, and invarrably me. The result of such actions would be severe.

DARPA, in coorperation with INTC are striving towards that goal, I applaud it, in fact, I demand it. I for one, will never forget.

You asked the wrong guy about tech, I love it. I look foward to future advancements daily, and I am on this site because I respect and admire some of the most brilliant individuals I've ever known.

And, these guys are right on the edge. The tech must go forward, especially now, and WE WILL get there first.

SPARKS

SPARKS said...

"Ahh... the victim card"

"Shame on them...."


You know, I thought I had the market cornered on sarcasm, you sir, have achieved levels I can only aspire to.

"What is thy bidding, oh master?"

SPARKS

hyc said...

Nice equating copyright infringement/protection to patent protection... In any system there will be extremes (and I fully agree the RIAA and MPAA have gone way too far). But are you going to paint a broad stroke over all intellectual IP based on these organizations or perhaps some of the extreme examples like the patent troll shops
There are extremes because the laws written on paper are meant to be interpreted by reasonable people, and so they don't attempt to specifically enumerate all of the ludicrous uses of the law that should be prohibited.

Patent trolls are scum, no doubt about it. And when trolls go after Intel and Intel wins, I cheer for Intel. There's such a thing as fair play.

Ahh... the victim card, if only the big bad Intel were not around, AMD would surely be giving away their chips at virtually no profit and powering computers with baby smiles and puppy's kisses

I never said anyone should work for no profit. I work with open source, but my company still makes a profit. We all have to earn enough to sustain our R&D and keep delivering new, improved products/services. I just think that after you've covered your R&D, expenses and overheads etc., you're golden, and anything beyond that belongs to the workers who put their sweat into creating the wealth in the first place.

My beef with the concept of corporations, and the legal fiction that gives them the rights of natural persons, is that it devalues the individual people who are actually the key to any enterprise's success. I think it's wrong that companies can own IP, and that they can force their employees to sign over IP that the employees created. *People* create IP, *people* should own it, not corporations. If you have an employee creating valuable IP for you, then it's in your interest to treat that employee well and make it worth their while to stay with you. Simple, and *fair*.


The fact is, Intel went off the deep end, ...

Shame on them.... they should have just stopped development and waited for the rest of the industry to catch up.... or better yet given their R&D money to memory makers, MOBO ,makers, LCD flat panel designers, software firms... DAMN those crazy Intel engineers for making their product too fast! If only they would slow down like everyone else.


10GHz Prescott.

Intel pursued CPU clock speed to the exclusion of anything else, and that was clearly a mistake. Now we're engaged in core-wars instead of megahertz-wars and again, day-to-day computing experience just isn't improving. That tells me that Intel is again focused on the wrong thing.

Serious question: In your world, since the CPU has outpaced the rest of the system, what would you have Intel do?

Take a step back and look at the landscape, and see what real bottlenecks remain. In that respect, I think that Nehalem is a good move, it shows a real willingness to break from the FSB past and it shows really substantial gains over predecessors.

I don't think it's carried to the logical conclusion yet, though. The bus was a fundamental problem because contention increased so quickly as more cores came online. The system really needs a crossbar switch and dedicated point-to-point access to every high speed off-chip device - memory, chip interconnects, etc. If we plug 8 DIMMs into a machine, we should be able to have memory accesses running on all 8 DIMMs simultaneously, all the time. E.g. the Niagara T2 has 60GB/sec of memory bandwidth thru 4 memory controllers. That's also a step in the right direction (but the rest of the Niagara, forget it).

Sparks: your examples are all good ones, and hell, I love technology too. But they don't answer any of my 4 bullet items - in fact those RIAA/MPAA guys are the anti-technology guys. Time&energy spent on DRM is just wasted effort, and a diversion of good talent that could otherwise be advancing technology for everyone's benefit.

For gaming, I have no response; I haven't played games on my machines since the mid 90s. Nowadays I spend most of my time compiling/debugging code, and the occasional web surfing like anyone else. My web surfing experience hasn't really changed in 10 years, and disk speeds are such that my coding experience hasn't changed substantially either. (And I realize the lack of progress in disk storage is not Intel's fault. Their new SSDs seem pretty cool though.)

re: HAL 9000 - I think you'd be very irritated when you really do come face to face with an AI smarter than you. Kasparov was pretty pissed off when Deep Blue beat him, and that wasn't even AI... As long as tools only help us with our thinking, as opposed to doing all our thinking for us, we'll be OK. When that line is crossed, I think most of humanity will disappear.

Anonymous said...

I think it's wrong that companies can own IP, and that they can force their employees to sign over IP that the employees created.

You are ignoring the fact that a person signs this agreement in advance - if they don't like the IP agreement, work for someone else, or form you own startup company and figure out how to get the millions of dollars to get your idea off the ground. In your simplistic world, you also ignore the fact that IP is often generated by MULTIPLE people working together, not a singular person. What if one wants to walk away and go sell it or do something on their own? Are the others left to say OK, good luck, we're veru glad we could give you parts of the idea.

So your solution is what? The employees own it and are free to take the IP to the highest bidder and when they get bored with that bidder they move on to the next bidder. That sounds good on paper, but what happens to a company that is employing, say 5000 people making a product, and one of them decides - this is my patent and I'm going across the street and well if the other 4999 lose out and lose their jobs and have to stop making the product, so be it...

Well 'we' (meaning the gov't) can force licenses, right? No I got it we can have the government be the arbiter of value on all IP (but then instead of the corporation owning it, you effectively have the gov't controlling it). Surely a gov't would not be susceptible to the same issues corporations have...

I know we will make it free to everyone (no more worries about pesky IP theft or IP enforcement). I can just imagine walking into AMD with a patent (or 'idea' in this new world) thinking if I only had 10million I could get this idea off the ground which would change computing as we know it... AMD agrees it is a great idea but they worry after that investment in me and MY IDEA, I can sell it to anyone who I'd like to buy it (or in the IP free world anyone in the world can copy it, but conveniently wait until after all the development and debugging is painfully done by the sucker, ummm I mean initial company).

All I'm hearing is whining about what is wrong with an IP based system, but no actual solution that is better (or is not riddled with as many if not more holes) than the current system.

On a side note, out of curiosity, all those point to point additions you suggest, just how would we be able to do that without increasing transistor count (and, oh, I don't know, maybe benefiting from Moore's law?). Yup Moore's law is a BFD, until of course you need it...

Nahh... we could do it on 0.5um technology (which is ~15 years old to put things in perspective) and a die (with similar transistor count) today would be ~12,300 times bigger. We could do it on 6" wafers like the olden days... Of course that would make the die size slightly bigger than the wafer (~100X) so we'll just need to clip out a few 100million (or billion?)transistors... oh and since we are getting 1 die/wafer, we'll have to hope there are no killer defects and the chip may cost ~$2000-3000 (assuming of course we have perfect yield which most manufacturers with any sort of "luck" should be able to obtain). Yeah this whole Moore's law thing...clearly over-rated. I have no problems with folks questioning whether it is sustainable going forward or whether it approaching diminishing returns (there is a decent article on silicon strategies about this), but to generalize Moore's law into "BFD" is said by someone who has no actual understanding of HW and manufacturing.

SPARKS said...

"HAL 9000 - I think you'd be very irritated when you really do come face to face with an AI smarter than you."

If I may make a suggestion. Take a ride in a Cessna 172 in MS FSX from your local airport on a really good machine. You may be suprized at the level of detail and the functional acuracy of the simulator along with the aerodynamic sililarties of the real machines.

I build tube amplifiers as another hobby. I type in the parameters and out comes a schematic whose results have been breathtaking. I call my design programs like MS 'Visio', assisted living.



As for smart computers, negative, I would never compete with a computer. I try not compete with people.

I seriously doubt a computer couId compile an idea like, "To thine own self be true, and it must follow, as the night the day, thou canst not then be false to any man."

Compiling stored facts in memory registers are one thing, inspiration and the ability to leap beyong logic quite another. Think, V'ger and his need to join with the creator.

I have long realized that I am merely a bit player in very large machine. But the machine works. I'm happy, well fed, fat and sassy. Sure, the system is flawed. Then, so are we. In fact, we all are; I don't care you you are, there are no luggage racks on a hearse. A computer would never understand that sentiment.

Frankly, the older I get, the human condition becomes more complicated and difficult to understand. Then again, that's what makes us (humanity) unique and special. Therefore, could two computer argue the socialogical merits of intelectual property? I think not.

Kasparov was a pompous fool. Aurther C. Clark, Gene Roddenberry, Issac Asimov, were brilliant. In fact, I'll submit that a computer would be a fool to compete with a human on this extremely high level of thought.

A computer will never sit in a room and make journeys at the speed of light, and then mathematically explain the physical properties of the universe. It would merely wait for input.

Leave that kind of thinking to humanity where there is no equal, and no competition.

A computer may play chess with a man, but it can't cannot play dice with the universe, leave that one to Einstein.

SPARKS

InTheKnow said...

Hm, several of the 486 design decisions were superior? Please enumerate them... Having written compiler backends for both, and tons of assembly language for both, I obviously think the 680x0 was superior all around, but I'd like to hear your perspective.

From the article you posted, and therefore, not my opinion, but someone you felt worth quoting I offer the following summary.

CPU General-Purpose Registers
From the point of view of the CPU register file, the MC68040 has a very clear advantage

FPU General-Purpose Registers
Because of the stack organization the 80486 might have a slight edge from the standpoint of compiler generation (for that part of the compiler dealing with floating-point operations).


MMU on Chip
I call this an edge to MC6840, but external bus limited.

Cache on Chip
The difference is that while the 80486 on-chip 8k cache is mixed, storing both code and data the MC68040 cache is subdivided into two equal parts: a 4-kbyte data cache and a 4-kbyte code cache. Each cache is controlled by the respective MMU, mentioned above. The advantage, as in the MMU case, is the provision of two parallel paths for code and data, resulting in an overall speedup of operation.
What isn't mentioned here is the advantage of a larger block of cache if needed.

Segmentation
Therefore, as far as segmentation is concerned, the 80486 and MC68040 are comparable. The 80486 has some edge, since it allows the user to implement segmentation if needed and avail oneself to its advantages.


Paging
The MMUs of both systems feature paged virtual memory management. The 80486 offers a single standard page size of 4 kbytes. This results in reduced complexity in the MMU hardware and in the OS software, one of whose tasks is to support the management of virtual memory. The MC68040 offers two page sizes, selectable by the user: 4 kbytes and 8 kbytes. This tends to complicate the MMU logic and the OS.

TLB (or ATC) Size
In this case, a strong advantage of the MC68040 is obvious.

Levels of Protection
The protection mechanism of the 80486 offers more reliable protection, it also results in more complicated on-chip logic. More time is taken up with protection checks on the 80486.


Instruction Pipeline Stages
The 80486 instruction pipeline has five stages, while that of the MC68040 has six. This means that the 80486 pipeline can handle five instructions simultaneously and the MC68040 can handle six. This certainly
gives an edge in favor of the MC68040


Performance Benchmarks
A clear edge to the MC6840 in the selected benchmarks.

Memory Access Clock Counts
Edge to MC68040

So of the 11 items listed, your article gave the edge to Intel in 4 of them. If that was all there was to it, then the moto chip was easily the winner. But that isn't all there is to a microprocessor.

There is size, which controls manufacturing cost and power which drives thermal limits.

After a bit of digging, it looks like the MC68040 was manufactured on 0.8 micron process while the initial 486 was manufactured on a 1 micron process.

The result should have been a cost edge to the MC68040 with a die size of 153mm^2 compared to the 486's 294mm^2 (this should have dropped to ~185 when production moved to 0.8 micron).

So what killed what looked like an obviously superior chip? Thermals. The moto chip used 6.5 watts compared to 3.5 for the Intel chip.

And power scaling was atrocious for the MC68040. Power jumped to 9.5W when speed was scaled to 33MHz. Plans to scale to 40 and 50MHz were canceled.

The follow on chip the MC68050 continued to have thermal issues and never entered mass production.

So while the MC68040 was the better chip at the moment the design didn't have longevity. I call that a fatal flaw in the grand scheme of things.

InTheKnow said...

Now we're engaged in core-wars instead of megahertz-wars and again, day-to-day computing experience just isn't improving. That tells me that Intel is again focused on the wrong thing.

First of all AMD started the "core wars". Maybe you remember dual core for dummies? Second, unless you can come up with a real game changer, if multi-core is a dead end, they you have declared advancement in computing to be at an end. Because the power wall is real and we've hit it.

I see that you propose changes in the memory structure to do this. I believe this was what Intel's polaris project was all about. 80 cores, each with their own piece of a shared memory structure. So Intel is not only looking to address this, they already have working Si.

As for Moore's Law - BFD.

So let's work backwards. The atom chip which you admit is enough processing power is enough for many people is 25mm^2 on 45nm. If we assume a very conservative size increase of 60% per process node, then by the time we have backed up to 3 micron node where things started we are up to 2749mm^2. That is a 100 fold increase from where I sit. The form factors I get with that don't seem to desirable.

Intel's roadmaps also show a number of SOC solutions at 32nm. If they could have done this with acceptable form factors at a larger node, they would have. So yeah, it is a big deal.

Anonymous said...

Sure, the system is flawed. Then, so are we. In fact, we all are; I don't care you you are, there are no luggage racks on a hearse.

I must say this is/was very well said. There are those who would rather just complain about the flaws and offer up no real solutions, then to consider some of the positives and consider that despite the flaws, maybe things are not as bad as they seem.

Look at the economy... 93 out of every 100 people in the US who want to work, do work. Yet 7% unemployment is considered financial Armageddon justifying trillions of stimulus, and politicians including our President of hope and change are negatively hammering away in the press to justify all sorts of emergency spending. (Will money for the DTV transition really be an actual part of the emergency stimulus package? Are you freakin kidding me?)

To put things in perspective during the great depression unemployment was ~30%, during the late 70's/early 80's it was ~15%, and now it has skyrocketed (GASP) to an out of control... wait for it...7%! (GASP Again!) To put things in further perpsective, most economists consider 3-4% the base unemployment level in a booming economy (people moving, looking for new jobs, etc) - so we are cringing and knee-jerk reacting to an additional 3 out of a 100 people unemployed. Don't get me wrong this is not insignificant but to hear the Obama lackeys at CNN ad NBC talk about it, you would think it is the end of the world without a trillion dollars of additional spending. These are the same folks who criticize the politics of fear, yet they apply the exact same fear mongering - it is just turned toward the economy (or the nig bad oil companies, or monopolies or banks or the [FILL IN CONVENIENT WHIPPIN BOY INDUSTRY OF THE DAY HERE]) to get their programs pushed through.

The real problem is we are living in the ADD generation where everything is instantly analyzed and nothing is put in actual perspective and everything demands an immediate solution (which invariably involves more gov't spending).

SPARKS said...

"...consider that despite the flaws, maybe things are not as bad as they seem."

"The real problem is we are living in the ADD generation where everything is instantly analyzed and nothing is put in actual perspective and everything demands an immediate solution...."

Gospel. The goal ultimately, permit me to add, is perfection. At risk of being redundant, it was my original premise, 'looking at the donut and not the hole.' It's like some twisted psychotic trying to obtain absolute cleanliness 24/7. How much can someone scrub their ass before their hospitalized?

Can there be 100% unemployment, of course not. Can there be a "victimless society"? My grand ma said, "If-a everybody gonna go to-a college who-a gonna clean-a the fish?

Is there a perfectly flawless chip out of 100 wafers. As a scientist, you and others here, have taught me about manufacturing tolerances. It's a great word, and that is the key, tolerances "put in actual perspective" as you say.

Can a motor be built with 100% efficiency, never! The fundamental principles of thermodynamic prohibit such a RIDICULOUS notion. We (who are trained to understand those principles and live within those tolerances) are compelled to accept this as physical law.

Politicians, law makers, rabble-rousers, whatever, are trained to sell us, the public, perfect solutions to all our societal ills. It never happens, because it can't happen, no mater how pure and altruistic the intent. Someone, somewhere is going to bitch. Someone, somewhere is going to be a victim! (The Victim Card)

Take our friend, he dispises the "lucky Bill Gates", and undoubtly, the billions he has made over the years. However, the donut, if you will, is that Gates and his wife have, along with the likes of Warren Buffet (and others), turned charity work into goddamned industry, amounting to billions anually. Do these 'capitalistic pigs' get credit, no they're theives, charlatins, and slave drivers of the masses.

Clearly, as we both know, given our backround in the sciences, tolerances are a fact of life. Perfection may be acheived in thought, never in practice. I mean really, how much money can an engineer throw into a perfectly round steel ball? It looks great, it can't do much for you, and you and both know it can never be ABSOLUTELY perfect. Ultimately, here will be someone behind you complaining, "all that money and you can't even use it as a paper weight. It will roll off your desk"

This is the real problem in our times, in our society, trying to make the impossible, possible. It can't happen, there are too many pitfall created with every solution given.

Taking profit away from corporations who support millions of people, taking from the rich to feed the poor, and throwing exorbitant amounts of money at issues, is certainly NOT the answer. Finding the best, cost effective rational solution is. That takes work, and time; and it will never be perfect This is indeed a rare commodity in our times, and getting rarer by the minute.

That, my friend, defines value, and those that can objectively appreciate it.

SPARKS

hyc said...

[I think it's wrong that companies can own IP, and that they can force their employees to sign over IP that the employees created.]

You are ignoring the fact that a person signs this agreement in advance - if they don't like the IP agreement, work for someone else, or form you own startup company and figure out how to get the millions of dollars to get your idea off the ground.


No, I'm not ignoring anything. Case in point, the company I joined in 1995 was acquired in 1999, and we all had to re-interview to keep our jobs. And sign a new IP agreement that claimed ownership of everything we produced, whether on the job or off. In my case that would have meant signing over rights to all of my freeware and hobby activities as well as my professional code. I chose not to sign, not to re-interview, and start my own company. Which is how I got where I am today.

In your simplistic world, you also ignore the fact that IP is often generated by MULTIPLE people working together, not a singular person. What if one wants to walk away and go sell it or do something on their own? Are the others left to say OK, good luck, we're veru glad we could give you parts of the idea.


And again, no, I'm well aware of this fact. Not just from the dev team I worked with (6 of us co-founded my current company) but also from the Celtic band I formed in 1995. I drafted a contract to ensure that all band profits were divided equally among us. We recorded a CD, the same applied to that. Some of the band members left, but they continue to receive their portion of profits from the CD sales.

I could easily have said "I'm the leader of the band, I'm more important than all of you so I get a bigger share of the profits." But the reality is that a band is synergistic, the whole is greater than the sum of its parts, and when any member is removed from the picture, the whole thing suffers. Ultimately having the product is binary - you either have a great sound, or not. Everyone who contributed gets rewarded equally.


So your solution is what? The employees own it and are free to take the IP to the highest bidder and when they get bored with that bidder they move on to the next bidder. That sounds good on paper, but what happens to a company that is employing, say 5000 people making a product, and one of them decides - this is my patent and I'm going across the street and well if the other 4999 lose out and lose their jobs and have to stop making the product, so be it...


Well... whenever that single person walks, he still has to share 4999/5000 of his new sales with everyone else listed on the patent.

And when you make your employees shareholders of your company, they have stronger reasons not to job-hop arbitrarily. Again - people have free will, but you can make it worth their while to stay. You can make it in everyone's best interest, not just in management's best interest, or not just in the individual's best interest.

hyc said...

InTheKnow:

Your conclusion on power trumps everything else, but I'll still mention a few things...

FPU - stack vs registers - I disagree with the original article, directly-addressable FP registers are superior for ease of programming.

Caches - you're right, the larger unified cache gives a degree of flexibility. But nowadays x86s have separate I & D caches too - history has proven this to give superior performance given the same constraints.

Segmentation - the 680x0 family's first MMU (68451) was segmented as well and it's all-but-forgotten. Again, history has shown that paging is superior to segments, in all cases.

Levels of Protection - 4 levels aren't more reliable than 2. In practice, until paravirtualization hit the scene, x86 OSs only used ring 0 and ring 3 anyway, so that was a degree of complexity carried thru many years for nothing.

Fyi, there are 40MHz 68040s, so your info was partially wrong. But yes, the 50MHz version was cancelled due to thermals.

The 68050 was also cancelled, but apparently that's because the 68060 was progressing faster than anticipated, and the 80486 competition wasn't catching up fast enough to warrant the effort.

hyc said...

[Sure, the system is flawed. Then, so are we. In fact, we all are; I don't care you you are, there are no luggage racks on a hearse.]

I must say this is/was very well said. There are those who would rather just complain about the flaws and offer up no real solutions, then to consider some of the positives and consider that despite the flaws, maybe things are not as bad as they seem.


That we live in an imperfect world is undeniable. But that's not an excuse to not strive for perfection anyway. You criticize me for whining about the flaws; to me it seems that you are unable to envision something better.

I don't think I've ever played an entire fiddle tune perfectly. But I know that, in the thousands of hours I've spent practicing and performing, I've played a perfect note here and there. And knowing that, I keep trying to play a couple more, and so on.

Of course I don't have the 100% perfect solution, and I never will. But I'll keep trying to get closer to it. It's not practical or cost effective, but it's still worth doing.

Anonymous said...

hyc: I think it's wrong that companies can own IP, and that they can force their employees to sign over IP that the employees created. *People* create IP, *people* should own it, not corporations. If you have an employee creating valuable IP for you, then it's in your interest to treat that employee well and make it worth their while to stay with you. Simple, and *fair*.

Many companies offer their employees stock options, so in effect they get rewarded for their own efforts. Well, in normal economic times that is :). However my belief is that if a compnay makes available the resources and pays the salaries, then of course what the employees produce (product, IP, etc), then the company has a vested interest in that product or IP and should own it. I know that IBM offers its engineers bonuses based on the patents their research brings. There would be an impossibly fine line trying to distinguish what an employee develops in their own time vs. what they are paid to develop during work hours, if what they develop on their "own" time is the same or similar to what they were paid to develop.

As for the whole x86 discussion, I agree that from what little I know, there are far better architectures out there for specific areas (e.g., GPUs, maybe VLIW and RISC in certain scenarios). However, both Intel and AMD gave the people what they wanted - backwards compatibility with old programs. If you had to start at ground zero and buy all new apps with your new hardware, sales would tank and people would feel the CPU companies were exercising too much power and influence, and that would leave the door open to an upstart making backwards-compatible hardware. As somebody else pointed out previously, look at the itanium experience.

My understanding is that the Chinese have been working on a CPU (whose name escapes me at the moment) that is supposed to be able to run x86 code at about 70% efficiency, using emulation for the most part, to get around the lack of x86 license.

I also recall that around the time the IBM PC became popular, some company had a dual-CPU machine (Rainbow?) that would run both CP/M and x86 natively, although not at the same time. I don't think they sold too many of them either...

SPARKS said...

"But I'll keep trying to get closer to it. It's not practical or cost effective, but it's still worth doing."

Of course you will, that's your humanity. I sincerely hope you won't cut off an ear if you don't reach your perceived/personal goals, while striving to complete a masterpiece.

I spoke of thermodynamic laws, I remind you of the law of diminshing returns, which is the cornerstone of this practical debate.

SPARKS

Anonymous said...

hyc: As long as tools only help us with our thinking, as opposed to doing all our thinking for us, we'll be OK. When that line is crossed, I think most of humanity will disappear.

Considering that history is littered with such examples of "humanity" as Hitler, Stalin, Nero, etc. that may not be the best reason to hope that machines don't take over the planet :).

I recall reading a Scientific American article on the future of AI several years ago - many scientists think that even following Moore's law, we won't have a machine-equivalent to human intellect until about 2050 or so. So many of us here won't be around to see that day, I'd guess :). I also recall reading that present-day supercomputers have less equivalent intelligence than a housefly.

Anonymous said...

You miss and inother casea conveniently exaggerate the point:

Well... whenever that single person walks, he still has to share 4999/5000 of his new sales with everyone else listed on the patent.

Well what if those people weren't part of the patent and merely working to produce a product BASED on the patent? They are left out in the cold. In companies with 10,000 people, there are not 10,000 people on the patent. The company owning the patent gives these employees SOME (not complete) protection that the patent holders won't just walk away and leave the rest out in the cold.

And when you make your employees shareholders of your company, they have stronger reasons not to job-hop arbitrarily. Again - people have free will, but you can make it worth their while to stay. You can make it in everyone's best interest, not just in management's best interest, or not just in the individual's best interest.

Let's talk about the value of a patent on the free market... your stock participation plan is going to offset this? FOR EVERY EMPLOYEE? Are you on drugs? What's your plan give everyone in the company a $5,000,000 stake in the company in case they invent something valuable down the road?

You chose to walk away with your IP (as there had been no prior agreement in place). That is fine, it's called free will and you had no agreement in place. I have no idea what your patent was, but what was the impact on the people who stayed in your old company? Were those employees worse off when the company lost access to your patent?

Suppose the person who held several key graphic chip architecture patents just walked away from AMD right now and decided to go work for Intel or Nvidia in exchange for a nice raise and some stock. What would that do to AMD's employees working in graphics?

This is the problem I'm getting at... if you work for a large company and you utilize the resources of the company to develop and debug a key idea, I don't think it is your right to just walk away and say thanks for the help, now I'll just go sell it to the highest bidder or now that the key R&D is done, I'll start up my own company (note - I'm not implying you did this).

Anonymous said...

But the reality is that a band is synergistic, the whole is greater than the sum of its parts, and when any member is removed from the picture, the whole thing suffers. Ultimately having the product is binary - you either have a great sound, or not. Everyone who contributed gets rewarded equally.

And you think everyone who has or will own a patent will be able to live up to your standards and share your ideals? What happens when joint owners of IP don't agree on the path? What if say one wants to start a new company and one wants to stay? Well that's easy right they could just split things up right?

Ok what happens when the 10 different people on a patent have 10 different views? One wants to sell to a patent troll company, 1 wants to sell it to Russia, another to North Korea, 1 wants to form a startup, 1 wants to stay in the current company and the rest want to sell it to various comapnies in exchange for promotion/more salary/etc? Suppose 1 of the 10 is a greedy SOB who just sells it again and again and again to 10 different companies as he is not happy with a 1/10th share?

Anonymous said...

You criticize me for whining about the flaws; to me it seems that you are unable to envision something better.

No I'm criticizing you for trying to force a solution that requires everyone to be perfect into what is an imperfect world.

It's not about envisioning something better - the systems you put in place have to recognize the flaws in existence and as those flaws are removed/reduced then you can evolve the system. You don't just build a system hoping people will magically evolve into it and do it knowing that it won't work.

If I were a city planner and was envisioning a world without crime and poverty would I build a city (TODAY) without soup kitchens, homeless shelters, police station, or jails? If I envision that same world with no fossil fuels and 100% renewable energy, should I just leave out the gas stations and assume people will get to it eventually?

Getting back to Sparks - it's about tolerance. The system needs to be able to handle people who will sell their ideas for a quick buck no matter what impact that has on peole around them.

If you utopian IP society was implemented today several things would likely occur - numerous companies would fail due to skyrocketing compensation to "encourage" employees not to leave when they develop valuable IP. Other companies would fail when the owner of the IP decided he's had enough and walks away from a company and goes somewhere else, and takes his individually owned patent with him.

In his greed he decides to go somewhere with no child labor laws in order to produce the product dirt cheap and mazimize his own income, so the old employees are not welcome as they make too much money and cut into his profit margin and his ability to buy a 3rd private jet. Would I like the world to not have people like this? Sure. Does the world currently have people like this? Absolutely. Should I put my hands over my ears and yell I'm not listening and design a system where people like this can better and more fully exploit good natured and well intention people? Well if my name was HYC, it seems the answer is yes.

hyc said...

Many companies offer their employees stock options, so in effect they get rewarded for their own efforts. Well, in normal economic times that is :). However my belief is that if a compnay makes available the resources and pays the salaries, then of course what the employees produce (product, IP, etc), then the company has a vested interest in that product or IP and should own it.

I partly agree with you, but I think you're emphasizing the wrong aspect. Yes, a company that makes resources available for the development of a product deserves to be repaid for providing those resources. But ultimately it's the people creating ideas that are unique and irreplaceable.

Say you have a well funded research lab, and a team of good researchers working there. You can't get a good result without both pieces of the picture right? But the lab isn't unique - those researchers could produce good work in any lab funded with the same amount of money. On the other hand, a different team of researchers in the same lab might never achieve the same results. Lab equipment is interchangeable, human insight isn't.

So why should I give ownership of the IP to the company, when they play no special part in the creation of a product? The IP comes into being because of the people and their ideas.

Now, if you used company resources to develop your idea, then I think it's fair to require that all initial profits from the product first go toward repaying the development costs. Whether you stay or leave, you have a debt that must be repaid. But once that's done, the company's claim is over.

hyc said...

Ok what happens when the 10 different people on a patent have 10 different views? One wants to sell to a patent troll company, 1 wants to sell it to Russia, another to North Korea, 1 wants to form a startup, 1 wants to stay in the current company and the rest want to sell it to various comapnies in exchange for promotion/more salary/etc? Suppose 1 of the 10 is a greedy SOB who just sells it again and again and again to 10 different companies as he is not happy with a 1/10th share?

Remember, my basic premise is that you *can't* sell IP. The creator owns it, period. To "sell it to North Korea" means to actually go to North Korea and live and work there.

The greedy SOB can't sell the same thing over and over again. Aside from the above impracticality, there's the reality that you can't burn your bridges in this world, things you do to screw other people tend to come back and bite you in the ass. Once you become known as a greedy crook, how many other companies are going to want to do business with you? Once you've licensed your idea out to 3-4 companies, how valuable will it be any more? How many other companies will be willing to pay top dollar for an idea that's no longer novel, no longer gives them a competitive advantage?

f you utopian IP society was implemented today several things would likely occur - numerous companies would fail due to skyrocketing compensation to "encourage" employees not to leave when they develop valuable IP. Other companies would fail when the owner of the IP decided he's had enough and walks away from a company and goes somewhere else, and takes his individually owned patent with him.


How many people who've tried to sell trade secrets from Intel to AMD wound up working for AMD? Or vice versa? Most of them wind up getting turned in to the cops by the prospective buyer, don't they? Companies learn pretty quickly not to do business with people they can't trust. Why will company B pay to lure you away from company A, knowing full well that company C could come in the future and just screw company B the same way? Once word gets around that you're an unscrupulous SOB, no one will talk to you, because they know they'll just be the next company you screw.

Anonymous said...

How many people who've tried to sell trade secrets from Intel to AMD wound up working for AMD? Or vice versa? Most of them wind up getting turned in to the cops by the prospective buyer, don't they?

None that I'm aware of and thank you, you just proved my point with this statement. Why don't people sell trade secrets? Because in the world we live in today they DON'T OWN THEM and they would face criminal charges from the owner of the trade secret (the company) if they did. It is also illegal for the receiving company to use a wrongly received trade secret which also explains why the company would turn those people in! This isn't about "trust" are you really that nieve?!?

I can't really believe you think it is simply trust and karma that keeps IP from moving from one company to another.... thanks I haven't laughed that hard in a really long time.

If the person owned the trade secret (in your individual IP world) there would be no such restriction and all you would have to rely on is trust (which may work for some folks, but if you think this will discourage everyone you need to get out in the world a little more often).

Again you are talking about how you would like things to be, not how they actually are.

Anonymous said...

Remember, my basic premise is that you *can't* sell IP. The creator owns it, period.

Fine, let's expand this out... Someone has a critical piece of IP, say a major breakthrough in computer architecture. It takes, let's say AMD, 4-5 years to validate this guys invention (going from design, to incorporation into existing designs, to actual Silicon validation). This process cost 100's of millions of dollars and 100's of thousands of man hours to determine it works.

The guy talks to a friend at IBM and says I own this fantastic idea (remember he owns the IP), it will work with Cell and if you hire me for 2 million dollars I will train you how to implement it on your design. You think IBM says no to this? Why would they, there is nothing wrong/illegal, it is this guys IP and he is free to work whereever he wants. After IBM he goes to Intel with the same offer. After that a small startup with no money for R&D also hires him as a "consultant"...

Now AMD invested millions and ridiculous manhours into this and suddenly everyone and their brother has access to it. Then they realize - if the idea hadn't worked they'd be out millions; if the idea does work they'd get a minor competitive advantage until this guy takes it to their competitors. In this world why does AMD (or whatever company) take all the risk - there is large downside, but limited upside.

So the guy can't 'sell' IP; he can give it away / go work for another company / train people how to use it / consult (for, ahem, a fee). Suppose he just sticks it on the internet (it's his idea, I assume that is allowed)? He could do this free or offer it up as a newletter subscription for a minor fee.

Think about drug development - it takes 10+years in some case to do this, and to go through a multitude of trials and FDA(or other gov't) approvals and when done the owner of the patent on the drug can go work for a generic drug company which doesn't spend a dime on development and merely takes the formula and manufactures the drug in short order. And in your world "trust" would stop this guy from moving to another company?

hyc said...

Fine, let's expand this out... Someone has a critical piece of IP, say a major breakthrough in computer architecture. It takes, let's say AMD, 4-5 years to validate this guys invention (going from design, to incorporation into existing designs, to actual Silicon validation). This process cost 100's of millions of dollars and 100's of thousands of man hours to determine it works.

The guy talks to a friend at IBM and says I own this fantastic idea (remember he owns the IP), it will work with Cell and if you hire me for 2 million dollars I will train you how to implement it on your design. You think IBM says no to this? Why would they, there is nothing wrong/illegal, it is this guys IP and he is free to work whereever he wants. After IBM he goes to Intel with the same offer. After that a small startup with no money for R&D also hires him as a "consultant"...


You're forgetting that the guy doesn't get to spend the profits until the R&D costs are paid back. Anything he does to market the idea and get more revenue from it benefits "AMD" first. If licensing it to other companies, to get it into more products faster, produces more revenue faster than just keeping it in one company, then big deal, AMD recoups their investment faster, that's all.

Anonymous said...

You're forgetting that the guy doesn't get to spend the profits until the R&D costs are paid back.

What profits are you talking about? The company gets the profits on products sold, last I checked individuals (even if they own the IP) don't receive profits directly if they work for a company, they get a salary (and if luck some stock or stock options).

Are you talking about profit from selling (licensing) the tech? If so this is even more lunacy, as what if the guy just goes to work for another company at a higher salary and takes his own IP with him? Are you saying the new company has to give their profits to the first company? How would this work LOGISTICALLY, not the 30,000' fantasy land view.

How would the actual development cost be calculated? Who would do it? Some 3rd party arbiter? What if during the development of this IP it was coupled with other development costs (consider something integrated like development of an entire Si process flow - how would you take out the development cost of just a small portion of that?).

How would the profits of the new company be calculated as they may not be using the invention on all their products - do you separate out just the profits on the products that use the invention? Is this treated as revenue for the original company? How do you actually calculate out profits on a single product line when often some costs are SHARED between multiple product lines? Who does this calculation? What if the company doesn't want to share very sensitive cost data on all their products (which would be needed to filter out the cost structure on the product in question).

You just sit in fantasy, wave your hand and say well we'll just compensate the company for the hard work they did, but you have no idea how difficult (impossible?) this would be to do? What if it is more than 1 person involved in the IP and several companies?

The complexity is ridiculous in what you are proposing and would grind things to a halt.

Let me give you a simple example. The engineer with the high K integration patents at Intel, decides he wants to move for personal reason - maybe family, sick relative, etc... and interviews at TSMC and gets a job as an engineer there. As he owns the high K IP, he decides to implement it at TSMC.... now what?

He hasn't sold anything. He doesn't have any profits for Intel to take. Does Intel get some of TSMC's profits? Over what time frame? And how would this be calculated? If TSMC uses several processes (some with some without) does this get filtered somehow?

And this is probably a relatively simple (1 sole owner of the patent, only 2 companies involved)...

Anonymous said...

What a sorry state of affairs

Looks like the clock has run out on AMD and it didn't even do a full tick tock tick tock, LOL

Of IP WTF: Of course the company owns it, not the person. Who do you think was paying the employee. Its another thing for the company to own the experience or wisdome.

Of Moore WTF: If we hadn't had the last 4 or 5 generations of process node, you really think you'd be happy with orginal windows, 5' boot times, 800baud dial up and AOL. Without moore you'd have no cell phones, no nano, no GPS, no handheld phones etc. etc. People who don't know what moore did for them don't know shit. Its a seperate question as to whether the MegHz wars or the core wars really got as much as AMD and INTEL plowed into it. But without INTEL leading the charge in scaling there are a lot of things you wouldn't have.

But so boring now that the green boys got nothing to talk about as AMD is dead as I predicted many years ago.

hyc said...


What profits are you talking about? The company gets the profits on products sold, last I checked individuals (even if they own the IP) don't receive profits directly if they work for a company, they get a salary (and if luck some stock or stock options).

Are you talking about profit from selling (licensing) the tech? If so this is even more lunacy, as what if the guy just goes to work for another company at a higher salary and takes his own IP with him? Are you saying the new company has to give their profits to the first company? How would this work LOGISTICALLY, not the 30,000' fantasy land view.


Yes. How does any licensing deal work between companies? There's nothing different about this.


How would the actual development cost be calculated? Who would do it? Some 3rd party arbiter? What if during the development of this IP it was coupled with other development costs (consider something integrated like development of an entire Si process flow - how would you take out the development cost of just a small portion of that?).


This is why people go to school to become accountants. Give them something interesting to do...


How would the profits of the new company be calculated as they may not be using the invention on all their products - do you separate out just the profits on the products that use the invention? Is this treated as revenue for the original company? How do you actually calculate out profits on a single product line when often some costs are SHARED between multiple product lines? Who does this calculation? What if the company doesn't want to share very sensitive cost data on all their products (which would be needed to filter out the cost structure on the product in question).


Again, this is no different from the way royalties and licensing is already handled today. The only difference is that all of the licensing deals are between real people, not between legal fictions/corporations.

Anonymous said...

Yes. How does any licensing deal work between companies? There's nothing different about this.

Ummm... there's something HUGELY DIFFERENT about this, licensing deals are done when one party OWNS the patent (in your world neither company owns it... how could they license what they don't own?). What if the IP owner is not selling/licensing the IP they are simply using what they own at another company? Who pays the bills then?

Or in an ironic twist of fate if the IP owner say doesn't have the money to recoup the expenses is he now forbidden from using his own IP elsewhere?

Licensing deals also aren't cost based deals; is one company going to share every accounting of cost associated with the development costs (and perhaps give away valuable data to a competitor) or is the other company or IP inventor just supposed to take their word for it?

This is why people go to school to become accountants. Give them something interesting to do...

So accountants are going to be able to separate out specific costs to the granularity associated with say high K development... Do you have any idea what you are talking about... think of the development Si that is shared and may have 5,10 maybe more development experiments on a single lot of wafers - how do you itemize out the cost of one particular experiment or one particular technology? in the real world it is not just "well... that's what we have accountants for" Then there's shared tools, shared technicians, shared power and facilities (exhaust, water). Should all companies have specific metering on every single tool in the fab in the case they need to recoup cost if some IP ever comes out from an individual? Do they need punch clocks on every tool for technicians who work on several tools so they can get an accurate cost associated with a specific technology? Or will "estimates" be relied upon (and pray there are no disagreements on the accuracy of the estimates?)

Again, this is no different from the way royalties and licensing is already handled today.

Again you lack any significant knowledgeable detail as to what you are talking about. Royalty agreements aren't handle today by we'll just get paid until we recoup our development costs - we'll let you know when that is ("trust us" on the math).

And again how is a company, who no longer OWNS any IP, collecting royalties from another company. Royalties are simple....$x or X% of each product. Who chooses the length of time it is amortized over? What if the 2 sides can't agree (I know, in you're Utopian world, both sides will be noble negotiators and come to a fair agreement... but what if that doesn't happen)? What if the new company refuses to pay? The previous company doesn't own any of the IP, do they seek an injunction for IP infringement on IP they don't own?

You sir are living in fantasy land. The fact that your responses are "no different than today" and have no actual specificity, shows me you have really not given this any intelligent thought and just have this vague notion that somehow it must be able to work. Again it is a fine vision to have, but it is simply foolhard to think such a system could work in today's society.

The only difference is that all of the licensing deals are between real people, not between legal fictions/corporation

Walk us though a simple example:

HYC works for Intel and changes jobs to go to work for TSMC and brings his IP (that he invented and thus obviously owns) with him... He developed an advance process technology that makes transistors consume vastly less power. Intel had overall development costs for the entire process tech node at 3Bil and they estimate the specific technology in question accounted for 300Million of the 3Bil...Intel is unwilling to give a detailed breakdown of development costs as that would entail releasing IP on the other processes developed, and is also unwilling to share sensitive fab data (dev yields, capital equipment pricing) to an ex-employee who is working for a competitor.

So Intel wants 300Mil to recover the cost associated with validating HYC's invention, now what exactly does HYC do?

Is he forbidden from using his own technology until he can cut a check? Is there a timeline involved or is it simply you can't use until Intel has 300mil in its hand?

Is Intel free to use the technology even though the person who owns it has left the company?

What if HYC doesn't like Intel's estimate?
- Is he forced to hire his own team of accountants to verify?
- Is Intel forced to reveal sensitive cost/pricing data to support their cost claim?
- What if the 2 sides can't come to an agreement is the technology just frozen out of use from both parties?

...waiting for the 'how is this any different than today' response? Well what is different today - Intel (who would own the IP) would negotiate a deal with TSMC - the deal would involve either a lump sum payment or some % of every product sold using that technology (with perhaps a max cap on it) or a combo of both. There would be no "recouping of development cost calculations", no need for Intel to demonstrate or itemize the specific costs of developing the IP (so no chance of disagreement there) and if the parties couldn't agree on terms, things would be simple, Intel would continue to own the tech and TSMC would have no access (until the patent period expires).

InTheKnow said...

The 68050 was also cancelled, but apparently that's because the 68060 was progressing faster than anticipated, and the 80486 competition wasn't catching up fast enough to warrant the effort.

Let's look at this a little more carefully, shall we?

i486 - 1989
MC68040 - 1990
i568 (Pentium) - 1993
MC68060 - 1994

4 years is not stunningly fast for development time. Intel is currently on a 1 year cadence. Though it seems like 4 years might have been par for the course up until the mid '90s.

The real competition for the 060 was not the i486, but the Pentium. And the Pentium and the 060 were very comparable architecturally from what I've read. The 060 was better than the Pentium on mixed code but the Pentium smoked the 060 on (2x the speed) on well written code.

Couple this with the fact that the MC68060 was released with a 50MHz clock speed compared to a 75MHz speed for the Pentium and you have the MC68060 at a 50% speed deficit. They did get the MC68080 up to 75MHz, but the Pentium made it up to 200MHz.

Ultimately, it suffered from one of the main shortcomings of it's predecessor. The inability to increase clock speed on the design.

So if you want to go for the lowest common denominator, (i.e. code any old bozo like me can throw together) you can argue that the 060 was the better chip on a clock-for-clock basis. But in terms of getting the job done it just couldn't compete with the Pentium due to low clock speeds.

I come away from all this convinced that Intel was quite capable of designing a good chip, though not every design was a winner. This whole series of events is like deja-vu when looking at the P4 --> C2 transition. Intel reaches deep into their bag of tricks when they find they are behind and pulls out a winner.

hyc said...


Ummm... there's something HUGELY DIFFERENT about this, licensing deals are done when one party OWNS the patent (in your world neither company owns it... how could they license what they don't own?). What if the IP owner is not selling/licensing the IP they are simply using what they own at another company? Who pays the bills then?


If the IP isn't in commercial use, then perhaps it's of no commercial value. If it's helping streamline a company's internal processes, then some bean counter in the company knows how to calculate their ROI for implementing the idea.

(By the way, if you'd been following the Novell vs SCO court case, there's an example of one company (SCO) licensing IP that it didn't own (Novell's). Their arrangement was that 100% of license proceeds collected by SCO would be turned over to Novell, and Novell would pay 5% back to SCO for their stewardship of the IP.)


Or in an ironic twist of fate if the IP owner say doesn't have the money to recoup the expenses is he now forbidden from using his own IP elsewhere?


That would certainly be a strong disincentive for leaving the original company, wouldn't it?

But no, that's not the idea... The owner simply doesn't get to keep any profits from using the IP until the original cost is repaid.

Licensing deals also aren't cost based deals; is one company going to share every accounting of cost associated with the development costs (and perhaps give away valuable data to a competitor) or is the other company or IP inventor just supposed to take their word for it?

Use a 3rd party auditor...


- how do you itemize out the cost of one particular experiment or one particular technology?

Or will "estimates" be relied upon (and pray there are no disagreements on the accuracy of the estimates?)


When Company X brags "we have a patent portfolio worth $10billion" what did they base that number on? Estimates are good enough for investors to sink actual numbers of dollars into something...


Again you lack any significant knowledgeable detail as to what you are talking about. Royalty agreements aren't handle today by we'll just get paid until we recoup our development costs - we'll let you know when that is ("trust us" on the math).

And again how is a company, who no longer OWNS any IP, collecting royalties from another company. Royalties are simple....$x or X% of each product. Who chooses the length of time it is amortized over? What if the 2 sides can't agree (I know, in you're Utopian world, both sides will be noble negotiators and come to a fair agreement... but what if that doesn't happen)? What if the new company refuses to pay? The previous company doesn't own any of the IP, do they seek an injunction for IP infringement on IP they don't own?


Nobody applies for a patent on something without at least a vague notion of what it's good for, and what it's worth. The process is too much of a hassle for that. If you have IP that you believe is valuable, then you have an idea of how much that value is. You can set license terms in pretty much the same way as is already done today - $x per item or X% per item for a manufactured product.

And as I already noted above, there's already a precedent for one company collecting royalties on IP it didn't own. A license is a contract, once two parties enter into the contract it's enforced the way any other contract is.

If the inventor stays at the original company, how long will it take to recoup their costs? When does any company ever know that all the resources they poured into a research project were well spent?

If you spend millions creating a product that eventually sells for billions, you have a pretty clear indication that the costs were recovered and the effort was worth it. But how do you really know?

And - if you're just working for your flat rate salary and you singlehandedly invent this wonderful idea that your company then sells for billions, don't you think it's fair that you get some percentage of those billions, in addition to your base rate?


Walk us though a simple example:

HYC works for Intel and changes jobs to go to work for TSMC and brings his IP (that he invented and thus obviously owns) with him... He developed an advance process technology that makes transistors consume vastly less power. Intel had overall development costs for the entire process tech node at 3Bil and they estimate the specific technology in question accounted for 300Million of the 3Bil...Intel is unwilling to give a detailed breakdown of development costs as that would entail releasing IP on the other processes developed, and is also unwilling to share sensitive fab data (dev yields, capital equipment pricing) to an ex-employee who is working for a competitor.

So Intel wants 300Mil to recover the cost associated with validating HYC's invention, now what exactly does HYC do?

Is he forbidden from using his own technology until he can cut a check? Is there a timeline involved or is it simply you can't use until Intel has 300mil in its hand?


No. He can use it anywhere, but he can't keep any profits from its use until the debt is paid. Annual payments, quarterly, whatever.

The incentive to change jobs would have to be pretty high to overcome the disincentive of undertaking all this hassle. Why is HYC changing jobs?


Is Intel free to use the technology even though the person who owns it has left the company?


Yes, until the debt is paid. After that Intel pays the same royalty scale to continue using it, until the patent expires.


What if HYC doesn't like Intel's estimate?
- Is he forced to hire his own team of accountants to verify?
- Is Intel forced to reveal sensitive cost/pricing data to support their cost claim?
- What if the 2 sides can't come to an agreement is the technology just frozen out of use from both parties?


The numbers would be agreed on at the outset of the project. As such, Intel would be sharing data with a valued employee, not a disloyal ex-employee. Based on the original project plan, plus any cost overruns, you have a pretty clearly documented number.

Why are we having this conversation? Are you trying to convince me that the entire idea is stupid and unworkable, or are you trying to find specific problems that can be fixed, eventually rendering it workable?

hyc said...


The real competition for the 060 was not the i486, but the Pentium. And the Pentium and the 060 were very comparable architecturally from what I've read. The 060 was better than the Pentium on mixed code but the Pentium smoked the 060 on (2x the speed) on well written code.

Couple this with the fact that the MC68060 was released with a 50MHz clock speed compared to a 75MHz speed for the Pentium and you have the MC68060 at a 50% speed deficit. They did get the MC68080 up to 75MHz, but the Pentium made it up to 200MHz.


The Pentium was originally released at 60MHz, and even that design had thermal problems.

http://www.cpu-world.com/CPUs/Pentium/index.html

According to this article,

http://en.wikipedia.org/wiki/Motorola_68060

However, with properly optimized and scheduled code, the Pentium's FPU was capable of double the clock for clock throughput of the 68060's FPU.


You said "well written code", but the actual point is FPU-intensive code, where "mixed" meant mixed integer and FP. The majority of code they targeted was obviously int-heavy. Unless you were heavy into scientific computing, the FPU difference didn't have any impact on users.

Anonymous said...

The incentive to change jobs would have to be pretty high to overcome the disincentive of undertaking all this hassle. Why is HYC changing jobs?

Now you are just meeting antagonistic.. what if you had a sick family member and wanted to work close to where they were located? What if your boss was a complete jerk? What if you found there was no upward mobility in your current company? What if you were getting married and moving to where your wife is from? What if you needed a better salary to pay off a debt or cover some medical bills? Come on now... people move jobs all the time and believe it or not it can be for non-financial or non-greedy reasons. IP ownership should not be a dissuading factor for attempting to change jobs just in case the guy is greedy (what if it's not a case of greed?)! IP should not be used as a tool to disincentivize people from moving jobs by creating a "hassle" to basically filter out people who aren;t committed to wanting to change jobs - that is just plain idiotic.

And AGAIN YOU MISS THE BOAT on profits... the IP owner, if he is working for a company might not see a single cent in profit when he needs to or wants to move jobs. His patent may still be years from getting to market... what then? You wave around these vague notions of quarterly payments but who's paying them? We could be talking about an engineer earning 80K/year and a patent that cost a company 200million to do development on. If that product hasn't gotten to market when the engineer wants to leave, what quarterly or yearly plan is he going to pay these costs back with?
Or is he stuck with the current company, through the hassle of the new IP model? Or is he forced to sign it over in exchange for his freedom to move (of course in your world companies cannot own IP so I don't know who he wold sign it over to)? And before you say ot could just be future profits for the company what if the company doesn't recoup cost with the profits (and yet the new employees company eventually turns a tidy profit from it?)

The numbers would be agreed on at the outset of the project. As such, Intel would be sharing data with a valued employee, not a disloyal ex-employee.

Again, I find it hard to believe you actually work in the real world. You file a patent application, it will take well over a year to even know if it gets approved. And the employee and the company are suppose to pre-align on the outset of the project? What if their is scope creep or cost creep? Is their a quarterly/yearly audit to keep track of these things? How many patents does a company like IBM or Intel have every year? Several thousand? And in your world, the company is aligning with the patent holder on a # for EVERY ONE of these patents? Now how many PATENT APPLICATIONS? (times 2 or 3?) Since the projects often start long before the patent ruling is in, you are talking about aligning on EVERY PATENT APPLICATION in the event an employee leaves. And I'm ignoring the mutiperson patent which is another complete minefield in your scenario)

And your notion that ALL data should be shared with the employee is laughable... you are confusing data on the technicals of the project with specific cost data that a company may not want a lot of people knowing (and before you think this is evil read on). In this case Intel will share all the salaries of people on the project with the patent holder (I'm pretty sure there may be some privacy regulations), all the capital tool costs (this is not something an engineer may typically knows), all the high level fab costs, contracted wafer costs, expected profit margin on every part that this process might be used on (which of course don't exist as the process is probably 3+ years out). This is not about loyal/disloyal, you are talking about several thousand people where this data may now need to be shared with... While in utopia all information can be shared with everyone, in today's companies not all data is shared with all employees.

And again we are talking about several thousand patents a year in some companies so this is a process that you fully expect to be done up front for each patent, or even worse each patent application (as you don't know which ones will eventually be approved). Are you staring to see the absurdity of some of your suggestions/solutions?

Are you trying to convince me that the entire idea is stupid and unworkable, or are you trying to find specific problems that can be fixed, eventually rendering it workable?

Mostly the former - the idea is not workable in today's society. It is not a matter of it being stupid - it just won't work... it may be laudable and an ideal solution, but that doesn't make it a good solution. In whatever society you live in where everyone is working for the collective good, perhaps it would work... but in today's world... no... And I don't think the problems can be fixed.

I'm trying to point out the folly of this "we'll just use accountants", "we'll just agree on development cost for EVERY PATENT application going forward upfront, just in case you decide to leave and we therefore have a fair #" (yeah that won't suck up too many resources on several thousand patent applications), and "we'll just share every bit of valuable data with any employee with a patent application because they are loyal". Part of the problem is you expect everyone to be working toward a common ideal - that may not be the case whether it be for personal reasons, political reasons, or in some case just plain greed. To ignore these non-idealities and try to implement a system which won't be able to withstand this or even consider that perhaps not everyone is out for the common good, is in my view, rather shortsighted.

Anonymous said...

Think about drug development - it takes 10+years in some case to do this, and to go through a multitude of trials and FDA(or other gov't) approvals and when done the owner of the patent on the drug can go work for a generic drug company which doesn't spend a dime on development and merely takes the formula and manufactures the drug in short order.

There's a difference between licensing a patent and selling it (or a share of it) outright. Clearly any company that is going to invest millions in developing a product based on somebody's IP is going to put in restrictions on who the owner can sell it to, in the license contract. Or else buy it outright. If not, then that company would have to compete against all other comers and so would take that into account as part of their business decision process.

Tonus said...

"We could be talking about an engineer earning 80K/year and a patent that cost a company 200million to do development on."

This comment makes me think that the scenario that hyc is describing would be a real problem for companies. Many ideas, no matter how great they seem on paper, will simply not pan out. This means that under the system being described (assuming I am understanding it properly) a company assumes all of the risk involved in developing a technology, while only profiting to a limited degree when one of them pans out.

So company A sinks $300 million into one idea, and $500,000 into another. The first idea goes nowhere, $300 million gone. The second idea is a huge seller, and the company recoups its $500,000 plus a few million that they agreed to with the patent owner(s). The patent owner(s) then cut a much more favorable deal with Company A or leave for another company.

This works really well for the employees who work on ideas that they will own at some point, but not so well for the company that must sink lots of money into developing those ideas. I can't imagine that the person who owned the patent that had $300 million invested in it would be expected to pay it back.

When the risk/reward equation becomes that lopsided, I think it actually severely restricts development and design of new ideas. Better to let your competitor spend the money to develop an idea, and then swoop in with a sweetheart offer to anyone who has a viable patent. The real winner are those "second tier" companies who live off of the work of others.

SPARKS said...

Gentlemen, my practical personal exposure to IP rights seems to be very limited. In the IBEW our foundation is teamwork. Perhaps there isn't enough 'intellect' or personal 'property' to warrant an issue, we are in essence a collective by pure definition.

However, this extremely informative and interesting philosophical debate, the individual vs. the collective, calls back to an author by the name of Ann Rand.

Forgive me if you are familiar with the works, 'The Fountain Head' and more poignantly 'Atlas Shrugged.'

May I suggest these reads as a supplemental/alternative perspective for your personal study of this clearly profound debate.

Thank you for the glimpse into an interesting facet of corporate American and its relationship with the individual and the individuals rights. Excellent, to say the least, most impressive.

SPARKS

Anonymous said...

Better to let your competitor spend the money to develop an idea, and then swoop in with a sweetheart offer to anyone who has a viable patent.

Exactly... I think In HYC's world the patent owner would be so noble and loyal that they would not leave for the money (or for some other reason) and companies would not recruit talented people who at one time had worked for a competitor. While this is a great ideal to envision for the world... it's not the world we live in today.

Innovation may, in an ironic sense, be somewhat diminished as companies are less willing to bet on the big idea and only willing to fund "more likely" or lower cost projects. This leads to progress through evolutionary ideas but discourages revolutionary or inflection point type ideas from being worked on. You already have been seeing this in research at major companies.

Given the dry up of VC money (there still is money, but not nearly as much as there used to be); who will fund the work for the truly revolutionary ideas, especially if the upside might be limited as a patent holder may just walkaway and go to a competitor at some point in time.

hyc said...

Now you are just meeting antagonistic.. what if you had a sick family member and wanted to work close to where they were located? What if your boss was a complete jerk? What if you found there was no upward mobility in your current company? What if you were getting married and moving to where your wife is from? What if you needed a better salary to pay off a debt or cover some medical bills? Come on now... people move jobs all the time and believe it or not it can be for non-financial or non-greedy reasons. IP ownership should not be a dissuading factor for attempting to change jobs just in case the guy is greedy (what if it's not a case of greed?)! IP should not be used as a tool to disincentivize people from moving jobs by creating a "hassle" to basically filter out people who aren;t committed to wanting to change jobs - that is just plain idiotic.

You're right, it shouldn't be used as a tool as you describe. But most of your examples are pretty flimsy. Usually, if you want to take care of a sick family member, you take a leave of absence and return, you don't quit/change jobs. And remember, if you have valuable IP, it's in your company's interest to make it worth your while to stay. If your boss is a jerk or you've hit a ceiling, then first the company should make a good faith effort to improve your situation, otherwise, leaving is appropriate.


And AGAIN YOU MISS THE BOAT on profits... the IP owner, if he is working for a company might not see a single cent in profit when he needs to or wants to move jobs. His patent may still be years from getting to market... what then? You wave around these vague notions of quarterly payments but who's paying them? We could be talking about an engineer earning 80K/year and a patent that cost a company 200million to do development on. If that product hasn't gotten to market when the engineer wants to leave, what quarterly or yearly plan is he going to pay these costs back with?
Or is he stuck with the current company, through the hassle of the new IP model? Or is he forced to sign it over in exchange for his freedom to move (of course in your world companies cannot own IP so I don't know who he wold sign it over to)? And before you say ot could just be future profits for the company what if the company doesn't recoup cost with the profits (and yet the new employees company eventually turns a tidy profit from it?)


No, you missed the point. Repayment comes from profits. If there are no sales or profits, then there is no repayment. If sales don't start until 10 years after the invention, then repayments don't begin until then. Obviously you can't pay someone from money that hasn't been received yet. In the entertainment industry, actors get annual royalty payments whenever one of their films is shown in a given year. If it's not played, there's no payment.


[The numbers would be agreed on at the outset of the project. As such, Intel would be sharing data with a valued employee, not a disloyal ex-employee.]

Again, I find it hard to believe you actually work in the real world. You file a patent application, it will take well over a year to even know if it gets approved. And the employee and the company are suppose to pre-align on the outset of the project? What if their is scope creep or cost creep? Is their a quarterly/yearly audit to keep track of these things?


In any well-run project these audits are already being done.


How many patents does a company like IBM or Intel have every year? Several thousand? And in your world, the company is aligning with the patent holder on a # for EVERY ONE of these patents? Now how many PATENT APPLICATIONS? (times 2 or 3?) Since the projects often start long before the patent ruling is in, you are talking about aligning on EVERY PATENT APPLICATION in the event an employee leaves. And I'm ignoring the mutiperson patent which is another complete minefield in your scenario)


Again, why would the company engage in the first place, if they didn't have an idea of how important the project will be?


And your notion that ALL data should be shared with the employee is laughable... you are confusing data on the technicals of the project with specific cost data that a company may not want a lot of people knowing (and before you think this is evil read on). In this case Intel will share all the salaries of people on the project with the patent holder (I'm pretty sure there may be some privacy regulations), all the capital tool costs (this is not something an engineer may typically knows), all the high level fab costs, contracted wafer costs, expected profit margin on every part that this process might be used on (which of course don't exist as the process is probably 3+ years out). This is not about loyal/disloyal, you are talking about several thousand people where this data may now need to be shared with... While in utopia all information can be shared with everyone, in today's companies not all data is shared with all employees.


Then today's companies are broken. I used to work for the State of Michigan, way back when... By state law, all government employees' salaries are public info, and they're printed in the newspapers every year. So again, yes, *in the real world* this is already being done.

In the case of government, the public has a right to know how their tax money is being spent. In the case of a corporation, the shareholders have a right to know. When every employee is also a shareholder, the rest follows.


And again we are talking about several thousand patents a year in some companies so this is a process that you fully expect to be done up front for each patent, or even worse each patent application (as you don't know which ones will eventually be approved). Are you staring to see the absurdity of some of your suggestions/solutions?


Many of these things have already been done. As such, I find most of your objections invalid.


Part of the problem is you expect everyone to be working toward a common ideal - that may not be the case whether it be for personal reasons, political reasons, or in some case just plain greed. To ignore these non-idealities and try to implement a system which won't be able to withstand this or even consider that perhaps not everyone is out for the common good, is in my view, rather shortsighted.


I think you can arrange so that common good and self interest are aligned a lot more closely than they are today, and thus minimize the impact of the outliers you're identifying. Even with the Prisoner's Dilemma there are equilibrium scenarios where cooperation is the best strategy. If people aren't playing to cooperate they're only going to hurt themselves. Eventually they'll self-select themselves out of the system.

Anonymous said...

HYC... time to stop looking at trees and start looking at the forest... you remind of Scientia in attempting to attack every single point in a vacuum but ignore the overall argument. A few examples of your complete loss of perspective:

If your boss is a jerk or you've hit a ceiling, then first the company should make a good faith effort to improve your situation, otherwise, leaving is appropriate.

Again you are mindlessly stuck on how things SHOULD be as opposed to how they are! Not every single person/company is as noble as you'd like them to be. You can bury your head in the sand and say it SHOULDN'T be that way, but it is. And some people cannot afford to take a leave of absence to take care of a family member and will move to the location and look for a job (you know, kinda to pay the bills?). You seem to refuse to acknowledge there are a cicrumstances where a person may move jobs for non-financial reasons or may move jobs for less than noble motives?

Your perspective seems to be how you would like things to be but you give no actual consideration of how things exist today. Either that or you choose to do things this way so you can refute a point and continue to think your IP plan is anything less than fantasy land.

Even with the Prisoner's Dilemma there are equilibrium scenarios where cooperation is the best strategy.

Please don't talk game theory, it is clear from your commentary that you live in a bubble world and don't acknowledge that some of these "outliers" are actually more than just outliers. You know what, I'll let you in on a little secret, there are some circumstances where people choose not to cooperate, where CEO's are irrational or vengeful, where people may solely do what they think is best for them and not what is best for the world or the Borg collective. I know this is all crazy talk, and it SHOULDN'T be this way, but they are (and these are not just mere outliers)

By state law, all government employees' salaries are public info, and they're printed in the newspapers every year.

Fantastic... ummm.... you are talking about GOVERNMENT/PUBLIC jobs, you idiot! Can you give me a SINGLE EXAMPLE where salaries of ALL employees of a private company are listed? Let me know where i can get the salary of every employee for AMD, Intel, IBM, TI, etc (I'm not talking Sr mgmt which is required by law, but lower level employees). ARE YOU SERIOUSLY MAKING THIS ARGUMENT!?!?

Many of these things have already been done. As such, I find most of your objections invalid.

I've worked on programs where there have been maybe 20-30 patent applications. As a program manager, I can tell you the total cost of the program, but you again are not living in the real world if you think you can say patent application A is costing us 50,000 of the 20million program, application B is 100,000. There is just not that level of granularity. While you again can go back to your bubble and say it SHOULD be tracked that way, it is not and would require ENORMOUS resources and likely a loss of productivity) to do. For a company like IBM you are asking them to track costs on individual patents (not programs, but patents as often there are a large # of patents for a given development project) when they have on average ~40 applications per WORKING DAY. Again this is a case of you WISHING and WANTING THINGS to be a certain way, or taking an overly simplistic example of a POSSIBLE SCENARIO (maybe a company that has 1 or a handful of applications can track it painlessly) , but it is not representative of the complexities that exist in the real world.

In the case of a corporation, the shareholders have a right to know. When every employee is also a shareholder, the rest follows.

Actually this is completely wrong (unless this is an "HYC" perceived right). By law, shareholders (and actually not shareholders but the public) have a right to know what the salaries are of upper levels of management. You may be talking now about a PERCEIVED RIGHT, or HOW YOU THINK THINGS SHOULD be (again), but if you could point me to a single law, SEC regulation or any other item that defines knowledge of every salary as a shareholder right. I have some AMD stock - can you let me know how I can get a listing of every employees salary? I would like to exercise this stockholder right you refer to.


No, you missed the point. Repayment comes from profits. If there are no sales or profits, then there is no repayment. If sales don't start until 10 years after the invention, then repayments don't begin until then. Obviously you can't pay someone from money that hasn't been received yet

Fine let's use your method... just curious what happens when there is no profit for 10 years, but the employee leaves and goes to another company (I know this whole 'leaving a company' is just crazy talk in your world, but it does happen). Is the employee allowed to uss HIS PATENT at the new company? Does the new company now have to foot the employee's bills to the old company since there are no profits yet? Or does the new company just get a free ride on the other company's development work to date? THAT IS (ONE OF) THE MAJOR HOLES / FUNDAMENTAL FLAWS you keep ignoring in your system and is one that will discourage companies from working on developing long term, high cost patents (and an inventor will have a hard time get financing for)

Do you think there would be significant drug development if as soon as a new drug was proven, an employee could jump ship (I know in your world the employee would never leave or the company would discourage it). He could then take the invention to a new company who could mass produce the drug without having to invest in that pesky validation work? (Again, I know a competitor would never ever ever do such an evil thing, but I'm crazy this way and consider this a distinct possibility)

The problem with this discussion is you are basing your system on how you think things people should or companies should act, instead of appreciating/considering how they do act. You keep saying people shouldn't leave or they should be loyal, or companies should reward them more... but that is not always the case. The way you have devised your system it would only take 1-2 outliers to completely exploit your system as your IP proposal is not tolerant to anything but ideal behavior. I'm struggling to see how you cannot see that - you seem unwilling to accept that not every company or person will behave ideally or only perform for the collective good.

To wrap up - would I like everyone to behave themselves? Absolutely! Would it be nice if patent holder could always own their own IP? If people behave ideally, then yes. But considering how EASY it would be for a small minority to exploit this system and MORE IMPORTANTLY how much damage that could do to good companies which act in good faith with goo intentions and employ thousand of hard working people, it is lunacy as many of these good companies could be crippled (which in turn would mean the people employed would also get the brunt of it through layoffs, or a decrease in the value of shares they now own in their own company). To ignore this is to ignore reality.

hyc said...

Again you are mindlessly stuck on how things SHOULD be as opposed to how they are! Not every single person/company is as noble as you'd like them to be. You can bury your head in the sand and say it SHOULDN'T be that way, but it is.

Yes, my entire objective here is spelling out the way I think things should be, and stating that the way things are today sucks.

You keep hammering the point that people change jobs frequently, for varied reasons. I acknowledge that this is true today, but I don't accept that it must always be true.

In the 1950s it was still common for most people to be "lifers" at their particular place of employment. Somewhere along the line it became unfashionable for companies and employees to remain loyal to each other, but loyalty did count for something in the past. My belief is that it's possible to find the causes for those changes and to fix them.

It starts by giving employees more than just a gold watch as a reward for long service. It starts by not firing long-term high-salary employees and replacing them with cheap fresh graduates. It means structuring the business environment such that it is detrimental for companies to treat employees as interchangeable faceless drones, and detrimental for employees to blithely job-hop every few months. It means creating tangible benefits and rewards for preserving loyalty and tangible costs for disloyalty.

hyc said...

Fine let's use your method... just curious what happens when there is no profit for 10 years, but the employee leaves and goes to another company (I know this whole 'leaving a company' is just crazy talk in your world, but it does happen). Is the employee allowed to uss HIS PATENT at the new company? Does the new company now have to foot the employee's bills to the old company since there are no profits yet? Or does the new company just get a free ride on the other company's development work to date? THAT IS (ONE OF) THE MAJOR HOLES / FUNDAMENTAL FLAWS you keep ignoring in your system and is one that will discourage companies from working on developing long term, high cost patents (and an inventor will have a hard time get financing for)

I've already answered this question, many times. Yes, the inventor can change jobs and use his IP at the new company. The debt obligation doesn't change; if the new company starts selling a product with this IP then they don't get to keep any profits until the original company is paid off.

The reality is, for any complex new technology, the original company is still going to have a sizable head start on everyone else acquiring the IP for the first time, and all of them will still owe their license fees to the original company. As such, very few companies are going to want to play this acquisition game. The company that pays the R&D costs recoups the costs regardless of where the inventor goes, so in that respect they're protected. And they'll be first to market, so they get the most value.

Anonymous said...

Simple example. I take a patent on a drug to cure cancer. I work for Merck, they see the value (and the public good) of the patent and spend billions developing the drug, doing the trials and we are enter the final stage on FDA approvals. I'm excited because as the patent holder I'm going to have a huge impact on humanity and also make a boatload of cash (as I should as the inventor - after of course the company gets their expenses back). I know that the company will be making $2/pill profit and spent $3.5 Billion (because in your system they have told me exact costs as per the upfront negotiation). I don't get too greedy as the company has gone into significant debt to fund the development, so I agree to keep only 1/2 the profit on every pill after they get paid for development expenses (and the company gets the rest to spread among the employees and fund future development work).

I find out Bristol Meyer has a much more efficient manufacturing process where they can make the same pill at 2/3 the overall price to the public and can make $4/pill profit - I contact them and they offer me a job and the same 1/2 the profits deal I had with Merck.

So I decide to go to BM; some may call this greedy, but I'm doing this for the collective good - they are selling the pill at a lower price so the drug will reach more people. And after an attack of conscience, I even keep the same $1/pill profit (meaning 1/4) of the profit) because I'm vain and I don't want people to think I switched companies out of greed. the company again gets the remainder (in this case 3/4 to spread among the employees and fund more development)

The drug finally gets approved, and of course Merck doesn't sell a single pill (as they are more expensive). BM eventually sells enough pills where I can send all the development money to Merck so they break even on the cost (per the HYC patent system requirements). BM (and me) of course make money hand over fist and society benefits because they now have gotten the pill at a cheaper price and it therefore can reach more people at lower overall cost to society.

So is this the optimal solution, right? The product is brought to market at the lowest price and therefore reaches the most people. Per the new IP system, Merck recovers their intital costs. And as the patent holder I get a good share of the profits (in fact if I wanted to I could have been greedy and gotten more, but chose not to).

See anything wrong with this picture?

Does Merck ever invest in developing another patent again? If not, their business fails and all those employees are now out of work. Why would Merck ever do this again if the patent holder can go to a competitor and then at best Merck may break even.

Some may say, well Merck was uncompetitive in manufacturing and thus should have failed... this would of course lead us to monopolies where the cheapest to manufacture something wins. This also has a nasty side effect of encouraging/enabling countries or companies where wage structure and employee treatment may be lower than other countries or companies so costs can be kept lower.

So I'll wait for the theoretical arguments...

- Theoretically I shouldn't have left Merck out of loyalty? (but why not? my goal was to reach as many people as I could with the pill at the lowest possible price, which I accomplished)
- Theoretically every company should have similar cost structures and there shouldn't be the disparity that I mention (unfortunately not all companies/countries operate under the same minimum wage requirements or environmental standards or tax systems)
- Theoretically my own company could have offered me more of the profit (but again greed/profit wasn't my motivation, my main reason for leaving was so I could produce the pill cheaper and reach more people)
- Theoretically all the Merck people out of work can find new jobs with zero impact to their lives (they'll find a job of similar salary without having uproot)
- Any other theories?

Going down this road you now have an IP system designed to generally reward the cheapest (most efficient manufacturer), instead of also valuing the innovation and development. This (in the extreme) will lead to minimal competition and potential less innovation. There are PLENTY of industries where time to market is not that severe and being first to market is not a significant advantage

InTheKnow said...

The Pentium was originally released at 60MHz, and even that design had thermal problems.

It did have issues, but they fixed them and got the architecture up to 200MHz eventually. The same can't be said of the MC68060. It didn't scale up well and I see that as a major flaw in the architecture.

You said "well written code", but the actual point is FPU-intensive code, where "mixed" meant mixed integer and FP. The majority of code they targeted was obviously int-heavy. Unless you were heavy into scientific computing, the FPU difference didn't have any impact on users.

Let's look at the quote in its entirety, shall we? From your source the full statement was...

Against the Pentium, the 68060 could perform better on mixed code, Pentium's decoder could not issue and FP instruction every opportunity and hence the FPU wasn't superscalar as the ALUs were. If the 68060's non-pipelined FPU could accept an instruction, it could be issued one by the decoder. This meant that optimizing for the 68060 was easier, no rules prevented FP instructions from being issued whenever was convenient for the programmer other than well understood instruction latencies. However, with properly optimized and scheduled code, the Pentium's FPU was capable of double the clock for clock throughput of the 68060's FPU. (emphasis added)

I read that to mean that the FPU, was the weak point of the Pentium. So the weakness was the FP performance not the integer performance. And you just said that integer was the more important for most users. Doesn't that render your claim invalid?

The way I read the statement was that even though the FPU pipeline had issues, if the code was well written that weakness could be offset.

Even if the MC68060 were better clock-for-clock than the Pentium (which I don't think the statement above says), the Pentium was able to out-muscle the MC68060 with higher clock rates.

So far I've seen nothing to convince me that the Motorola chips didn't have process/design flaws that Intel's offerings exploited to offset Motorola's advantages.

hyc said...

Ok. I have no answer to that.

I'm not sure the scenario you described is actually worse than where we are today though. All of our consumer electronics are already manufactured in China/Taiwan/etc. for cost reasons. People here have continually been losing their jobs as they got outsourced overseas, so how is this worse?

From a completely different angle - it's clear you believe the current system we have today is better than what I'm describing. Do you believe our current system has problems that need to be (and can be) fixed?

Sparks: "In capitalism, man exploits his fellow man. In communism, it's the exact opposite."

Is what we have today really the best we can do?

InTheKnow said...

To wade into the fringes of the patent debate, I'm with Tonus on this. I don't see any reason for a company to put out anything to do the research in the first place.

IP moves freely from company to company with the employee, competitive advantage is fleeting, and there is no incentive for long term investment on the part of the company since they bear all the risk.

It becomes incumbent upon the company to keep the patent holder happy after they have provided all the resources to make the patent a reality. The inventor/patent holder risks nothing. I just don't see this as equitable. Great for the creative genius, but not for the company (read all the other employees).

Under this system I see a condition which in the long run must inevitably lead to a single source manufacturer for any given item.

And when you have only one employer, personal choice is limited and incentive to improve the product is gone.

As so often seems to happen, the desire to redress a wrong goes too far the other direction and creates a new set of "wrongs".

Dare I be politically incorrect and hold out affirmative action as an example?

hyc said...

InTheKnow: I'll concede the point. Motorola switched focus to the PowerPC at that point, they must have had a strong reason to switch to RISC and not to keep developing the 680x0.

But, with Intel's focus on the speedup of the Pentium FPU, they also made a serious screwup.

http://en.wikipedia.org/wiki/Pentium_FDIV_bug

http://www.cs.earlham.edu/~dusko/cs63/fdiv.html

InTheKnow said...

But, with Intel's focus on the speedup of the Pentium FPU, they also made a serious screwup.

Worse than that, they not only had an issue, but Andy Grove tried to tell the great unwashed masses it wasn't a problem (at least for them). Right or wrong, Grove's statement was a terrible PR move if ever there was one.

And they paid dearly for it. I believe they ended up offering a replacement chip to anyone that wanted one.

Of course I don't recall ever seeing anyone accuse Grove of being tactful. :)

Anonymous said...

blah blah blah

In the end the epilog is the same. INTEL wins the battle and only one making money. Nothing else really matters does it. They are the only ones with the money and motivation to keep pushing. Others have no money and thus it matters not what great ideas there are without billions and billions and the factories and connections to sell hundreds of millions even the likes of IBM, Moto, TI and the Japanese failed. Sorry AMD backed by the arabs don't have a snowballs chance in hell either

Khorgano said...

hyc said...

Ok. I have no answer to that.

I'm not sure the scenario you described is actually worse than where we are today though. All of our consumer electronics are already manufactured in China/Taiwan/etc. for cost reasons. People here have continually been losing their jobs as they got outsourced overseas, so how is this worse?

From a completely different angle - it's clear you believe the current system we have today is better than what I'm describing. Do you believe our current system has problems that need to be (and can be) fixed?

Sparks: "In capitalism, man exploits his fellow man. In communism, it's the exact opposite."

Is what we have today really the best we can do?


I think HYC is a reasonable guy and making an honest effort in his argument, he's done a decent job explaining himself, although, some of his details could have been spelled out earlier to guide the conversation. However, I think you are confused on this point. Yes, this system we have isn't perfect and on the surface the ends may resemble the exact thing we're arguing against (Labor/work moving to cheaper countries), but you have to realize, that this is only because we DON'T live in a true capitalist society.

In fact, we really haven't since the early 1900's and the creation of the federal reserve. Ever since the Great Depression and FDR crapped on the constitution, we've been more like a neomercantilist, interventionist corprate welfare state than a free market. This is has become even more clear in the current economic environment and the notion that the gov't can "stimulate" us back to health, but I digress.

If we were in a truly free market relying on sound monetary policy, the shifts in labor/cost model's to other countries would never have happened.

To reiterrate Sparks, you really should check out Ayn Rand and might I suggest Murray Rothbard.

hyc said...

InTheKnow said...
It becomes incumbent upon the company to keep the patent holder happy after they have provided all the resources to make the patent a reality. The inventor/patent holder risks nothing. I just don't see this as equitable. Great for the creative genius, but not for the company (read all the other employees).


You're forgetting something: The inventor is also, like every other employee, a major shareholder in the company. Again, it would take pretty extreme circumstances to make it worthwhile to damage his own company. His risk is proportional to how valuable he is to the company.

If "just breaking even" is the issue, then we change the payoff - the originating company is owed 2x their cost. 100% return is a pretty good ROI...

hyc said...

Interestingly, google turned up rumors that the 68060 was prototyped at 200MHz in 1993. And it appears there's a group preparing to resurrect the 68060 design.

http://www.natami.net/68070.htm

Seems a bit odd to spend all this effort on a new 32 bit machine today; but all the same I'm curious to see what it would be capable of.

hyc said...

Khorgano: thanks. Ok, I'm off to read.

hyc said...

RIP Transmeta

http://www.theinquirer.net/inquirer/news/703/1050703/transmeta-patents-trolled-company-dies

SPARKS said...

Sparks: "In capitalism, man exploits his fellow man. In communism, it's the exact opposite."

"Is what we have today really the best we can do?"

'Man exploiting capitalism,' is precisely why the former Soviet Union, and its puppet satellites, failed so miserably in the later half of the 20th century. They destroyed incentive and pride.

I must emphasize and reiterate these sociological theories are constrained by their own definitions. In PURE communism there is no CAPITAL, per se. The term as applied in the quoted statement is referring to a communist regime definitely not the communism preached by Christ (the purest form) and the one we enjoyed in Roddenberry's 'Star Trek'. Money is abandoned. It has no value in favor or a pure collective. In fact, it is frowned upon with disgust, "profit!", Piccard once exclaimed of the Ferengy in one episode.

John Lennon's 'Imagine' is a wonderful utopian dream by one of the most commercially successful and critically acclaimed artists of our time, yet he realizes the limitations first hand by the line "You May Say I'm A. Dreamer."

Therefor what have we? We have these ideoligal dreams, when put itno practice fail miserably. Why? Is that all we have, perhaps? But, what we have it ain't bad. Our American Capitalist/Democratic society works in practice! And, as this debate has proven, there are flaws. Sure, the system needs a tweak, maybe like your perfect musical score or lines of code, but it's the best machine we've got.

Again, I must stress tolerances and variables. Would you have a better shot at perfect piece with a Stradivarius? Maybe. Would Itzhak (Isaac) Perlman be able to do a better job, of course. Therefore, do you put the VALUE on the 'Strat' or on Perlman? (A wealthy owner donated the instrument to Perlman. Now that's getting your money's worth)

More to the point, from my own experience. As a union workers we are the closest you can get in this country to a working socialist class. However, within those ranks there are exclusions, unwritten ones that allows TOLERANCE for the outstanding employees. Is one worker, better than another? Yes, Have I carried another worker who couldn't perform his job as well, yes. Have carried a worker who wouldn't perform his/her job, NO. We strive for equality, we demand effort and we work within those tolerances. Do I get the credit? ALL THE TIME, by my supervisors and the Owners! Do a get more money? (read: capitalism here) Yes, but nothing that going to get me a 54 foot Bertram. Should I get a percentage of the owners profit (I usually turn between 30 and 40 percent net) If he's giving, I'm taking. IF NOT, so be it. That's the key. I'm not expecting anything. If you work hard, and your good, the money will come.

I call it approach, attitude, and aptitude. "Triple A, baby!"

Fuck one of those up, somebody somewhere ain't gonna like you. That's a fact.

Here's the thing, and there no getting around this, I NEED the owner of the company make the most PROFIT. I WANT his company/our company to be strong. I HAVE to cover/help my brothers/sisters who may not have my abilities and knowledge. This is as close as I get to IP.

I don't want the owners headaches, his infrastructure, his payroll, his deadbeat customers (about 4M, I don't know he sleeps).

That's his job. He earns every penny of it.

We have a strong well respected healthy company. We have wealthy and successful owner(s). Finally, we have a good place to work, good wage/benifits, for someone THAT WANTS TO WORK AND CONTRIBUTE to support his/her family and OUR COLLECTIVE ENTERPRISE. Some guy/gal may not be the sharpest tool in the box, but you can bet your ass I'll find something they can do, AND BE PRODUCTIVE. What if he/she is really good? They'll be running their own work in a New York minute. I'll make sure they do.

That's my job. I earn every penny of it.

Now I don't know if this is capitalism or socialism. But, it's all I have, and I HELP MAKE IT WORK. I'm proud that it does.

Our owner hired 2 more estimators this week, in these times. He says he's planning for the turn around.

I'm right behind him.

SPARKS

Anonymous said...

Do you believe our current system has problems that need to be (and can be) fixed?

Of course it has problems. And no, probably not everything can be fixed... does that mean you throw out the system for one that is more flawed?

"Demcoracy is the worst form of government...except for every other one." Democracy has problems, do you get rid of it for a more flawed system?

I'm not sure the scenario you described is actually worse than where we are today though. (regarding business moving to the cheapest manufacturer)

You don't see an issue with this or this being worse today? I suppose then you don't think Nike using child labor to lower their production cost is a bad thing? Or elimination of any working standards (such as safety) in order to avoid costs having to upgrade facilities. You think outsourcing is bad today... imagine lowering the IP barrier even further and making it even cheaper for some places to manufacture (and exploit workers).

Anonymous said...

You're forgetting something: The inventor is also, like every other employee, a major shareholder in the company.

Define major? If you are 1 of 50,000 employees how exactly are you a major shareholder? (I'm also under the assumption there are outside investors or is that not allowed).

And if everyone was a shareholder, in the example I give - the employee would just be giving up his share in one company for a share in the new company. It doesn't change things at all as his new share would also be valuable (even more so in the example I give). This wouldn't be a disincentive at all.

Perhaps this person doesn't think about it as 'damaging' his company, but rather trying to do the most good by getting as much of his product out to help the world... surely that outweighs any 'damage' to a few thousand people?

If "just breaking even" is the issue, then we change the payoff - the originating company is owed 2x their cost. 100% return is a pretty good ROI...

It is now clear you are just throwing things out and are not really thinking them through. 2X might be fine or in the case of a really expensive investment it may completely swing in the opposite direction and FREEZE out all other competition. Again, when there are times when the IP (person) is moving or wants to move before profits are known. There may be now away of telling what the upside is... what if it is less than 2x?

If it is less than 2X, do you realize what you have just done? you have given all the profits to the company and now you have given absolutely ZERO to the inventor. Next you'll say we can mix this up and do some sort of 50/50 or 70/30 (or whatever) split until the investment and the 2X is recouped... what if it never gets there?

The problem is the uncertainty - the uncertainty in predicting profits, the uncertainty in measuring the actual costs (not matter how many accountants you want to hire), the uncertainty for the company to take the risk, the uncertainty of whether the employee will want to move.

The system you describe may be better under ideal circumstances and if you have perfect information and perfect predictive capabilities (then you could implement game theory and assume people will make informed decisions to maximize value and an equilibrium will be achieved). The problem is those conditions don't (and perhaps never will?) exist and rather than attempting to understand these limitations and the (in)tolerance of your system to these limitations, you just assume these can be engineered out and/or fixed.

The IP system today may be much, much worse if we were operating under ideal conditions - but I believe it is far more tolerant to the flaws and imperfect conditions in the world today then the system you are proposing.

Anonymous said...

And one last major issue on this whole concept of recouping (or 2x'ing the costs) and I promise I'll stop. I like the discussion but I realize there is a lot of repetition and some may be getting bored with it.

AMD makes a K10 chip, they make a certain profit on it (OK, let's ASSUME they do!). There are hundreds (thousands?) of patents that apply to that chip. How could you assign a specific, individual cost to each and every one of those patents? Ok, we've been down this road - let's assume you somehow can; whether you hire hundreds of accountants to manage all this or put in some tracking system (running on 64bit SW, no doubt) that can magically deconvolute all costs.

One of those patent holders moves to Intel, they make their own chip the RoboCore i999(Tm pending), this chip also has thousands of patents that apply to it, including the one the guy left AMD with.

Intel makes a bunch of profits on the new chip... some probably due to the patent the ex-AMD guy brought, but a lot from the other couple of thousand patents that went into making and designing the chip as well. AMD of course should, and wants to be, reimbursed... but how? How do you separate out the profit on each Intel RoboCore chip that is specific to that one patent. If you don't and just take a percentage of Intel's profit on each chip (until AMD is recouped or 2x'd on the cost) then the system is potentially screwing over all of the other individual patent holders.

Suppose Intel complains, hey AMD already made back the development costs from their K10 sales? Why are we reimbursing them for costs they already made back? AMD says "no we didn't, trust us". Intel doesn't believe this (or is unwilling to take it on faith) and they ask to see all of AMD's finances so they can understand profit/loss on the K10 chip (only to better understand the intricacies of this specific issue and the reimbursement - of course they wouldn't dream of using it for competitive reasons!). This of course is not info that AMD believes they can (or want to) share so a third party gets involved and let the inefficiency and bureaucracy begin! The level of knowledge the 3rd party would need to even begin to understand costs and values, not to metnion the mountains of data from each company to analyze it, is incredible. Now picture needing this sort of capability across hundreds of different industries.

The drug development example, I think, shows the inherent hole/flaw in the system hyc proposes - but even if the system can engineer that specific issue out... that is just the tip of the iceberg in terms of complexity.

This trickles down everywhere - can you imagine the impact to venture capitalism dollars? (which is a huge driver of the money that drives the innovation wagon) A lot of the VC investments into some of these small startup companies are based in part on the value of the IP (obviously there are numerous other factors). This is extremely difficult to do in the current system with the company owning the IP. When you add in the complexity of the individuals owning it, you now don't just ask the question of how much is the IP potentially worth, but how much is it worth if the employees stay with the company and how much is it worth if one or several of the patent holders leaves the company along with their IP. And yas part of that you also have to evaluate the likelihood (probability) of the various folks staying - do any of them have health problems, personal issues that might cause them to move, etc...etc... I have no doubt that would impact the willingness of VC's to put money into startups.

hyc said...

[You're forgetting something: The inventor is also, like every other employee, a major shareholder in the company.]

Define major? If you are 1 of 50,000 employees how exactly are you a major shareholder? (I'm also under the assumption there are outside investors or is that not allowed).


I got the impression we were talking about a superstar in the company, one whose creative output was critical. If you have 50,000 employees and 10,000 of them are cranking out good patents every year, then sure, the influence of any single one of them is reduced. But also, I'd say that their impact on the company's wellbeing is also reduced, and the overall significance/risk/damage from one of them leaving is minimal.

[We never talked about outside investors in this discussion yet. I suspect that will also trigger another thread that most of you will dislike. In short, my system would allow outsiders to invest, but only employees can vote. Again, my belief is that outside investors tend to take too much control of a company and invariably destroy them. Control should remain with the people whose livelihoods depend on the success of the company. Investors are, by definition, tossing around disposable cash, so if they make a poor leadership choice and the company tanks, they will continue their lives pretty much unchanged. But all the workers at the company are screwed. If you keep the leadership within the company, within the ranks of people who understand the technology and care about developing it, I believe you'll get healthier companies in the long run.]

[re: 2X]
It is now clear you are just throwing things out and are not really thinking them through.


I concede the point. I didn't even want to mention changing the payoff because it would have been such an obvious bandaid attempt. But if the overall shape of the system was sound, then some fine-grain tweaking would have been OK.

[I'm not sure the scenario you described is actually worse than where we are today though. (regarding business moving to the cheapest manufacturer)]

You don't see an issue with this or this being worse today?


Yes, I see the issue, but the issue is already with us now and I didn't believe my proposal made that any worse.

It's also not an issue that can remain forever, because as more businesses move into a location, their standard of living raises and their costs eventually rise sufficiently to offset the advantage of outsourcing to them. This has happened multiple times in recent history and is already happening in India, with companies pulling out and moving to Brazil, etc. Eventually there won't be any un-developed markets left to tap, and everyone will just be better off using their local labor.


This trickles down everywhere - can you imagine the impact to venture capitalism dollars? (which is a huge driver of the money that drives the innovation wagon) A lot of the VC investments into some of these small startup companies are based in part on the value of the IP (obviously there are numerous other factors). This is extremely difficult to do in the current system with the company owning the IP. When you add in the complexity of the individuals owning it, you now don't just ask the question of how much is the IP potentially worth, but how much is it worth if the employees stay with the company and how much is it worth if one or several of the patent holders leaves the company along with their IP. And yas part of that you also have to evaluate the likelihood (probability) of the various folks staying - do any of them have health problems, personal issues that might cause them to move, etc...etc... I have no doubt that would impact the willingness of VC's to put money into startups.


IMO, this is exactly what needs to happen. In any small startup company, there's generally one (or maybe a few) idea man who was the main inspiration into forming the startup. So fine, in today's world, the company owns all the startup's IP. But if these key people leave, their health fails, they die, whatever, the company is hosed, IP or no IP. Ultimately my point here is that individuals are more important than corporations, and nothing makes it more obvious than a startup company.

Earlier we talked about how I left my previous company in 1999. (Sorry, I searched back but couldn't find the exact comment.) You asked if my former company was harmed by my taking my IP with me, and I never addressed the question. In this particular case, the project continued their work without me, there was nothing preventing them from using the IP we had already developed up to that point. But they never completed it, and never brought any product to market. Today that entire division is gone, it was downsized out of existence. My belief is that they never understood the concepts well enough to develop it to completion. We completed the product in my startup in the following 2 years, and sold it to a few large clients.

Again, my point is that IP itself is frequently rather pointless. What's valuable is the creative minds that generated it. Intellect is more important than intellectual property. If you don't hold on to the people who create value for you, if you don't give them incentive to stay, then you're missing the point and your company deserves to fail.

hyc said...

Going down this road you now have an IP system designed to generally reward the cheapest (most efficient manufacturer), instead of also valuing the innovation and development. This (in the extreme) will lead to minimal competition and potential less innovation. There are PLENTY of industries where time to market is not that severe and being first to market is not a significant advantage


Backtracking a bit, I think you're downplaying time to market too much. Today there's a lot of ED drugs on the market - Cialis, Levitra... But the one that defined the market is Viagra. It came first and got the most mindshare. (Hell, when I was in Ireland in 2001 this 60-some year old fiddler told this joke "the other day I tried to take a viagra, but it got stuck in my throat. - what happened? - it made my neck stiff" - that's serious brand recognition, and the value of that can't be underestimated. Nobody tells that joke about "cialis" or "levitra" ...) There are brand-name drugs in the drugstore, and frequently right next to them are cheaper generics. And yet, a lot of people still buy the brandname product even though the generic is chemically identical, and cheaper.

As such, I think your B&M/Merck example's outcome was bleaker than called for. Being first to market makes a difference, especially for medicines. You think a cancer patient is going to spurn a $4 cure today, knowing that 6 months later a competitor will be able to sell the pill for only $2? Hell no, they're going to buy the cure today, ASAP.

SPARKS said...

"If you don't hold on to the people who create value for you, if you don't give them incentive to stay, then you're missing the point and your company deserves to fail."

That's what I'm talkin' about, the value of your labor force. I couldn't give a flying frig if you worked at Cern smashing atoms to bits in the LHC, calculating their vectors, or sweeping the place to keep it contaminant free. Get the best people for the job, pay them well, and you will succeed, PROVIDING MANAGEMENT KNOWS WHAT THE HELL THEIR DOING!

I give you Wrector Ruinz, and his board, as an example of what NOT to do.

SPARKS

Anonymous said...

If you don't hold on to the people who create value for you, if you don't give them incentive to stay, then you're missing the point and your company deserves to fail.

Completely agree... but it is up to the company to do (or not do) this. Installing a system in an attempt to force this to work if the company doesn't want to, will just lead to failure eventually anyway; and the system you are proposing to fix this problem brings a bunch of other problems with it.

You make it sound as if some folks here want companies to dump on employees - I don't want that but there are somethings you can't regulate. If the company does not value the employees to begin with, trying to force this through IP management is not the answer and will not make it happen. What it will do is introduce a bunch of problems and logistical nightmares that will have to be born by companies that do value their employees. If there are 10 (or even 50) criminals in a neighborhood of 100, you don't put up baracades, barbed wire fences, install ID checks and impose a bunch of conditions on EVERYONE in the neighborhood in an attempt to change the habits/values.

And of course at the end of the day the inventor knows what he is getting into and has free will (at least in most places)... if you don't like the rules of the company as they pertain to IP, don't take a job there, or leave (like you did).

Tonus said...

AMD/VIA lose 0.9% share of the x86 processor market to Intel. LINK

Apparently, a good portion of this increase was not just due to Atom. According to Barron's, Intel's share of the server market grew 3%, from 86% in Q3 to 89% in Q4. Presumably this is the x86 server market. This is particularly bad news for AMD, as Intel hasn't really made its move into the x86 server market with Nehalem. I have no idea how Shanghai fits into those numbers, though it was released in Q4.

Anonymous said...

Tonus... you and the link are clearly mistaken - I have been told by good sources and that Atom is a failure, so I don't see how it could be responsible, data be damned!

BTW - Nehalem is not going to make a huge difference in the numbers as the unit volumes for server are small, so even a 5% swing in marketshare would not have a real huge impact on the overall X86 market #'s. Also the space it will likely to eat into the most is even smaller (4P+) in terms of server market compared to 1P and 2P. There is a lot of talk about HPC systems and scaling, but at the end of the day well over 1/2 the server market is 1P and 2P which Core2 has already put a dent in.

Anonymous said...

For a change of subject

big down turn and every company looking at costs and ROI, big blue is laying off in the chip business and talk is again surfacing of them spinning it off.

Why should big blue keep investing billions into a business that returns little? If they do what happens to the poor alliance and AMD.

The clock is ticking

SPARKS said...

I'm a little confused about this 1P, 2P, 4P thing. I realize the numerical means the numbers of processors on the board, sockets? I read/heard somewhere that the O.S. manufacture charges a customer a certain amount per processor. So, if I purchase an eight core Beckton and install it on a 1P board, could I a save a good deal of money by reducing the number of sockets in a current 2P or 4P space?

Sorry, if I didn't get it right. Like I said, I'm a bit confused.

SPARKS

hyc said...

Sparks: it's easy to be confused, because different vendors charge differently. Some charge per-socket, some charge per-core, etc...

SPARKS said...

Ok, since we're on the subject. Obviously, you write code. What exactly do you do when you write code for multithreaded applications? Is there some kind of "tag" (instruction), for the lack of a better term, that only addresses each specific core? Are there multiple 'tags' that address multiple cores? How do they split the loads. Is it more difficult to write code for multithreaded APS? If so, how much more 'labor intensive" is it percentage wise to recompile something from single threaded to multithreaded?

I recall one game, Quake 4, that was compiled for a single thread upon its initial release. Subsequently, on the release of the first patch, they recompiled the game to utilize multiple cores.

In fact some games like C.O.D. 4 detect the presence of multiple cores. The performance improvement, I might add, is substantial.

SPARKS

hyc said...

Ooo, that's the $64,000,000,000 question.

It usually takes a fair amount of source-level rewriting to take a program that was originally single-threaded and make it multi-threaded. Usually the result looks extremely different from the original code. It's not as simple changing a few flags and then re-compiling.

You can get a flavor of the work from this paper I wrote in 1991:

http://highlandsun.com/hyc/#Make

It's not talking directly about code, but the problems are the same. You start out with a linear piece of code that says "do A, then B, then C, then D, etc...". The fact that all of these parts occur in order A-B-C-D is purely incidental, that's just how things happened because you wrote for a single-threaded system. But in reality, some parts of the process *must* occur in a fixed order, and others don't need to.

E.g., maybe B depends on the result of A, so therefore B can only run after A, never before. But C is independent of A, so it can be started any time. And D depends on B, so it can only run after B completes (which also can only run after A completes).

The work involved in making single-threaded code into multi-threaded code requires that you identify which chunks of the process truly depend on preceding steps, and which are independent. As a first cut, you can toss the independent parts into their own thread, and presto, instant performance boost. This is the same work that my paper describes - identifying the real dependencies in a build procedure, and splitting out the things that can safely be done in parallel, and identifying which steps must occur serially.

The work in the paper is brain-dead simple though, because it doesn't have to consider tasks that are expected to occur simultaneously, and various other situations that arise in threaded code. But at least it shows you the magnitude of visible changes you might see in rewriting something.

It can be a staggering job, and sometimes it's better just to throw out all the original single-threaded code and start from scratch.

hyc said...

Oh, and as for "tags" - most application programmers don't have to worry about assigning threads to specific cores - they just say "create a new thread for me" and let the operating system figure out where to run it. Of course, the OS scheduler may not always choose the best core for a given thread (the Windows scheduler is notoriously bad at this, especially for a hyperthreaded CPU) and so usually the OS provides API calls that a programmer can use to be very specific. But this tends to make code very non-portable, so it will only run well on a single machine configuration, so generally, programmers shouldn't use these calls.

SPARKS said...

Brain dead simple? Maybe for you, a software guru! I had trouble with BASIC!

Anyway, that was excellent, thank you. Humph, it took a long time for the multicore/multithreaded aps to gain traction, more than a decade, obviously.

Now, one more $64 question to add to the $64B. Has anyone, besides MS with its poorly implemented "scheduler", writtened/designed/produced a program that would examine these order dependent lines in code to automatically recompile single threaded aps, gaining an instant performance boost, thereby eliminating the need to rewrite the entire application from scratch?

If someone did, in addition to logically and automatically tighten up the inefficiencies in poorly written code, wouldn't this have a very lucrative potential to programers/developers in all facets of the industry? I'm sure computers would be much faster and more reliable in identifying these 'order dependent' lines of code.



SPARKS

Tonus said...

I saw this story linked at Ace's forums.

IBM cut 1200 jobs in its semiconductor division. The article speculates about IBM's future and whether or not it will drop its IC division any time soon. While they may do that in the future, it doesn't strike me as something that they would do now or in the near term.

SPARKS said...

INTC sweetens the pot for ole SPARKS! Yes, Core i7 975 with the new (DO) stepping has hit the tabloids. (But a lot of you inside boys allready knew this, and you get the sons-of-bitches free. %( )With a tweak here and tweak there, they've pumped up the volume to 3.3 Gig. I'm willing to bet this thing will clock a bit better than i965.

I had a hunch this would happen. Besides, I never buy a first stepping. (I've learned a lot from you boys.)

BTW, I'm in on this one.

http://www.hexus.net/content/item.php?item=17069


Lets play.

Memory-

http://www.scan.co.uk/Product.aspx?WebProductId=928197

Mobo-

http://www.newegg.com/Product/Product.aspx?Item=N82E16813131352

Drives- (2X RAID 0)

http://www.newegg.com/Product/Product.aspx?Item=N82E16822136296

G Card-1 for starters.

http://www.newegg.com/Product/Product.aspx?Item=N82E16814143143

Yeah, that about covers it.

SPARKS

Anonymous said...

Raptors? No SSD's? I'm disappointed!

I just picked up one of the Samsung 640GB drive (a mere 7200 RPM) on the cheap from the egg and I guess that'll have to do for me, still waiting for the Penryn quads to cheapen just a little more. Some of us don't have Sparks' budget! :) (I'm just jealous)

hyc said...

I just swapped out an OCZ CoreV2 120GB SSD for a G.Skill Titan 256GB SSD in my laptop. The Intel's still are above my palatable $$/GB threshold, and the other brands are approaching performance parity. This Titan is very nice.....

Sparks, for your $64 question ... I don't think compiler technology is quite there yet. You talked about being able to use one of your circuit design programs and saying "give me X" and out pops a nice complete, usable circuit diagram. There's no such equivalent for code yet.

Back in the early 90s there were a lot of parallel supercomputer vendors and most of them had some form of auto-parallelizing compiler. But generally this stuff was aimed at the HPC world, and the HPC world uses Fortran. Fortran is a fairly inflexible language, which makes it relatively easy to write such automation.

But most general purpose code is still written in C, and C is, by its nature, a low level language that allows you to do pretty much anything conceivable to a given processor/machine. This flexibility makes it a bitch to auto-parallelize, because there's a lot of data-dependency knowledge that can only be determined at runtime. As such, very few vendors ever released good auto-parallelizing C compilers.

These days, multi-threading takes the place of parallelization, but the obstacles are still the same: identifying data dependencies that require serialization.

In Fortran you typically run across code that loops a calculation over a big array, e.g.:
x[i] = y[i] * z[i]
This is extremely easy for compilers to recognize and auto-parallelize.

If you have a different computation like so:
x[i] = x[i-1] * y[i]
the compiler also knows it *can't* parallelize it, because the information in any given iteration depends on the result of the previous iteration.

When you have a lot of computations of the latter form, you either find a way to rewrite it without the serial dependency, or you say "screw it" and don't get any multi-processor speedup on that code.

In C, there's a concept of a "pointer" which is a variable that can point to any location in memory. Pointers are a powerful tool, and used heavily. They can be used analogously to arrays, but unlike arrays, whose dimensions are known, pointers are unbounded. So, while a compiler can look at "x[i]" and "y[i}" and know exactly what memory and variables are being referred to, it can't analyze "*x" or "y->foo" because x or y could literally point anywhere in the machine's memory. As such, even common array operations in C are difficult to auto-parallelize, and for the more esoteric uses of pointers, forget it.

If you can't unambiguously identify what memory is associated with a given variable reference, then you can't unambiguously determine if it has a data dependency on any other code. And so, you can't automatically break the single-threaded code up into multi-thread-friendly chunks.

On the other hand, Java is a pretty popular language these days (ack ptui) and there are no pointers in Java. And, JVMs typically include Just-In-Time (JIT) compilers that analyze and re-optimize code at runtime, so it's possible for a Java program to automatically multi-thread itself. I don't know of any JIT compilers that do this yet, but I'm sure people are working on it.

It feels to me like a lot of the IP in this area, established in the 80s and 90s, disappeared when all those supercomputer companies folded, so a lot of the techniques are being rediscovered by the current generation of compiler writers. I was a gcc maintainer (GNU C compiler, open source of course) back then, for the m68k and i860, and I know that most of what I contributed to gcc is gone now; at the 3.x revision a crowd of x86 hackers came in and ripped out a lot of the old plumbing and started from scratch. (Fancy that.) So I figure it will be a while before we see something really competent here.

Tonus said...

I'm more interested in hearing how the new stepping affects prices, OCing and heat (in that order!). ~100MHz bumps in clock speed at this point are of little interest in and of themselves. But if this pushes prices on the other i7 models down a bit, I'll be happy. With motherboard and memory prices dropping and/or due to drop further, a Core i7 920/940 on the new stepping could be a real monster that doesn't break the bank.

SPARKS said...

"Raptors? No SSD's? I'm disappointed!"

So am I. %< hyc said it for me.....

"'The Intel's still are above my palatable $$/GB threshold, '


I look at mechanical hard drives the way 'Spock' looks at vacuum tube based nemonic memory banks as "stone knives and bear skins."

Don't get me wrong, I'm lusting after two of those INTC SSD's like a masseuse waiting to give Christina Aguilara upper abdominal massage therapy. Besides, like anything in this industry, once they start cranking out the bastards like Doritos, the mainstream hard drive as we know it will be relegated to the 5.4 inch floppy bone yard.

hyc- Brilliant! 'auto-parallelizing!' A new concept to add to my repertoire.

I've got to jump on board with you on this one. Those idiots at Microsoft should be all over this like white on rice, instead of pushing that fat slob bloatware called Vista, and its new lipstick on a pig, WIN7.

Take a ~ 15% performance hit with an OS change, and to add insult to injury, render my old apps (and hardware) obsolete!?!?! They've got multicore running mainstream and they're sitting on their hands!? What part of F**K YOU don't they understand? Even INTC corporate told them to shove it!!!

Powerful and flexible auto-parallelizing software would be a boon to performance given today's powerful multicore rockets like my beloved QX9770 which, incidentally, is tragically under utilized.

Thanks for defining in the 'software gap' for this hardware freak. Brilliant.

SPARKS

SPARKS said...

Hey, are there any INTC boys out there (who get free chips) willing to discuss the new (DO) stepping without compromising company trade secrets, or undermining older stepping sales? A link perhaps?

A very excited,

SPARKS

Orthogonal said...

Hey, are there any INTC boys out there (who get free chips) willing to discuss the new (DO) stepping without compromising company trade secrets, or undermining older stepping sales? A link perhaps?

A very excited,

SPARKS


Yes, we can get free chips, but unfortunately it's not a buffet. Supplies are limited and SKU's can vary wildly from one site to the next and from time-to-time. You'd have to know someone in assembly test to get ES's or other goodies on a regular basis. That's why you always see random folks on the net in Malaysia get a hold of them first. I just help build the damn things ;)

Tonus said...

"I just help build the damn things ;)"

Can't you steal a few parts and assemble it at home? :)

Orthogonal said...

Can't you steal a few parts and assemble it at home? :)

Yeah, I'm almost done with my C4 Sort facility in the basement.

Anonymous said...

HYC - wasn't Intel working on (or at least researching) a sort of add on compiler (not sure if that is the right terminology) to eseentially take compiled code (I think this was C) and attempt to parallize it/distribute it ober mutliple cores where possible?

I'm a bit skeptical of this - but I am pretty sure I something in the press about this (maybe 9-12months ago). From what I can recall it was in the early phases and nothing near actual demo (more at the conceptual stage).

SPARKS said...

"You'd have to know someone in assembly test to get ES's or other goodies on a regular basis."

(sigh) ES

Bro, I'm drooling. That guy would never buy lunch, I'd get it. Christ, I'd install towel warmers in all his bathrooms.

Hook up a new 60A Jaccuzzi in the backyard? No problemo. Hell, I'd trench the son of a bitch myself.

"goodies" (sigh)

SPARKS

Anonymous said...

Comment from Sept 27:

- AMD exercises option sometime in '09 (this will likely have to be after Asset Removal is implemented), but has to be before the option expires mid-09 (unless NY extends it)
- Assuming they break ground in late '09, figure about a year to get the shell (foundation, walls, major structural, facility elements). This puts the shell "done" in H2'2010
- The equipment move in is the schedule swinger. If AMD is really driving the schedule hard they can actually start moving in equipment while the shell is being finished up - though I would find this unlikely. Figure ~1 year to install and qualify pilot line equipment. Then figure another 6 months to certify the process. This would put production ready likely early 2012 best case.


Feb 4 - Actual AMD schedule (EEtimes)
March 17: Site clearing and grading begins
June 16: First concrete poured
July 17: First steel erected
July 2010: Building shell completed
July 2011: Fab completed
2012: Full-scale chip manufacturing begins

I was a little off on the start time (as at that point the Paula Abdul's had not given out the money and Asset removal was still "a secret" and in Dementia's world a natural extension of APM); but the rest of the schedule?

Whadya know? Looks like 1 year for shell (June'09-Jul'10), 1 year for tool move in (Jul'10 - Jul'11), and 6+ months (Jul'11 to sometime in 20102)for certification/revenue starts.

But what do I know, I'm just one of the fanboys on Robo's blog? (and proud to be posting here) Elsewhere, Abinstupid was projecting possibly late 2010 or sometime 2011 for production. I'm sure he will now claim that the lag behind his schedule is merely a strategic choice by AMD and if they wanted to they could go faster (you know like the "choice" to releases the 2.3GHz Barcies first)

Anonymous said...

AMD will be BK well before tool movein in Luther Forrest begins.

The plans are nothing but a wet dream.

pointer said...

Blogger Orthogonal said...
Yes, we can get free chips, but unfortunately it's not a buffet. Supplies are limited and SKU's can vary wildly from one site to the next and from time-to-time. You'd have to know someone in assembly test to get ES's or other goodies on a regular basis. That's why you always see random folks on the net in Malaysia get a hold of them first. I just help build the damn things ;)


Well, I am not sure those that in the assembly test can send out ES on their own will. I do not think that Intel allows that, unless those folks get it thru not so proper channel ...

btw, i am building Bloomfield base system very soon... :)

Anonymous said...

HYC - wasn't Intel working on (or at least researching) a sort of add on compiler (not sure if that is the right terminology) to eseentially take compiled code (I think this was C) and attempt to parallize it/distribute it ober mutliple cores where possible?

I'm a bit skeptical of this - but I am pretty sure I something in the press about this (maybe 9-12months ago). From what I can recall it was in the early phases and nothing near actual demo (more at the conceptual stage).


I think I read that as well ... not by Intel ... but some other company (maybe with Intel support) ... which claim to make singlethreaded apps faster in multicore CPU ...

hyc said...

HYC - wasn't Intel working on (or at least researching) a sort of add on compiler (not sure if that is the right terminology) to eseentially take compiled code (I think this was C) and attempt to parallize it/distribute it ober mutliple cores where possible?

I'm a bit skeptical of this - but I am pretty sure I something in the press about this (maybe 9-12months ago). From what I can recall it was in the early phases and nothing near actual demo (more at the conceptual stage).


I don't remember reading any announcements like what you're describing, but I wouldn't be surprised. Intel's compiler team is certainly top notch; it would make sense for them to be pushing in this direction. It would be stupid not to...

But it's no cakewalk. My SOP when I first got my hands on those multiprocessor compilers was to examine the assembly code they produced, and then recode it by hand. There's a lot of information about the dataflow in a program that a programmer knows, that can't be encoded into the program source, and the compiler is bad at guessing.

[another rant, feel free to ignore]
These days it's even worse; they've added restrictions to the syntax of C99 that didn't exist before, aimed to reduce data ambiguities, to give an optimizing compiler a better chance at analyzing the data flow. But these same restrictions also prevent you from writing what used to be perfectly valid C code, and they prevent you from writing optimal source code. So now you're *forced* to write less efficient source code, and rely on having a good compiler+optimizer to trim out the inefficiencies, instead of just relying on *being a good programmer in the first place*. Needless to say, this pisses me off immensely. Catering to the lowest common denominator ... A genius programmer can write a program suitable for anybody to use, but a mediocre programmer can't write code suitable for a genius. We shouldn't be making it easier for mediocre programmers to write code. We should be training them to be better, or telling them to pursue a different line of work...[/rant]

SPARKS said...

Core i975 is blasting away all world records. Indeed, I was right (thanks Guru, the mega gurus at INTC tweaked the recipe with the second stepping, (DO). Man, are they on a roll or what?

FUGGER, overclocker extraordinare, clocked it to 5.259 Gig.

http://www.xtremesystems.org/forums/showthread.php?t=216495

http://www.fudzilla.com/index.php?option=com_content&task=view&id=11805&Itemid=1

I'm in.

SPARKS

SPARKS said...

Now that INTC has put the icing on the cake in the desktop segment, they are now going to souflee and flambee the enterprise segment come Feb 8.

INTC is giving a new meaning to execution, and it ain't pretty to the competition.

hyc, 8 cores 16 threads, all at 64-bit, right up your alley. HPC on a single die? Perhaps?

They also mention the Boxboro-EX chipset, very hush-hush, kicking some serious ass at 128 threads.

The server assualt has begun in earnest. OMG! AMD will have a new FAB up and running by 2012! (with part of the 5K+ I paid in NYS income tax). What will INTC do? Estimates may vary, but it looks like INTC is in trouble now!

So am I, as I watch my hard earned money (over a hundred bucks a week) go right down the fucking toilet, or the middle east. The choice is yours, same thing, really.


http://www.hexus.net/content/item.php?item=17103


SPARKS

Anonymous said...

Hyc: I just swapped out an OCZ CoreV2 120GB SSD for a G.Skill Titan 256GB SSD in my laptop. The Intel's still are above my palatable $$/GB threshold, and the other brands are approaching performance parity. This Titan is very nice.....

I'm considering an SSD boot drive for my next build, but am a bit concerned over the longevity of SSDs compared to hard disks. I know flash memory is only good for a certain number of write cycles, and that the SSD firmware tries to spread the writes evenly across the memory addresses, but still I'd hate to see my SSD lose a significant part of its capacity after a couple of years usage :). You have any idea on how long one lasts under average use?

Sparks: INTC sweetens the pot for ole SPARKS! Yes, Core i7 975 with the new (DO) stepping has hit the tabloids.

I was back to waiting on Westmere when I saw your post. Now I'm waiting for June or so to see how much a D0 stepping of the i920 goes for :). If I can overclock it to 4GHz on air and stock core voltage, then I'll be building this summer instead of next winter.

Anonymous said...

Well our favourite UAEZone moderator Sciatica has another whopper of a marketshare post up:

For example, let's compare graphic chips Q3 2008 to Q4 2007. Why are we comparing different quarters? Because Q4 is usually the highest quarter of the year but in 2008 the sales were unsually high in Q3 and much lower in Q4. So the cross comparison should be more accurate. These numbers are for all GPU's including chipset, discrete, and mobile.

I guess he didn't like the Q4 numbers too much. Unfortunately, his analysis is of historical interest only, until such time as the economy recovers. Certainly it's irrelevant to extrapolate "good times" trends when you're in free-fall off a cliff. Maybe he got a Pollyanna Twoshoes hot air balloon for Christmas.

Total GPU's for Q4 2007 is 100.50 Million. Total GPU's for Q3 2008 is 111.26 Million.
Total volume growth is 10.7%.
AMD: 18.36 -> 22.90, 24.7% increase
Intel: 43.71 -> 54.95, 25.7% increase
nVidia: 33.78 -> 30.93, 8.4% decrease

So, while AMD and Intel saw similar growth, nVidia lost share. This is pretty much what I expected. However, let's compare GPU's sold to CPU's sold.

In Q4 2007 AMD sold 96.5 GPU's for each 100 CPU's it sold, In Q3 2008 it sold 133.9 for each 100 CPU's.
In Q4 2007 Intel sold 69.2 GPU's for each 100 CPU's it sold, In Q3 2008 it sold 70.2 for each 100 CPU's.

This makes it even clearer. Intel's ratio of GPU's to CPU's has hardly changed in the past year. However, AMD's ratio has increased dramatically.


Heh, I guess it never occurred to Mr. Scientific that if the denominator decreases significantly, the ratio increases. AMDs CPU marketshare decreased to something like 18% in all of 2008, dunno what the Q3 numbers are but I seem to recall they were lower than in Q3 of 2007, while Intel increased theirs.

Tonus said...

Heh, that's pretty good. Ratio of GPUs to CPUs sold, with a conclusion that is 100% correct. Relevance? Unknown! But the conclusion was 100% correct.

Nonny, I saw an article on SSDs talking about performance and longevity, but I can't remember where it was linked from. :x It was either the Aces forum or HardOCP's news page (now that I think about it, it's very likely to be from a HardOCP link or a Q&A they held with an Intel guy).

Anonymous said...

Heh, I guess it never occurred to Mr. Scientific that if the denominator decreases significantly, the ratio increases. AMDs CPU marketshare decreased to something like 18% in all of 2008, dunno what the Q3 numbers are but I seem to recall they were lower than in Q3 of 2007, while Intel increased theirs.

This is no different then when he cooked, ummm 'analyzed', the CPU market share #'s a while back. Rather than actually looking at the #'s, he decided, the Via/other component had to be subtracted out and then you need to look at the AMD/Intel ratio to see who was growing share.

The assumption of course was the ratio of Intel/AMD CPU's in Via's market was the same as the rest of the market so if Via's share grew or shrank he assumed it was proportional to (and representative of) the rest of the market - a really flawed assumption, and one that is symbolic of a his empirical approach to everything as opposed to understanding fundamentals. At the time he did this analysis Via's share shrank and Intel increased a tiny bit and AMD grew more. His #'s of course lead him to the conclusion that this meant AMD was in general gaining share from Intel (even though Intel's overall share was slightly higher than the previous period)

I proposed that perhaps this was AMD eating Via's market share (as it is low end business where AMD tends to fare better), and perhaps was perhaps not representative of AMD gaining ground against Intel except in the Via market segment. And in an extreme case (though not probable), perhaps AMD ate all of Via's market share losses and it was Intel eating some of AMD's. The following quarters Intel continued to grow and AMD shrank (as Via was so low, there was no more of that share to eat). This theory, analysis of market share trends were suddenly never seen or heard from again.

Looking solely at his #'s, it seemed like a reasonable argument - the problem again was him trying to fit the #'s to his perceived (and preformed) conclusion. If you took a look at the overall environment - like the fact that AMD's growth came at a time when Via dipped quite a bit, and Intel was fairly flat overall, and also knowing where AMD was competitive at the time - it was fairly easy to see this was a flawed use of the #'s.

As 'nonny' stated, his latest example is remarkably similar - he is using a change in CPU ratios to drive a change in market share ratios for graohics.

I would propose that AMD is going to sell discrete cards in a ratio vs Nvidia that is pretty much independent of Intel vs AMD CPU sales. (there is a small interdependency due to SLI and obviously chipset #'s are much more dependent on ratio ofn Intel/AMD CPU's). So if the ratio between Nvidia and AMD graphics stayed fairly constant, but the AMD CPU sale drop; you will suddenly see an artifical GPU/CPU ratio change. The problem is what happened if the reverse situation was happening? If AMD grew to 80% CPU market share, then suddenly this ratio would drop... and then we would conclude what about graphics market share? (absolutely nothing) Would Dementia be talking about a rapid decline in graphics share?

It's a matter of making the #'s fit the conclusion. AMD gained significant graphics share this year, though the Q4 #'s clearly showed a leveling off (of course Dementia excluded these for some 'other' reasons) It's hard to say if this is a trend or not, but clearly his use of #'s is bogus. It will be another quarter or two to definitively show this (like the CPU case I mentioned above), and by that time he will have moved on to another topic where he will cook, sorry I mean 'analyze', the #'s to fit again. And the cycle in the life of an AMD fanboy goes round and round!

Anonymous said...

For a minute there I was considering employing Sci to do my taxes. Then I realized I'd likely be spending 10 to 30 in jail :)...

I see that he never publicly (on AMDZone or his blog) acknowledged his fiasco of a prediction about Asset Smart. His trouble is that he writes in absolutes, tolerates no dissenting POV, and then winds up with a carton of eggs on his face when reality rears up and bytes him on the arse. I suppose that is why we enjoy seeing him trip up so much. If he were a bit more reasonable and willing to listen to people who actually know the subject matter, he might garner some sympathy instead of ridicule.

But, I think he's too set in his ways to change, so his career as the butt of our jokes remains safe.

pointer said...

Anonymous A Nonny Moose said...

I'm considering an SSD boot drive for my next build, but am a bit concerned over the longevity of SSDs compared to hard disks. I know flash memory is only good for a certain number of write cycles, and that the SSD firmware tries to spread the writes evenly across the memory addresses, but still I'd hate to see my SSD lose a significant part of its capacity after a couple of years usage :). You have any idea on how long one lasts under average use?


Heard from a Si architect some time ago that flash devices normally has bigger capacity than what it advertises so that when one or more blocks have issue, the capacity still maintain for some years to come. Not too sure if this is also apply to SSD too (which i think it should)

Anonymous said...

For a minute there I was considering employing Sci to do my taxes. Then I realized I'd likely be spending 10 to 30 in jail :)...

Or get nominated for a position in Obama's new administration (honest tax paying folks need not apply)! Perhaps this is just a new creative method for generating new revenue streams as all of the potential appointees suddenly become "aware" (purely coincidence of course) of tax errors and suddenly pay a bunch of back taxes.

Tonus said...

"If he were a bit more reasonable and willing to listen to people who actually know the subject matter, he might garner some sympathy instead of ridicule."

If I am not mistaken, he has held to his views about DFM (that AMD only recently began to use DFM, and it's another reason that they have closed the gap on Intel in regards to process tech).

InTheKnow said...

His trouble is that he writes in absolutes, tolerates no dissenting POV...

And that is what sets my teeth on edge.

It also looks like he is badly misrepresenting my position viz-a-viz Atom over in the Zone. Ironic considering how he feels others are always doing that to him.

I quote:
Mobile is surprising. AMD's average mobile CPU ASP for 2007 was about $64 and Q3 2008 it was about the same at $62. However, Intel's mobile CPU ASP dropped from a 2007 average of $119 to just $87 which is a 27% drop. This is rather bad news for Intel since its mobile ASPs are dropping faster than its mobile volume increases. Don't believe it? Well, Intel shipped 22% more units in Q4 2008 than Q4 2007 but brought in 13% less revenue. I recall when one Intel fan described Silverthorne as Intel's silver bullet but I don't think he had Intel's revenues in mind as the target.

What is funny about this, is that one of his own zoner's corrected him on the difference between ASP's and margins.

That was what I tried to explain to him a long time ago.

The Atom is dirt cheap to make. If I can make it for 1/4 of the cost and sell it for 1/3 as much (I'm just pulling these numbers out of the air), I have a winner. Despite the apparent "loss" in revenue.

He obviously never did grasp the concept.

I was also amused to note that a 6.25% drop in ASP for AMD was "about the same". Unless their cost structure changed, that is a direct hit to the margins, which didn't look too stellar at their last earnings call.

Anonymous said...

The Atom is dirt cheap to make. If I can make it for 1/4 of the cost and sell it for 1/3 as much (I'm just pulling these numbers out of the air), I have a winner. Despite the apparent "loss" in revenue.

Now this is just crazy talk... how is Scientia supposed to know about earnings when AMD has had none for the last 7 quarters? You have to look at revenue because that's just good business sense... who cares whether you earn money?

The other crazy thing is all of the talk about cannibilization - Intel talked about it during their Q3 earnings call... if there is any cannibilization (and they wouldn't even acknowledge there was back in Q3) it would be on the ultra low end notebook which would actually mean IMPROVED margins (due tot the lower unit cost as ITK points out). So yeah revenue/ASP may go down but you make more from the lower revenue (margins improve). It is rather stunning that many folks, including many people in the media, can't understand this rather simple concept.

Ahhh... the cannibilization talk reminds me of the days of the "Osborne effect" when Conroe was benchmarked ~3-4months prior to release. (This of course was from people arguing that the benchmarks weren't real and were cooked)

Anonymous said...

If I am not mistaken, he has held to his views about DFM

Actually when Intel made a big deal about some of the 65nm and 45nm design rules, end there were a few web articles on the benefit of RDR's (restrictive design rules) - Dementia initially dismissed this as a big bunch ado about nothing - he even wrote a blog about it, if you go through his old blogs (assuming he didn't delete it).

Then when AMD's 65nm process turned out to be 'craptastic' (Sparks - that's another one of those insider technical industry terms), despite the 40% improvement claims AMD had made - he suddenly completely changed his tune and said MOST of Intel's lead was due to DFM/RDR's (of course the process couldn't have anything to do with it).

You see now this concept which he at first thought was magic beans could suddenly be used to explain away the difference in process performance - it was now the magic bullet AMD needed. The other reason he could do this is that AMD suddenly started talking up DFM for 45nm (as if it were some newfangled concept).

You guys have the wrong assessment of Scientia - he doesn't doggedly stick to a belief or a theory, he only conveniently sticks to it if it explains away an AMD shortcoming or gives AMD a theoretical advantage. He just doggedly spins anything to the advantage of AMD (or disadvantage of Intel)

SPARKS said...

Sorry, I have a pet peeve with this one. (As you know) I don't think there is another acronym on the planet that bothers me more than DFM.

I'm a practical mechanic by trade and by nature. (E.g. It looks great on paper, it won't work, it's waste of time. we'll lose money, but build it anyway!) If you're a manufacturer, I keep asking myself, why would anyone/any company design for anything else? It's like me reminding someone to breathe, just in case they forget.

I came up with a few acronyms myself to explore the possibilities.

DFAC- Design For Arab Consumption
DFNYS1B+- (you fill it the blanks. I hate it)
DFAB- Design For Altruistic Beauty
DFMM - Design For Marketing Morons
DFKH- Design for Knuckle Heads
DFPPP- Design For Power Point Presentations
DFCJ- Design For Circle Jerks
DFMCJ- Design For Multiple Circle Jerks
DFIMR-Design For IBM Media Release
DFTS- Design For Tech Sites
DFSG- Design For Shits and Giggles
DFCA- Design For Company Advancement
DFTB- Design For Technical Beauty
DFFB- Design For Fan Boys

And my favorite.....

DFWDHAFTTWSLGTT- Design For We Don't Have A Fucking Thing That Works So Lets Give Them This.

Sorry, rant off

SPARKS

Anonymous said...

Pointer: Heard from a Si architect some time ago that flash devices normally has bigger capacity than what it advertises so that when one or more blocks have issue, the capacity still maintain for some years to come.

Saw that Intel dropped the price on its 80GB SSD to $400 but that's still too expensive - half that and I might consider it. I see 8 and even 16GB flash drives selling for under $25 on Amazon and buy.com so it can't be the memory per se that makes it so expensive. Heck, I just bought a 4GB SD card for my wife's camera. Unfortunately I failed to notice it is SD-HC so it doesn't work. But seeing as how I paid a whopping $2.50 for it, I won't bother returning it. I'm sure I will eventually upgrade my own camera and I think the new ones take the HC version.

Anonymous said...

Anon: This is no different then when he cooked, ummm 'analyzed', the CPU market share #'s a while back. Rather than actually looking at the #'s, he decided, the Via/other component had to be subtracted out and then you need to look at the AMD/Intel ratio to see who was growing share.


I thought it funny how Sci went off about GPU marketshare, shortly after the news stories hit the wires about AMD losing CPU marketshare (down to a total of 18.5% in 2008, from about 21% in 2007). I'd really like to see Q4 results but I guess only Mercury Research has those numbers for now and I ain't paying $2500 to find out :). I did note that according to AMD's own Q4 report, graphics cards lost $10M in Q4, as opposed to (gasp!) making a profit of $50M in Q3. That prolly explains why Sci picked on Q3 numbers for his anal-ysis.

Anonymous said...

Scientia, Sharikou, and all the other AMD zone power/benchmark measurbators have nothing left to measurbate on. Try as they might they don't measure up.

AMD comes up limp when it comes to power, performance, future roadmap, damm even the little detail of measurbating 45nm they are late and on boring SiO2, slow power hungry just plain boring...

Oh I also forgot they are bleeding cash. Their arab sugar daddy is now running red as oil falls below 40 bucks so even hector won't be able fluff a load about of his arab daddy no matter how long he sits on his knees.

SPARKS said...

Far be it me to EVER criticize my beloved INTC.

However, I am fully aware of the strong relationship between HP and INTC over the years (close to 20), especially their commitment (and billions) to the RISC based processor called Itanium. It's new and improved iteration code named Tukwila has been ---um--- scheduled for some modifications, ---again.

I'm sure INTC/HP have built a solid infrastructure around the processor over the years which is valuable asset to both,---- yeah, that's it. I'm sure the boys over at INTC have poured blood sweat and tears into this platform (at gun point?), as not to diminish their efforts in the least.

Yes, Virginia, there is a question coming. Why is this thing so important? What can it do that a new up and coming Xeon W5580 won't do, besides get its ass kicked from here to, say, the Orion Nebula?

http://www.eweek.com/c/a/IT-Infrastructure/Intel-Tweaking-Itanium-Design-Delaying-Chip-Launch/


SPARKS

Anonymous said...

Itanium never turned out what INTEL or HP had dreamed of

HP is happy they got some sucker to take over the designers and throw billions at productizing it. HP would never have been able to turn that design into silicon with their sorry ass process

INTEL had hoped it would have replaced IA32 but after billion and billions it was years late and who would have thought that vernerable old x86 intruction would morph to IA64. INTLE learned the hard way pissing away billiosn and billions that they did to themselves what they did to Alpha, PowerPC, 6800, Transmeta, and others that the x86 conquers all and will never never die.

Soon even the inferior copier will die too, bye bye AMD

Anonymous said...

Can you imagine how lucky AMD is that INTLE never took all those Itanic designers and instead had them thousands designing IA64 and or larabee earlier.

We be done with Nvidia and silly AMD long ago.

Ah, but sorry ass mistakes those Intel executives made, hopefully they are all retiring like good ole Craig and they won't waste billions again

hyc said...

A Nonny Moose said...

I'm considering an SSD boot drive for my next build, but am a bit concerned over the longevity of SSDs compared to hard disks. I know flash memory is only good for a certain number of write cycles, and that the SSD firmware tries to spread the writes evenly across the memory addresses, but still I'd hate to see my SSD lose a significant part of its capacity after a couple of years usage :). You have any idea on how long one lasts under average use?


Well, we can establish an upper bound as the capacity * writes per cell. For my 256GB drive and MLC's typical cycle life of 10,000 writes, that means the drive will endure 2.56petabytes of writes - more than the lifetime of the plastic and metal it's constructed of, I'd expect. That assumes a perfect wear-leveling mechanism, and zero write amplification, of course.

My 120GB drive was only 40% full when I swapped it out for the 256GB drive. So now at 20% full it's got ~200GB worth of spare sectors available. With my usage patterns it will be many years before the flash cells wear out.

hyc said...

A Nonny Moose said...

Pointer: Heard from a Si architect some time ago that flash devices normally has bigger capacity than what it advertises so that when one or more blocks have issue, the capacity still maintain for some years to come.

Saw that Intel dropped the price on its 80GB SSD to $400 but that's still too expensive - half that and I might consider it. I see 8 and even 16GB flash drives selling for under $25 on Amazon and buy.com so it can't be the memory per se that makes it so expensive. Heck, I just bought a 4GB SD card for my wife's camera. Unfortunately I failed to notice it is SD-HC so it doesn't work. But seeing as how I paid a whopping $2.50 for it, I won't bother returning it. I'm sure I will eventually upgrade my own camera and I think the new ones take the HC version.


Visit www.dramexchange.com if you want to get pissed off some time.

The spot price for 32Gbit MLC is only $6.50 today. Just over $1.60/GByte. Most of the cost in all of the SSDs currently on the market is in the controller; the storage capacity of the drive contributes only a small fraction to the cost. The infuriating thing of course is that the controller is a fixed-cost item - it costs the same to put into the 80GB drive as it does in the 160GB drive. The markup on these things is outrageous, especially the higher capacity models.

That's why I went with the G.Skill Titan - 256GB for $499, just under $2/GB, is actually a very fair price for today's market. The sequential read/write rates on my drive are just as advertised, unlike the OCZ which was always 20% slower than spec'd. Random reads are still faster than any HDD; random writes are still the Achilles heel but with 4GB of RAM in my laptop, there's plenty of room for FS cache. So, I pretty much never notice any write bottlenecks...

Anonymous said...

Hyc: Well, we can establish an upper bound as the capacity * writes per cell. For my 256GB drive and MLC's typical cycle life of 10,000 writes, that means the drive will endure 2.56petabytes of writes - more than the lifetime of the plastic and metal it's constructed of, I'd expect.

That's reassuring to know, thanks. I had read some articles when SSDs were first on the market, professing concern over their longevity, but now I think the writers were more concerned over the exorbitant pricing and were using lifetime concerns as a reason not to buy.

As for the prices, think I'll just get a smallish one for my boot drive for my next build, and wait for prices to come down before getting one for data, games etc. I just hate waiting for Vista or XP to boot up :)

Anonymous said...

I just hate waiting for Vista or XP to boot up :)

Is it really the hard-drive? On my computer, it seems the BIOS/MOBO crap takes almost as long as the actual XP boot. (Don't know about vista as I see no reason to upgrade)

Anonymous said...

Is it really the hard-drive? On my computer, it seems the BIOS/MOBO crap takes almost as long as the actual XP boot. (Don't know about vista as I see no reason to upgrade)

It's partly the HD, and partly time I reformatted the boot drive and reinstall XP I guess - after 2 years of installing/overwriting software, drivers, etc etc the registry gets clogged up. I've used the registry cleaners in the past with just so-so results. I have a pair of Raptors striped in Raid 0 so the HD is actually pretty fast, but of course SSD is faster still and less noisy :).

What I really need to do is tear the entire machine down, lap the CPU IHS, do some better cable routing, and clean the dust bunnies out. I have it in an NZXT Nemesis case with a bunch of 120mm fans, but being that my "office" is in the basement which is carpeted, there's enough carpet dust so that I have to wipe the fan screens off. Realtemp shows anywhere from 28C to 23C idle on my QX6700 oc'ed to 3Ghz, but still it's not perfect. I'll likely work on this machine after my i7 build is up and running.

SPARKS said...

I am at a bit of a loss to understand why AMD's today's newly launched AM3 chips have a significant drop in speed. Fuddie is reporting, and has benchmarked, the new Quads and Triple Crpples. He saying the Cripples should clock faster.

What up with this, bugs in the memory controller, perhaps?

http://www.fudzilla.com/index.php?option=com_content&task=view&id=11922&Itemid=1

SPARKS

Anonymous said...

What up with this, bugs in the memory controller, perhaps?

It's hard to say exactly what, but that won't stop me from making some guesses! :)

The combo controller may add a bit to the thermals making it harder to fit the high bins in the same TDP/ACP's. (I don't know how many transistors were added to do the combo DDR2/DDR3 thing).

The other theory (which I consider more likely than the one above) is these high bins may just have poor bin splits and it may just take more time to build up an adequate supply of the AM3 versions. (AM2+ was in production first so AMD had more time to stockpile the upper bins). Naturally the lower bins will have higher volumes so AMD may just be wanting to get these out there to convince the support folks like the MOBO makers to get moving on fixing the AM3 boards.

And then again this simply could be another case of AMD "just giving customers what they demand" and they are not demanding the higher bins! (stifles laughter and reminisces about the good old Barcelona launch days)

Simply put - if AMD COULD release the higher bins right now, why wouldn't they? I'm sure we'll here some fantastic theories... but why wouldn't you release a higher margin part if you could? It's not like there is a 'special process flow' for 2.4/2.6 GHz chips vs 3.0GHz chips.

Anonymous said...

I have a pair of Raptors striped in Raid 0 so the HD is actually pretty fast, but of course SSD is faster still and less noisy.

So let me get this straight... you have STRIPED RAPTORS AND ARE STILL COMPLAINING ABOUT XP BOOT TIMES! Run LINUX!!!!! :)

Seriously, (maybe HYC knows?), will today's SSD's be significantly faster than striped raptors? (I have no idea).

Tonus said...

"And then again this simply could be another case of AMD "just giving customers what they demand" and they are not demanding the higher bins!"

They decided to let Intel deal with the headaches and bad press from having the high end of the performance range to themselves. :)

It is a bit of a letdown, though. But that has been the story of AMD for a few years running, now.

SPARKS said...

"but that won't stop me from making some guesses!"

I'll take 1 of your "guesses" or "speculations" over any of those shills/experts combined, anytime.

"The other theory...." et al., is what I figured (but don't have the technical backround to warrent a guess). Whatever it is, it's real fugly to introduce a slower chip (by half a GIG) to compete with a faster chip, Pheromone 2, on a cheaper, backward compatible platform with cheaper DDR2 memory. Uh oh.

I'm no marketing genus, but I know a bullet in the foot when I see one.

Anand has a this review for your enjoyment.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3512

SPARKS

SPARKS said...

"So let me get this straight... you have STRIPED RAPTORS AND ARE STILL COMPLAINING ABOUT XP BOOT TIMES! Run LINUX!!!!! :)"


I've been running Raptors striped for over a year now. Sustained read/writes are in the 130MB/s range, or better (X48 native SATA). That's not burst, that's an average from the edge, to the spindle. Nothing changes the seek time, typically ~8ms.

Striping two INTC X25's has got to be hardware freak Valhalla. The soul is willing but the flesh is weak. (or as it's translated in Russian, the Vodka is good, but the meat is rotten).

1600 bucks for cutting 6 or seven seconds on "bloat boot" just doesn't cut it, for me anyway.

SPARKS

SPARKS said...

To all,

Go to the Anand review. The stock clocked Q6600 is still a competive solution!

The best processor for the money I have EVER purchased. Stepping (GO), of course.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3512&p=1

SPARKS

SPARKS said...

Moose,

I found a nice link about AMD's market share LOSS. In good ole' black and white, and you won't need a $2500 script.

That's a confirmed 4Q '08, 0.7% loss, by the way. It has been eroding every quarter.

http://www.xbitlabs.com/news/cpu/display/20090204205823_Microprocessor_Market_Plummets_in_Q4_as_AMD_Loses_Further_Chunk_of_the_Market_to_Intel.html



SPARKS

Anonymous said...

Good link... there's a few scary things in there:

According to estimates, AMD’s desktop CPU ASP was $50 per unit, which is a good news considering the fact that the company only began to ship its new-generation Phenom II processors very late in the quarter.

Good news?!?!? An AVERAGE of $50, meaning OVER 1/2 of AMD's desktop processors were sold for <$50. Yup the focus on quads is really paying off... good thing they didn't make dual cores with K10(?). People blame Intel for the price war, yet look at what AMD has done with trying to commoditize tri-core and quad cores - this only further erodes pricing on the volume dual core market.

On the market of mobile processors – thanks to generally more competitive lineup as well as Intel Atom processor – Intel managed to boost its unit market share to rather unprecedented 89.4%. AMD controlled roughly 10% of the mobile CPU market, whereas Via shipped 0.6% - 0.7% of chips.

10% in the market that is growing the fastest and has overtaken desktop. I guess the glass is half full argument is that there is not that much downside left in this market for AMD? Scientia had this one pegged with Puma gaining continued acceptance, eh?

Even though AMD began shipments of its quad-core AMD Opteron processors based on improved code-named Shanghai design early in Q4 and by mid-November its leading partners began to ship their systems with the new chip inside, AMD’s server market share also declined by 2.4% quarter-over-quarter.

So with the new chip, AMD lost market share (but in fairness gained ASP) in what is probably their most competiitive segment. Again, while folks talk about 4P and up solutions, the money and volumes are made in the 1P and 2P space (where FSB is an issue but not as severe as it is in the 4P+ niche). And this is without Nehalem in any significant volume.

The good news is the declining AMD CPU share should make Scientia's GPU market share calculation look all the better!

InTheKnow said...

Yes, Virginia, there is a question coming. Why is this thing so important? What can it do that a new up and coming Xeon W5580 won't do, besides get its ass kicked from here to, say, the Orion Nebula?

Sparks, I've been told that RISC based processors account for 5% of the server market by volume. But that 5% of the volume accounts for 50% of the revenue. If you were Intel, would you walk away from that market?

I think the potential upside is just too big to walk away from.

InTheKnow said...

For a comparison of RISC vs CISC start here. There is a table of comparisons at the top of the page.

Anonymous said...

So let me get this straight... you have STRIPED RAPTORS AND ARE STILL COMPLAINING ABOUT XP BOOT TIMES! Run LINUX!!!!! :)

Heh, I really should. However I just have too much Windows software and too much inertia to switch :).

One reason why I haven't reinstalled XP yet is Trendmicro's AV - like a dummy, I paid for 3 years and only afterwards found out that they limit you to 3 total installs. Since I already used all 3 for my old P4 machine and laptop, I was trying to stretch it out until this summer. However I'm about ready to dump Trendmicro and just go with the freebie AVG. Trendmicro's spyware sniffer routinely flags a lot of websites including my work home page as being full of 'dangerous' spyware :).

Anonymous said...

Sparks, thanks for the link. I note from another article on Xbitlabs http://www.xbitlabs.com/news/cpu/display/20060728114723.html that AMDs highest-ever totals were in the 2nd quarter of 2006, when they had 33% of total server CPU and 22% overall. However mobile still sucked at 13.6%. Despite Sci's continued flogging that dead horse, it still appears lifeless.

Just saw the financial news where AMD doesn't yet have the 50% votes needed to approve the fab spinoff, so the stock is tanking at nearly 10 cents, or 10 percent take your choice :)

Sunnyvale-based AMD said the shareholder meeting, scheduled for Tuesday, will now be held Feb. 18 because only 42 percent of investors have voted so far, when the chip maker needs at least half to vote to go forward.

AMD, the world's second-biggest maker of personal-computer microprocessors, emphasized that nearly all of the votes cast so far — 97 percent — have supported the company's plans.

AMD, which has lost nearly $7 billion over the past two years, wants to sell a bigger piece of itself to Mubadala Development Co., an investment arm of the Abu Dhabi government, as part of a larger deal that would allow AMD to spin off its factories in a moneysaving move.

Shareholders are voting on whether to issue new stock that will be bought by Mubadala, a key part of the complicated arrangement.

SPARKS said...

ITK-

"I think the potential upside is just too big to walk away from."

Sure, so is the great relationship HP and INTC have shared over the years, no doubt.

However, they both have been ringing "the potential upside" bell for years. Ten years is an excruciatingly long development period, to say the least, and not many are sitting down to eat during the last ten.

Additionally, from your link in fact, their conclusion was.....

"Current architectures are a hodge-podge of features that embody a variety of trends and design approaches, some RISC, some CISC, and some neither.  In the post-RISC era, it no longer makes sense to divide the world into RISC and CISC camps.  Whatever "RISC vs. CISC" debate that once went on has long been over, and what must now follow is a more nuanced and far more interesting discussion that takes each platform--hardware and software, ISA and implementation--on its own merits."


This is precisely why I put the question forward. I certainly am no ITK or GURU on the hardware end, and without doubt no hyc on the software end. The conclusion above speaks volumes and reflects my sentiments exactly.

Pass the steak sauce, give me the 4 socket board (Boxboro EX), 4 Nehalem (EX), 32 Gig of DDR3, and lets feast on AMD's carcass, hold the Itanic.

This is something I can understand.

http://download.intel.com/pressroom/
kits/events/idffall_2008
/SSmith_briefing_roadmap.pdf



SPARKS

hyc said...

Seriously, (maybe HYC knows?), will today's SSD's be significantly faster than striped raptors? (I have no idea).

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=290&Itemid=60&limit=1&limitstart=7

For reads, yes, a single SSD can beat a striped pair of raptors. For writes, no. Striped SSDs however.....

SPARKS said...

"Feb. 18 because only 42 percent of investors have voted so far, when the chip maker needs at least half to vote to go forward."


Moose, this is interesting. Most of AMD is institutionally owned. Therefore, THEY are the ones not signing on to this "E-ticket" ride.

I suspect, as insiders, they are watching INTC's lawyers very closely on this deal.

SPARKS

Anonymous said...

What interesting contrasts

AMD delays a shareholder meeting and watches their stock fall

INTEL announces 7 billion capital expansion to bring 3 sites up on 32nm with first product out end of this year. Tick Tock Tick Tock, my prediction by the time we get to the Tock on 32nm AMD is done, finished and fully owned by a bunch of Arabs who have no clue what they just wasted billions on.

The business just take way to much money to develop and manufacture and AMD has neither the money nor technology now to compete. Even with IBM help they don't generate enough money or volume to invest the money and energy to keep up.

The recession, total dry up of capital, technology pull back is the perfect storm when combined with the piss away of the lead they got with Opteron and the distraction of ATI during the Hector RuinitAMD era has sealed their fate.

InTheKnow said...

According to this, Intel will be making their own version of the "Triple Cripples" all too soon.

I wonder how Paul Otellini's "we see a distinct advantage to having all our cores work" comment tastes now?

InTheKnow said...

However, they both have been ringing "the potential upside" bell for years.

I certainly won't argue with that. However, what would you propose they do? Leave 50% of the server market revenue on the table and just walk away while wishing IBM bon appetite?

I think they have to compete in this space, but you will notice that Itanium doesn't seem to be getting the development resources it once had. The last two iterations have both slipped.

Anonymous said...

Moose, this is interesting. Most of AMD is institutionally owned.

This is really strange; AMD is 59.6% institutionally owned and 21.7% insider owned (source - Yahoo finance) and I believed these are mutually exclusive... meaning these 2 account for over 80% of the outstanding shares.

While it is fairly normal to get an extremely low response rate from individual stockholders - I'm stunned that they are only at 42% (and they've said it is over 97% positive so that means something like a ~43% total response rate).

One would think they would instantly get the 21% insider ownership to vote and then it should be fairly easy to get 30% from the institutions (who do respond to these proxies)... while there are a ton of institution holders (~275), AMD merely needs to get the top 10 to vote (as they control more than 30%). Are they not soliciting the large institutions for a response? This would seem to indicate either poor management of the vote by AMD or some reluctance for the institutions to approve the deal.

Anonymous said...

Ahhh... Dementia at it again over at AMDzone refusing to accept ACTUAL DATA that Core i7 uses less total power for the given task then a PhenomII.

His argument? The Intel i7 is using DDR3 vs the Phenom using DDR2 and he thus deems it impossible to conclude anything. This would almost seem right except that he IGNORES the fact tat the i7 system is using THREE DDR3 memory sticks, while the Phenom is using TWO DDR2. So unless the DDR3 is operating at less then 2/3 the voltage of DDR2, the memory is actually working AGAINST Intel in terms of power consumption! If anyone has an account there could you kindly point out his ridiculousness?

You could also point out the ridiculousness of the bozos there that keep quoting the lost circuits calculation of core i7 power draw. Lost circuits is using a WP which specs max current allowable and then essentially converting that to ACTUAL current draw under full load, multiplying that by the various voltages to come up with their own power draw #'s.

The fatal flaw (besides equating max allowable current to ACTUAL current draw)? All three i7 models have the same WP specs! So by Lost Cicuits 'brilliant' analysis, I guess all Core i7 models have the same power consumption under load, regardless of clockspeed! (Yeah that seems like a sound conclusion) I'm so confused.... if the WP says they all have the same max current spec and same voltage spec, then they must all be using the same power? Wait max ALLOWABLE current is not necessarily the actual current usage under full load? But from my engineering background you always want the operating condition to be AT THE MAX spec right? Is my thought that the chip may be, I don't know, operating at lower than the max allowable spec just crazy talk? Is my though that even though all 3 models have similar WP specs they may not have the same power draws more cray talk?

«Oldest ‹Older   1 – 200 of 354   Newer› Newest»