7.17.2008

AMD Loses another $Billion in Q2'08 while Hector gets the boot

Bogged down by biased press and unfair benchmarks, AMD today reported second quarter 2008 revenue from continuing operations of $1.349 billion, a seven percent decrease (-7%)compared to the first quarter of 2008 and a three percent (3%) increase compared to the second quarter of 2007. As part of its previously communicated review of its non-core businesses, AMD decided to divest its Handheld and DTV product businesses, and therefore is classifying them as discontinued operations 1 for financial reporting.

In the second quarter of 2008, AMD reported a net loss of $1.189 billion, or $1.96 per share. For continuing operations, the second quarter loss was $269 million, or $0.44 per share, and the operating loss was $143 million. The results for continuing operations include a net favorable impact of $97 million, or $0.16 per share as described in the table below. Loss from discontinued operations was $920 million, or $1.52 a share, including asset impairment charges of $876 million, or $1.44 a share.

It may appear that AMD has mastered announcing earnings report in the midst of deflated expectations. Surprisingly, nobody seemed to care that AMD lost $270 million in the quarter and wrote off another Billion from its books. That's the AMD we're all familiar with, beating expectations the wrong way round.

But let's focus on the good news! No, Hector leaving can either be good new or bad news depending on which side of the fence you're sitting. His announcement in the midst of the report may have caught a lot of people by surprise, but like he said, it's been planned and talked about before. The real good news for AMD is that their new product line-up are gaining momentum. They never said which direction but still we're assuming it's getting design wins and a bunch of orders. AMD is also confirming together with Intel that the PC market is healthy. In fact it is so healthy that AMD just might lose a lot less in the 3rd quarter and probably break even in the 4th. AMD's promises might not worth much these days but this time the numbers do support it.

As for "SmartAss-et"(c), AMD didn't come out with any announcement but instead gave away a few more clues in the form of an easy riddle. They said, investment in the microprocessor business involves 3 things: process development, factories and chip design. And they said that, implementing smart-asset will take away two of those things. It doesn't take a genius to figure out that the two are investment in process development and factories. If AMD gets the deal then that would leave them to only focus on chip design . It does confirm that AMD is planning to go fabless. We've been discussing this several months ago while at the same time the blogger across the street thought that asset-smart was some kind of process improvement.

204 comments:

1 – 200 of 204   Newer›   Newest»
Anonymous said...

Ruiz just stepped down as CEO. Dirk Diggler, ummm Meyer, replacing him.

Khorgano said...

Stepping down, but not out. He will remain "Executive Chairman" of the board.

http://biz.yahoo.com/ap/080717/amd_ceo.html?.v=2

Roborat, Ph.D said...

Dirk "Mr Focus" Meyer.

Tonus said...

"Bogged down by biased press and unfair benchmarks"

lol

So they will do a phantom shuffle at the top to make it seem as if they are "taking steps?" Or is there some teeth to this change?

Anonymous said...

'Stepping down, but not out. He will remain "Executive Chairman" of the board.'

That just means he continues to get paid and does stargerizenessments for Dirk. Though I'm curious to see the conference call transcripts about Asset Innovation and whether Ruiz still owns this (I figured, I give it a new name as Smart and Light were getting old and with no progress a new name is warranted).

It's not like the AMD board has done anything for the last few years, so Ruiz should fit right in.

Anonymous said...

So they will do a phantom shuffle at the top to make it seem as if they are "taking steps?" Or is there some teeth to this change?

Short term, minimal impact... long term I think this is a good change. I don't see Meyer as being as ego driven as Ruiz and I do think he will be a little more detail/execution focused. I think AMD fell into the 'a good product is all it takes' mantra and lost focus on manufacturing, sales, execution, operating costs and marketing.

The move had to be made - I just didn't think it would happen until after Ruiz missed his Q3 berakeven promise. In this manner, AMD gets a mulligan on the promise and I already have seen some articles saying breakeven in H2'08 and if somehow AMD does breakeven in Q3, Meyer will look like a hero.

Anonymous said...

Scientia said:
I've had people complain that I've been too tough on Anand ...

who are these people? I thought no one really noticed or cared.

Anand has made a lot of mistakes lately that he has had to correct. But, aside from mistakes Anand clearly favors Intel.

this is hilarious coming from a man who's known for his impartiality.

Anonymous said...

Whats new here? AMD loses money, is not competitive?

If you look back in AMD’s history that is what they do best, lose money and bring up the rear and of course look to blame INTEL and sue people as its others fault they are losing money.

Opps, I forgot, something is different. This time they are making a change at the top.
Hector is gone and Dirk is in. We won’t have to be fixated anymore on how Hector screwed it up, but in the end it matters little at the moment whether Dirk or Hector or Bozo the clown is president. AMD’s situation as it stands today is in such seriously bad situation nothing as small as who is at the top will changes things. This is similar to the situation at Kodak or GE for that matter. Who is at the top can’t change overnight the larger macro issues that are causing the woes for the company. True GE isn’t in trouble and losing billions but it mattered not who was there Welch himself couldn’t change the issues that plague GE, nor can the next CEO of Kodak nor AMD for that matter fix the fundamental problems. AMD is like Kodak or Polaroid it will fail, its only a matter of when.

AMD will NEVER again rise to the leadership and prominence it enjoyed a few years past. That was an anomaly more INTELs doing then anything AMD did right. Hector should feel lucky that Jerry plucked him off that sinking Titanic called Motorola Semiconductor and he enjoyed it for a bit. I’m sure he is recalling those fond times where he challenged Paul to benchmark duels during the Prescott days. What he doesn’t realize those were the days were missed opportunity for AMD as that window of time they had a chance to permanently become competitive. Instead Hector, Jerry, and the merry lot at AMD management pissed away years and billions, and what is worst piled on billions in debt adding nothing to their competitive position. They could have made a stronger more competitive AMD but they pissed it all away.

Today AMD has no competitive technology with which to manufacturing competitive CPUs. Today AMD has no manufacturing base with which to flood the market with the required volumes to win market share back to generate the required cash flow to develop the next generation technology. Today AMD has no fucking chance of catching up to INTEL. It matters not whether Dirk, Hector or if they poached the most brilliant business or technology leader on earth into the corner office. There is NO product roadmap that AMD can pull out of its ass that will change this. AMD doesn’t have what it takes, and they don’t have what is required to catch up. Their only hope is if Nehalem has some latent flaw and is released late ( quarters later ) or released on time and they get a Pentium type bug disaster similar to the Barcelona mess at AMD. Right now INTEL’s 45nm looks by all indications on track, yielding well and way ahead of anything that AMD nor IBM can catch till 32nm. Penryn seems to be more then capable to answer anything from AMD. Once Nehalem launce in the 4 way space the last bastion where AMD held anything is gone. Nothing Dirk can possible marshal at AMD can hope to compete with the INTEL machine.

Others can whine and cry about my very boring and nothing new prospective. But I’m all eyes as to what AMD can do to recover and compete again in ANY segment. Please pray tell how AMD with Dirk can fix it?

Go Ass lite?
Sell to Samsung?
Sell to IBM?

Fold up CPU and compete with nVidia and INTEL in graphics?

None will matter,

Tick Tock Tick Tock AMD is finished.

SPARKS said...

“Penryn seems to be more then capable to answer anything from AMD.”

Attila The Tock- Heh, this gospel!

QX9770

Look at the name. Powerful, almost sinister, this is the monster AMD feared, and rightfully so. It became more than AMD ever hoped Barcelona would be, and some. This the one no one talks about, but they all know very well.

Eventually, this one too, will be added to my well lit Curio cabinet collection dating back to I8086. However, this one is very, very special. It will be on top for quite some time, and I just may frame it all by itself.

SPARKS

Anonymous said...

Hey Mr Tick Tock, or should I say Lex?, are you no longer allowed to post on Scientia's board?

I was just going through some of Scientia's old posts and there was quite a few comments with "tick tock" and a very similar 'style' posted under the name LEX.

We don't whine about you being boring, we whine about Mr TickTock / Lex trying to appear as an expert when he (or she) is clearly not!

In response to your question, if you meant it constructively and not rhetorically, Dirk can startto 'fix' things with a few actions.
1) Stop cutting prices and killing margins. Intel still does not yet have the production capacity to feed the entire market (especially as the market seems a bit healthier and growing more than expected); nor would regulators allow Intel to do this. AMD will sell some chips regardless - might as well get as high a price as possible.

Seeing this, Intel will likely slow down their price cuts as well - if they cannot supply more of the market there is no point cutting prices further. They also do not want AMD to completely go away as that would potentially open up regulatory hell (especially from the EU... could you imagine the fines/cash grab they would attempt if they saw AMD disappearing completely?)

2) SETTLE THE LASWUIT! AMD may have maximum leverage right now (from PR perspective) with the FTC jumping in and EU expanding their charges. Use the settlement to leverage a re-negotiation of the x86 license terms to allow outsourcing (if desired) and use cash to pay down some debt and start a crash course on their own Atom part. With their expertise/capability on graphics, if they had a small CPU die solution for this space (not simply a cut down Athlon) - they could make some hay.

3) Either bring GPU manufacturing in house or outsource CPU production. Manufacturing in the semi space is about scale and with AMD's market share, CPU's alone are not enough to efficiently load 300mm fab(s).

4) Continue to subvert / distract / slow down 450mm as much as possible (there are quite a few IC manufacturers and a lot of equipment suppliers who also want this) If 450mm goes mainstream, AMD will no longer have any hope of manufacturing - they do not have enough volume to recover the capital layouts required for 450mm.

5) Do not attempt to speed up the tech node cycle in an attempt to catch up to Intel. This is just poor business (it screws up depreciation models) and is simply not sustainable. They should go to new nodes when the tech is ready and there is a clear cost improvement and should milk the node for as long as they can.

6) De-emphasize quad core and tri-core desktop biz. They make more money selling two small dual cores than one large (and potentially lower yielding) quad. The quad desktop is killing all their pricing and has forced the tri-core band-aid and has shut out any K10 dual core (where would it fit from a pricing perspective?).

From a PR perspective all AMD has to do is thin down the # of SKU's and introduce 'shortages' on some of the remaining ones and shift that production to the fast growing 'low' end (sub$1000) notebook market and more dual cores.

Killing quads (or keeping just one or two high bins at a high price for the 'I only buy AMD' loyalists) would also potentially allow them to release K10 duals and give them an excuse to raise prices a bit (even if it is only marginally better than K8 duals).

If you killed tri-core (this is the same produtciton cost as a quad core) you could potentially slot the smaller dual core K10's in this pricing segment (make up some fictitious excuse like quad yields and bins are extremely high and just THROW AWAY the tri-cores)

7) Focus future designs on efficiency and size... follow the 48x0 model. Rather than going for elegance and the best engineering solution (native quad anyone?), go for the best tradeoff between design, real world performance, and cost. The 0MB L3 cache quad core may be a good step in this direction (assuming of course there is not a huge performance hit). If you are going to be behind on tech node scaling, you need to make it up on design (and specifically die size)

This is a huge mindset change and requires a close coordination between design and process development teams which will be difficult to do if AMD does indeed outsource manufacturing.

8) Swallow your pride. Dirk has a chance to make changes and will be given a pass early on. If SOI is no longer significantly better than bare Si, don't stick with it. If L3 cache isn't the answer drop it. If native quad yields on 65nm are not good enough - scale that production down. If 45nm isn't ready... maybe hold it off and try to marry that intro with high K. I'm not saying any of these things are necessarily true, I'm saying don't be married to the past philosophies and past roadmaps and do whatever is best going forward (in some respects Ruiz was cornered into having to deliver on his past roadmaps and commitments). Dirk should use his 'newness' to alter and adjust roadmaps and plans where it makes sense.

***********

Now none of this would give them significant market share, but I believe it would quickly get them in the black and make them more financially vaiable, help get back some investor confidence and clear some debt off the books. This would then presumably give them an option to keep manufacturing in house (if they want to) and raise money in 2009 if needed.

I'm an Intel fan, but if this site is just going to degenerate into a trolling/bash AMD site then it will wither and die just like Shari-kook and similar to what I think has been happening at Scientia's. I like the discussions here and people challenging others regardless of where the perspective (pro-Intel or pro-AMD) is coming from.

- GURU

SPARKS said...

GURU-

As usual, with Hubble Scope vision, and Swiss clock work precision, your thesis and analysis is absolutely flawless, like so many times in the past. Categorically, you’ve laid out the structured foundation for an eventual and possible AMD recovery.

This is no compliment by any stretch. This is just a simple observation by reviewing your previous posts during the past year, all of which have come to pass with clairvoyant accuracy. I may not have 1/1000 of your expertise, but I do remember everything someone says or does with photographic memory. I can recall it the conceptual analysis, backed by specific references, with equal aplomb. An example that impresses me most to this day is Barcelona’s speeds you with pegged at ~ 2.6 to 2.8 gig.

What is most interesting, setting aside any bias fanboyism, I really don’t think AMD can see, or let alone implement, the guild lines you set forth so succinctly. In fact, if I may be so bold and forgive me if I over step my bounds, I detect some frustration on your part regarding AMD, failures, misstep, and stupidity. Clearly, as I have said so many times before, it as if they were deliberately making the company look as bad as they possibly can, by design.

We have spoken about the board of directors either being corporate yes men, completely incompetent, or comatose, the dynamics of which I will never understand till someone, hopefully, writes a book about their progressive calamitous failures.

Therefore, speaking for myself, please forgive my lighthearted ridicule of AMD. Those AMD billboards on my Park Ave. walk to either the New York Palace, Waldorf Astoria, and Four Season’s are still VERY fresh in my mind as they publicly ridiculed INTC, for what seemed an eternity.

Frankly, and this is a compliment, I don’t think anyone at AMD has your vision, nor the discipline, to institute “a huge mindset change and requires a close coordination between design and process development teams”. I don’t think Derrick Meyer has the stuff to do it, as he was an active participant in the long term failure of Advanced Micro Devices. I’m still very short on the company on the whole.

Respectfully

SPARKS

Anonymous said...

Nice thoughts you get cyber A for effort

But in the real world those things are not enough for AMD.

Without leading edge technology they can't compete in CPU either for power or performance. INTEL will have the upper hand having the huge volumes of scale on 45nm and soon 32nm to push product down the stack with huge die and performance advantage.

Bringing GPU in house, they can't afford. Remember they just sold off 200mm capacity that might have been perfect. AMD would need to retool and or run a non SOI with different design rules to make GPU fly in house. They don't have the capacity or resources to do both. They are caught with a losing position either way.


Kill native quads, that is funny. AMD the ones that heckled INTEL for their cut and past dual core and quad cores. They can't do it.

450mm, they don't need to do anything, nor is it relevant. The tool guys lost a bundle on 300mm no way are they going to do 450mm it is a lot of FUD with no substance.

Pricing, they have no pricing leverage. INTEL holds all the cards and with 45nm efficiencies, the trimming in the company they hold all the cards with pricing. AMD would love nothing better if there was no price erosion. Guess what another round of cuts are coming from INTEL. AMD has to respond as they have inferior product, inferior roadmap and now have to sell for cheaper when they can least afford it.

Settle the lawsuite? You may drop it? you think INTEL will settle? No I doubt it, so AMD again without losing face has no option but to push own. Again a big mistake by Hector.

SKU reduction, quadcore reduction, tri-core elimination. None matters as INTEL again holds all the cards here. All this is FUD again to confuse the buying public. End this and people will see AMD for what it is, a totally no viable company producing non competive CHEAP products that have to be sold at lower price then equavalent INTEL parts. That is BK situation.

Make it up on design? You are saying AMD can out design INTEL on inferior process. You expect INTEL to pull another Tehjas and Prescott blunder?

AMD is already late to simple efficient designs like Atom, way behind.

All nice thoughts, these they should have done 3 years ago when they held the upper hand. Now its too little to late.

Scientia spews nonsense these days just like Sharikou. Not worthy of a post or even a moderated attempt. WTF cares about a review. In the end everyone can make their own decision. And its clear who has superior process, design, and products.

Life in CPU land is so boring. Larabee adventure will be something to watch.

SPARKS said...

Dirk is up against a serious roadmap here.

http://www.hexus.net/content/item.php?item=14468

Of course, this does not reflect prices cuts on ‘older’ products as new ones ramp.

SPARKS

Anonymous said...

In fact, if I may be so bold and forgive me if I over step my bounds, I detect some frustration on your part regarding AMD, failures, misstep, and stupidity.

The frustration lies in the fact that this could be a decent company if run like a business and the people who get screwed are the stockholders while folks like Ruiz get paid millions and then the golden parachutes.

Ruiz should be on the board after all of this? Really? This is the best option? He is the one who should be responsible for Asset light? Hasn't he 'lightened' AMD's assets enough already?

Now Intel may be guilty of anti-competitive practices (it's pretty gray in the area where they deal in), but we'll never know because of how poorly AMD was run during that time. They had a good design, a relatively stable & cheap 200mm manufacturing process and they had no clue what it would take to scale capacity.

If they had not been so worried about cutting a deal with Dell for credibility (when it was abundantly clear they did not have the chip capacity to handle this without screwing their loyal customers) and had focused on keeping ASP's up, generating positive cash flow and enabling more investment in capacity then we may have been able to see if Intel was truly locking them out of markets. As AMD was capacity constrained throughout that time period, we'll never know what impact it would have had EVEN IF Intel was operating illegally... it's not like they could have sold more chips during this period.

So who gets screwed? The customer of course - Europe will be looking for their payola soon and this money will not go to consumers, it will not go to AMD, it will go to lawyers and the EU governments. If/when Europe gets paid, you know the US (especially under an Obama administration) will be looking to get their piece of the pie.

Intel in turn will look to recoup that money in future products and guess what? Without price fixing and caps by the EU and US, the fines will simply get passed on to the consumer. And I'm curious as to when the EU will go after all these companies who willingly accepted these rebates, marketing money, etc... do they have no blame in all of this? Were they FORCED INTO taking rebates and ad money?

The new AMD is Yahoo... you have Yang who still thinks he owns the company and is DRIVING it into the ground out of spite. Rather than cede control, he cuts a deal with HIS COMPETITOR (google) to show a trickling of cash flow and hopefully hold off the proxy battle coming in August to remove him. In the meantime that stock WILL NEVER see the $33-35/share offer from MS. And just like AMD, the board is asleep at the wheel and not asking how Yang will get comparable value and instead accepts these poison pills like a RIDICULOUS multi-billion severance plan which will pay some Yahoo folks up to 2 YEARS of severance if taken over by Yahoo and they decide to leave.

Anonymous said...

Bringing GPU in house, they can't afford. Remember they just sold off 200mm capacity that might have been perfect.

This was 90nm tehcnology? Useful much for GPU's? (which are now 55nm). The problem with 200mm is that many advanced technologies now are only available on 300mm, so even if they could reuse some of the equipment it would be extremely difficult to run an advanced process on it. You should know this Lex!

Kill native quads, that is funny. AMD the ones that heckled INTEL for their cut and past dual core and quad cores. They can't do it.

They can do it, I think you mean they WON'T do it - if you read my post (which admittedly was a short novel), I ended by saying AMD needs to swallow their pride and do what's right. I don't think it's likely they would do this, but they could if a true businessman was running the company. Again, pay closer attention Lex, instead of just looking for an argument.

450mm, they don't need to do anything, nor is it relevant.

It is relevant because AMD is not going to be able to afford 450mm, and as soon as Intel moves there, they get a 30% cost advantage. And this will be in addition to any cost advantages from technology node scaling they might have at the time. What they are doing now is the right thing - try to drive a huge wedge in Sematech to slow things down. Unfortunately, Sematech is no longer a huge player in the grand scheme of technology roadmaps so ultimately it is going to come down to Samsung/Intel/TSMC vs the equipment suppliers.

Pricing, they have no pricing leverage. INTEL holds all the cards and with 45nm efficiencies, the trimming in the company they hold all the cards with pricing.

This is not entirely true and as usual reflects you naivety - If Intel cut their prices by 50% across the board it wouldn't matter that significantly as at this time they still can't supply the entire market - if you recall Otellini himself was saying that they were turning away some of the low end business (I think this was at the Q4 call). Because of this AMD does have some pricing power as they will be left to supply whatever Intel can't. It is not a huge amount of power, but it is enough where they should realize they are not compelled to keep cutting prices to maintain share (or lose share at a low rate)

As always you are so focused on the desktop space, you can't see the forest.

Settle the lawsuite? You may drop it? you think INTEL will settle?

For the right price, Intel would do it in a heartbeat! The lawyer fees alone will run into the 10's of (if not 100+) millions and even if AMD has a small chance of winning, Intel continues to take a PR hit, and there is some finite risk.

What you don't seem to understand is that there is a risk management piece to this and you need to look at expected value. If for example Intel has a 5% chance of losing and they could lose as much as say $8Bil (a loss would mean up to 3X the damages - I'm just throwing 8Bil out as an example), hit has an expected value of $400Mil. While the great and arrogant Lex may decide they can't lose and roll the dice, Intel understands their first responsibility is to the shareholders and if you can get a settlement less than the expected value than it is their duty to strongly consider it (and probably take it).

If they settle amicably it may also take some of the teeth out of some of the governmental suits.

SKU reduction, quadcore reduction, tri-core elimination. None matters as INTEL again holds all the cards here.

Again it seems like you are simply intent on arguing and not thinking about what I posted... this is not to compete against Intel, but to prevent competition to their own products! When you have ~$150 tri and quad cores, how much can you reasonably charge for a dual core?

The problem here is that AMD has so many products and the incremental cost to the consumer is not that much to go to the next product. However the problem is their is a huge production cost difference between a dual core and tri-core, yet AMD only gets a small incremental revenue gain. While many AMD fans considered it smart to sell tri-cores instead of scrapping them, they fail to understand that when you consider the impact this has to their dual core pricing, it may actually better to just toss the tri-cores in the garbage.

So thanks for answering the question I posed - it is clear now when you asked what could AMD do, you had no actual interest in this question and were simply trolling again...

I guess now that Lex can no longer troll Scientia's board, he can focus on trolling here. (Lucky us?)

SPARKS said...

“they had no clue what it would take to scale capacity.”

“If they had not been so worried about cutting a deal with Dell for credibility”



Let’s add the catastrophic blunder of unexpectedly dropping the very successful socket 939.

Consummating the ATI 5.4B purchase as opposed to corroboration and licensing.

Native Quad when they clearly weren’t ready. Neither was INTC, and they said so, if I recall. “It would be difficult for us to do now”

SOI

Wrecktor still gets paid millions, just to hang out with the boys?!?!

Therefore, the $64 question is, ‘Where was DIRK during all of this?’ In the basement playing DOOM3, perhaps?

SPARKS

hyc said...

Let’s add the catastrophic blunder of unexpectedly dropping the very successful socket 939.

Consummating the ATI 5.4B purchase as opposed to corroboration and licensing.

Native Quad when they clearly weren’t ready. Neither was INTC, and they said so, if I recall. “It would be difficult for us to do now”

SOI

Wrecktor still gets paid millions, just to hang out with the boys?!?!


S939, ATI - I totally agree there. Said as much when the news about each of those first broke. DDR2 offered no advantage over DDR at the time they shifted to AM2. The new memory controller was slower. All in all, the AM2 transition made no sense.

SOI - dunno. Just like Z-RAM, seems there was still some unrealized potential to tap, and it's obviously non-trivial to rework their designs for bulk silicon.

Native quad - dunno. For server workloads, it still beat Clovertown by a wide margin, I think it was the right choice. But, they might have been able to deliver on time without botching it if they hadn't blown the $5.4BN on ATI already.

CEO salary - ugh. Just ugh... I wish I got paid even a hundredth of that to make such big mistakes...

Anonymous said...

SOI - dunno. Just like Z-RAM, seems there was still some unrealized potential to tap, and it's obviously non-trivial to rework their designs for bulk silicon.

SOI was not implemented with ZRAM (or some other form of SOI based memory) in mind and these technologies will probably see the light of day in the 32nm node at the earliest (and that may be improbable). SOI is a significant cost adder and it's impact on the transistor level diminishes with scaling, so to say you never know or there may be potential ater being in use for over 3 generations, all the while incurring the added susbtrate cost seems like a poor tradeoff. AMD gained the main benefit on 130 and 90nm and chose to go for the short term benefit and now have wedded themselves to it - clearly on 65nm it is not having as significant an impact and I think the impact will be even smaller on 45nm. Annd while the process and integration is different, the tooling for SOI and bare Si is more or less identical, so it's not like AMD couldn't go back to SOI if there was a need. There seems to be a bit of stubbornness to the SOI roadmap and I don't hear AMD ever acknowledge the cost downside of using SOI.

Unlike most folks here I didn't see the 939-AM2 transition as a bust. At some point AMD had to switch sockets for compatibility with DDR2. I think the only real issue was the expectation that this was going to give performance initially (and quite frankly I'm not sure if this was the press or AMD setting such high expectations). With DDR2 prices coming down and getting more mainstream, AMD may have held out a little longer but I couldn't see them holding out that much longer. Maintaining 2 socket configurations for the same chip in the desktop space seems costly (and I'm not sure if intel is wise to do 2 sockets with Nehalem).

While impacting the enthusiats/build your own folks, you have to remember this is still a very small % of revenue and most CPU's are sold into new systems. If AMD had managed the expectations such that this was a simple memory transition, I don't think much would have been made of it (other than the enthusiast community who always like to upgrade without a new MOBO)

Native quad - just consider speed, yield and the thermals. Could AMD have got 70-80% of the quad performance with an MCM approach which would have been significantly cheaper and probably shorter time to market? This seems like AMD going for the elegant approach and forcing themselves into this approach after the dual core/quad core for dummies marketing and being too proud to do some more investigation on whether MCM was a reasonable alternative approach

HYC - When you compare AMD on server loads you are complicating the comparison with the fact that Intel has FSB and no IMC. Had AMD gone the MCM route they would not have had those exact limitations. The native vs MCM variable is not alone responsible for the performance differences you allude to and I don't see how you can conclude native was the right choice without knowing how an MCM version (with HT and IMC) of an AMD chip might perform.

But, they might have been able to deliver on time without botching it if they hadn't blown the $5.4BN on ATI already.

Why? This seems like a poor excuse for the CPU execution issues - was the ATI acquisition distracting the CPU design and manufacturing teams? As all of the ATI production is done in foundries I'm not sure how it would significantly impact K10 release. Could you go into more detail why you think this is the case? I don't think the issues AMD faced were ones that were a matter of throwing more money at them.

pointer said...


Blogger SPARKS said...

GURU-

As usual, with Hubble Scope vision, and Swiss clock work precision, your thesis and analysis is absolutely flawless, like so many times in the past.



are you serious? He might be good, but not to that extent of worshiping :)

Guru said
I guess now that Lex can no longer troll Scientia's board, he can focus on trolling here. (Lucky us?)


I do not know if he is lex or not and dun care. I don't think your points are all right, and I don't his points are all wrong. as long as he is not saying AMD is (and NVIDIA are) going to BK in every posts (ok to do that in Sharikou's blog :)) I do not see a much problem though. Normally I do not reply to your post as most of the time you talk about the process which i am not familiar with (little paper knowledge, 0 working exp) .. but when talk about the business sense, everyone can make some judgment on that and it won't be totally wrong.


Hyc said

Native quad - dunno. For server workloads, it still beat Clovertown by a wide margin,...


too generalized and too vague. All server workload are memory bound? Nope. AMD beats Intel in DP server? Nope. If you said about MP server usages and may be yes but still, not all workload, and on some usage, Intel's is better.

Anonymous said...

.. but when talk about the business sense, everyone can make some judgment on that and it won't be totally wrong.

Did I say that he was totally wrong? Lex made his typical trolling post which asked what can AMD do - I made the mistake of taking this at face value and as a constructive question and took a shot at some possible actions.

He of course proceeded with his typical, too little to late, tock tick I'm just a ----, and really had no interest in actually thinking about or truly analyzing any of the suggestions.

His interest was to shoot the suggestions down and shoot his mouth off some more, rather than having a constructive conversation (which I thought was part of the objective of the comments in the blog?)

I do not pretend to say all of my actions would work, nor am I saying they are 100% correct or Lex was 100% wrong. What I am saying is Lex has no interest in having any conversation which doesn't fit his myopic view of the world and will not consider that an opposing point of view may have some validity.

I realize now that I'm at fault for thinking that if I continue to attempt to reason he may actually consider the ideas at face value and evaluate them honestly, even though there is a mountain of evidence on this board (and at Scientia's board) to the contrary. I guess that's a bit of arrogance on my part for thinking I might be able to do something others haven't and my apologies to the other folks on the board for wasting their time on this.

And Pointer... don't be picking on my man Sparks, as always his comments on me are 100% spot on! :)

Anonymous said...

Not everyone is buying AMD's plan to recovery.

Goldman’s Jim Covello repeated his Sell rating and $3.50 price target on the stock. He is skeptical that AMD can meet its goal of hitting operating break-even in the second half. He notes that the company burned over $600 million in cash in the first half alone, and contends it will have to raise equity capital in the next few quarters, “which would be extremely dilutive.” He says the “cryptic” asset-light strategy the company continues to promise would actually be a negative for the business, “given the difficulties of optimizing design and manufacturing while working with several foundry partners.” And he notes that “AMD’s market share continues to deteriorate and it is having a difficult time closing the competitive gap” with Intel.

J.P. Morgan’s Christopher Danely sounds a similar theme, asserting that “it will be difficult for AMD to make money unless it can gain back some market share and improving its pricing.” He stays Neutral on the stock.

Bank of America’s Sumit Dhanda is a skeptic as well. “The path towards profitability is challenged, now that AMD is fighting multiple battles at the same time, including an aggressive competitive environment in graphics, lack of technology/cost leadership in client/server CPUs against Intel’s 45 nm products [and] share loss in notebooks,” and he says that “the potential encroachment of Atom into AMD’s turf [in] low end desktops and notebooks is likely to make matters even worse for AMD.” Adds Dhanda: “Meyer has his work cut out for him.”

Citigroup’s Glen Yeung is another skeptic on the company’s break-even expectations; in his note this morning he pointed out that he isn’t modeling AMD reaching operating profitability in Q3, or Q4, or even Q1 of next year. He maintains a Hold rating, and cut his target to $6, from $7.50.

Oppenheimer’s John Lau cut his target on the stock to $6 from $7. “AMD’s delayed product launches, vulnerable competitive position and low operating cash compel us to stay cautious,” he writes. “The difficult transition to 45-nanometer geometry and large net debt position compel us to maintains our Hold rating.”

hyc said...

There seems to be a bit of stubbornness to the SOI roadmap and I don't hear AMD ever acknowledge the cost downside of using SOI.

I recall seeing an AMD slide showing the relative cost, but I can't find it now. Anyway, this may do:

http://www.spectrum.ieee.org/jan07/4839

A 200-millimeter SOI wafer sells for about $275, while a plain silicon wafer of the same size goes for $65. Mark-Eric Jones, Innovative Silicon’s president and CEO, points out that as a percentage of total cost, the price gap collapses considerably once you factor in the $2400 in processing needed to turn a blank wafer of either substrate into a repeating mosaic of microchips. Still, all things being equal, the SOI chip will cost almost 9 ­percent more than the bulk-silicon chip.


S939 - AM2: that move seemed premature, if they had waited until DDR2-800 was more common, so there was an actual performance delta, it would have made more sense. I think they had the freedom to start this transition at whatever time they wanted; clearly the rest of the core was largely unaffected by the change.

MCM quad - if you're suggesting they could have used HyperTransport to glue the pair of dual-cores together, perhaps. But that design would limit their quadcore to dual-socket systems, because there aren't enough HT address bits for more than that. I guess they could have introduced HT3.0 sooner, but it would only have allowed them to go to quad-socket. There would be no performance advantage of a dual-socket MCM quad system vs a quad-socket dual-core system; in fact it would probably perform worse because there's less memory bandwidth for the dual-socket system.

I can't say anything conclusive about the CPU execution issues. When the news about the Barcelona TLB Bug first hit, a lot of folks commented that AMD must be cutting corners on their validation efforts. It had already been significantly delayed, and pretty much all their hopes were pinned on it being a success. I'd say whoever was working on it was probably under huge amounts of stress. If it weren't such a do-or-die project, if a few billion extra $$ were still sitting in reserve, people probably could have completed the project with a bit less stress and probably fewer mistakes. Just guessing.

re: server workload characteristics - IMO, the whole reason you're buying a quad-core server chip is because you have a lot of high-concurrency code to run. Otherwise, you can get faster clocked dual-core chips. In workloads that demand multiple threads, AMD won, whether or not it's I/O bound, memory bound, or compute bound. The only place where Clovertown still won was for compute-bound jobs where the entire working set of all 4 threads fit into cache without any evictions. As soon as you got threads competing for the shared cache, performance tanked and you would have done better sticking to dual core.

SPARKS said...

“are you serious? He might be good, but not to that extent of worshiping :)”

Yes and no. Allow me to say, in regards to my QX9770/X48/DDR3-1800/Raptors; this is indeed worship, no doubt.

With GURU, call it technical awe, down to the atomic level. In my industry, I’m considered one of the best. That said, there are many in my field who sound technically correct, but from a hands on practical level there very subtle differences that distinguish the talkers from the real players. Theory and practical knowledge differentiate the experts from the knowledgeable. It may be a small thing, but it is often true. The devil is in the details. Very minor contradictions and technical inconsistencies are the first clues. In reality, it ultimately becomes a make or break scenario between success and failure in regards to successful practical implementation.

As for me, if someone tries to bullshit me, or god forbid challenges my knowledge and expertise in the same vein, we’re talking lunch meat, thinly sliced. And as we’ve seen, GURU doesn’t take well to baloney.

Therefore, not worship, but more, professional admiration and respect. I’ll take that over worship any day. Besides, that’s reserved for QX9770, currently the finest chip on the planet.

“too generalized and too vague. All server workload are memory bound? Nope. AMD beats Intel in DP server? Nope. If you said about MP server usages and may be yes but still, not all workload, and on some usage, Intel's is better.”

Something tells me you know what I’m talking about.

SPARKS

Anonymous said...

If it weren't such a do-or-die project, if a few billion extra $$ were still sitting in reserve, people probably could have completed the project with a bit less stress and probably fewer mistakes

Not buying this, too many questionable assumptions. First off AMD didn't spend billions on ATI, they spent a little and borrowed a lot - unless AMD was going to borrow money to throw at K10 (which is unlikely in my view, beacuse it's not like they were getting great terms on the money) this is a flawed assumption.

The stress was there regardless of the ATI acquisition, I'm not sure how not acquiring ATI or throwing money at it at the working level would really have altered the stress state.

I also disagree with the stress being a probably cause of the execution issues (difference of opinion). I can see how you could potentially argue that it could have lead to missing the TLB bug (though I disagree), but it's not like the TLB bug was the major execution issue. In my view the process/design interactions leading to slower than expected clocks and higher than expected TDP's was the real issue with Barcelona and the TLB is really just a convenient excuse. And stress is not responsible for low clocks and high TDP's.

In my view (and I stated this at the time of TLB was announced), it was a great marketing move and was a convenient way of focusing K10's issues on a singular problem. This bought AMD some time, and a good excuse not to ship product in volume for a while, while they worked on new steppings (and yield/binsplits?) to address more than just the TLB bug. As usual the press did not dig any deeper and blindly assumed the major issue was this TLB bug, when clearly it was not.

InTheKnow said...

I guess that's a bit of arrogance on my part for thinking I might be able to do something others haven't and my apologies to the other folks on the board for wasting their time on this.

Nothing to apologize for. I've wondered myself what AMD could do to bail themselves out of the hole they are in. Your suggestions are worth consideration, even if the person asking the original question didn't think so.

I would add developing a Silverthorne competitor to the list of things AMD should do.

I saw a report today that indicated Meyer will steer AMD away from the MID space (though he will go for the "netbook" space). I think this is a huge mistake based on a short sited view.

The history of computing has been to move more computing power to ever smaller platforms. We've gone from computers that filled rooms and had less computing power that my pocket calculator to servers, to desktops to laptops. Now we begin to move to "netbooks".

MIDs may be a bit early, but they seems to be the next logical step in the progression. And I believe if x86 doesn't continue to move down into smaller devices, its days are numbered.

So from my viewpoint, AMD is condemning themselves to be behind at best or setting themselves on to the road to extinction at worst.

As an aside, if you want a sneak peak at what Menlow is going to bring to the table just look here. Note the unit is equiped with a 10wHr battery.

Anonymous said...

The Inq has forumlated their pet name for the new AMD CEO... they are calling him...

Captain Dirk

I thought it was funny anyway.

http://www.theinquirer.net/gb/inquirer/news/2008/07/18/captain-dirk-speaks

A Nonny Moose said...

Looks like my first prediction may be in jeopardy now - AMD's stock price closed down over 12% to $4.65 on Friday, while oil dropped to under $130 a barrel :) DAAMIT! :P

On 7/16/2007, AMD closed at $15.50. Just in case there's any lingering doubt, I think that speaks volumes on how K10 has busted AMD's wheels off.

However my 2nd prediction about Scientia continuing to spew nonsense is safe. From AMDZone:

"AMD's Financial Outlook/Plucking Numbers Out Of Thin Air"

..."However, they do not mention that with the release of the 4800 series the graphic assets have probably increased in value by about $500 Million. It could be higher later in the year if nVidia doesn't respond well but that is a good estimate for now. They also don't mention the value increase as a result of 790FX or Puma."

Gee, Sci - where did you get that "$500 Million"?? Maybe out of your arse if not thin air? And from what I've read recently, Puma may have missed its small window of opportunity before Centrino 2, thanks to AMD's incredible ineptness and piss-poor marketing.

But anyway my heartfelt thanks to Sci - he's much more predictable than AMD :P.
Along with his cohorts at the Zone, he can be depended on to shoot his pea gun at anybody bearing the slightest bit of bad news about his favorite company.

A Nonny Moose said...

Another point the "financial expert" Scientia tries to make:

From a Zoner: "It's interesting to think what would've happened if they had waited until after the 2900 flop to make an offer."

And Sci responds "Then no 4800 series; the earliest they could have caught up was early next year on 40nm. No Puma until 2009 which is now when the second version of Puma is planned. And no Fusion until 2010."

Hmm, Sci thinks the value of AMD should be increased $500M due to the success of the 48XX series (unlikely since it was released so late in the 2nd quarter), and this somehow justifies AMD spending about $2.5B more that what it would have if it had waited a bit. With that level of financial acumen & savvy, I'd like to get him into a Friday night poker game. Oh wait a minute - he's too broke to be able to buy any AMD stock, even at $4.65 a share. Nevermind :).

NV seems to have been caught resting on its (can of whoop)ass much like Intel back in the Netburst days. I'm sure they'd still be waiting around for ATI to release something competitive, so it's not like time was of the essence for AMD to acquire ATI. And without any new source of $$, how is AMD going to fund the R&D for anything more than another BS roadmap?

Ed over at overclockers.com has a pretty good analysis of where AMD went wrong and what they should have done instead.

I get the distinct impression that Sci is some middle-aged, underpaid programmer, no GF, who has a small (and shrinking) world of AMDZone and his own blog to try and exercise authority & garner respect. Sad, really.

SPARKS said...

A Nonny Moose-

I thought some of your predictions were bit over the top; however, you did get there with a couple. The Oil/AMD connection was a bit of a stretch. But, hey, what ever works for ya. It did that time.

More significantly, is AMD’s price at 4.65. Fugley. This time around the anal-ist weren’t taken in by AMD’s Voodoo accounting methods as they were during the past few quarters. 880 million is a pretty big nut to hide under a shell. As you, and others here have said before (me included) regarding AMD’s usual up tick after the smoke and mirror quarterly parade, it ain’t gonna happen. Not this time.

As of late, AMD is no longer the “darlings of Wall Street”, or the tech press. Obviously, to us AMD has had nothing substantial; I think the mainstream is starting to get the picture, too. Can you imagine the idiots that believed ALL the 2007 spew? No doubt there are a lot of people out there with egg on their faces. This is the prediction I made. It’s called backlash.

“Sure, no problem, the checks in the mail, and I won’t ---- -- ---- -----.”

Therefore, AMD has shot its last load with the retreat of Wrector Ruinz, and Dirk is no Henri Richard. They have one last vestige of hope; it has nothing to do with processor road maps. It has to do with lawsuits, instead.

Allow me to make a prediction. Uncle Sam will do nothing, aside from a token slap on the wrist, to hurt INTC, not now, not with this economy. MS and INTC are multi billion dollar breadwinners for the USA. Never hurt your most successful children. The new dictatorial world business police, The EU, however will get a substantial amount. They know they can pull this stuff on the U.S. They wouldn’t dare with OPEC. Christ the French would be sending their women over to the Middle East just to appease them.

Voulez-vous coucher avec Moi?

Wait! Dirk and company already are!

SPARKS

Anonymous said...

SEC 8K http://www.sec.gov/Archives/edgar/data/2488/000119312508153147/d8k.htm

Hectors employment agreement:

http://www.sec.gov/Archives/edgar/data/2488/000119312508153147/dex101.htm

Nothing has changed except titles, and a few lines on the org chart.

Hector gets his same compsenation, the SEC only changes the paragraphs about his title and then states replace everything instance of CEO with Executive Chairman. He hasn't left or stepped anywhere but perhaps a promotion with less responsibilities.

Dirk, however, got a substantial raise...
http://www.sec.gov/Archives/edgar/data/2488/000119312508153147/dex102.htm

What appears to be a changing of the guard looks more like simply moving the two who were running the company closer to 'two in a box'....

Only in the AMD fashion can they do one thing and fool others that they are doing the other.... amazing.

Anonymous said...

Earnings call transcript (for the financial geeks out there) - these are often more significant than the actual #'s

http://seekingalpha.com/article/85590-advanced-micro-devices-inc-q2-2008-earnings-call-transcript?source=yahoo&page=7

Some of the interesting parts:
Dirk Meyer

We still believe on the strength of the momentum we’re seeing in the marketplace based on the new products we introduced in Q2 that we can expect to remain profitable at the operating level in the second half.


REMAIN profitable? Ummm... someone tell the new CEO that you are not currently profitable - I'll assume this is a misstatement and he meant to say the PROJECTION of a return to operating profitability remains.

Ruiz on Asset Smart:
I can tell you we are very pleased with the progress. We have made enormous progress. We are looking forward to being able to share that with you. Certainly have incredibly high level of expectation we will be able to do that before the end of the year.

They are pleased! Fan-freakin-tastic! That is so meaningful! And then this bluff is called:

In terms of progress could you give any color on how you measure that or what you consider to be the elements of that progress?

Hector de J. Ruiz

I really can’t.



Here's a classic from the CFO:

Clearly with the introduction I’ll call it of the new Quad-Core architecture which really took place at the beginning of this year and servers at the end of the first quarter that is what we’re talking about as I call it the new microprocessor architecture.

So fixing a TLB constitutes 'a new microprocessor'. I know he is trying to differentiate it from the aborted Q3'07 launch, but I think this is taking some liberties!



And if there were any questions on Asset Smart, you need to read between the lines:

Clearly there is going to be cost increases in the gross margin as you pay for the cost of that technology on a per wafer basis and the cost of capital on a per wafer basis. (in reference to asset smart impact on gross margin)

45nm Shagnhai
Any projections on when it will launch?

Dirk Meyer

As I said we are in production now and we will ship for revenue early in Q4.


So any projections on when it will launch? Notice the artful dodge here...shipping for revenue means shippint to OEM's (not necessarily available for purchase). They've done this on virtually every recent product launch and the press never follow up!

A little bit of an itchy question but in terms of the Barcelona speeds are you guys shipping 2.4 ghz now and when will you begin shipping 2.5 and 2.6?

Dirk Meyer

The answer to the first question is no, not yet in volume. On the second question I’d like to hold off until we talk about future products and actually launch them rather than pre-launch announce.


Not shipping 2.4GHz Barcelona's? Nearly a year after release without ANY speed increase? That CTI philosophy sure is paying huge dividends!

Dirk can you give us a sense for when you will begin shipping in volume?...The 2.4 Barcelona.

Dirk Meyer

I don’t want to pre-announce product introductions on the call.


So a 100MHz speed bump is now a new product introduction? This comment might be innocent (don't want to set a precedent at earnings calls), but to me it potentially stinks of something else!

More Asset Smart:
Do you guys expect a major improvement to margins with this or something a little bit more minor?

Dirk Meyer

It will be a major reformation of the company from that standpoint. I don’t want to get into those kinds of specifics but it is definitely going to improve the balance sheet.


And this my friends is the dirty little secret... asset smart is actually probably going to take gross margins DOWN. he talks of improving the balance sheet as it will cut down capital investments, and may even have some novel debt re-structuring. But if you are paying someone per wafer it is not quire as good as doing it in house as the person making the wafer generally also likes to make a profit! Notice the dodge to the gross margin specifics.

Anonymous said...

It is the "perfect storm".. Where did I hear that one before?

Anonymous said...

Wasn't the flop called Barcelona done with the blessing of Dirk? What makes the next flop better?

Anonymous said...

It looks like AMD did believe Dr. Fakerikou about capturing the 40% of the CPU industry.. Here is an advice for Dirk if he wants to really rescue AMD:
1. Drop the Intel lawsuit and stop playing childish games. Save the lawyer money and Make a good processor and I will buy it.

2. Get Hector out of the company.. The opteron was not his idea.. HIs ideas screwed the company. Just look at data

3. Here is a shocker: if you can not beat them, WELL JOIN THEM. work with Intel to make chipsets and graphic cards for their platforms. Make money and deny nVidia that opportunity

4. Focus on the graphic part of the compnay and find an ATOM Killer before it kills you

5. Do not listen to idiots like Rahul and Sharikou.. When they give advices, One is looking for his best interest and the other is just a plain idiot!

Anonymous said...

AMD long on fancy double speak works about profits, desire, goals, and dreams

One small problem for little Dirk is that he needs to get from here which is billions of debt, no credibility, no technology, and really is totally short of any credible details on the path that will lead AMD there.

Talks is cheap, they reason they don't share details announce any more roadmaps and such is that they have none, any bold ones they have don't have any credibility

SPARKS said...

Gentlemen, if you haven’t read this article, may I suggest you do.

http://www.statesman.com/business/content/business/stories/other/07/19/0719dirk.html


In view of Mr. Meyer’s efforts and his promise to become profitable, I think a reality check is in order here in mid JULY 2008.

http://www.sharkyextreme.com/hardware/cpu/article.php/3261_3756681__11

And of course----

http://www.overclockersclub.com/reviews/intel_qx9770/16.htm


Ah, NO ONE, ANYWHERE else is making the comparison, toe to toe, head to head.


They only do this-------------

http://www.legitreviews.com/article/735/13/

The bottom line is this----

http://finance.yahoo.com/q/is?s=AMD&annual


SPARKS

Tonus said...

anon: "What appears to be a changing of the guard looks more like simply moving the two who were running the company closer to 'two in a box'...."

The impression I get is to place Meyer in the firing line for a while to satisfy the people asking for Ruiz' head. If things continue to go badly (or get worse) Meyer takes the beating, but will get a bit of slack since he inherited someone else's (Ruiz) mess. If things get better, Meyer will pull back the curtain and tell everyone that hey... it was really Ruiz who led them out of trouble and therefore should get the CEO spot back!

AMD doesn't seem to be getting any slack from the analyst community this time around, and offering them Hector's head in a symbolic manner might be their attempt to appease Wall Street while not making any real changes.

SPARKS said...

“AMD doesn't seem to be getting any slack from the analyst community this time around, and offering them Hector's head in a symbolic manner might be their attempt to appease Wall Street while not making any real changes.”


You know, ‘T’, I’ve been THINKING! (Imagine that!)

Timing, plain old simple timing!

Right at this moment, now, INTC is absolutely at the right place at exactly the right time, with the right tools for the job.

1) They’re smack-dab in the middle of the 65 to 45nM transition.
2) Nehalem is primed and ready for the 2008, 2009 server assault.
3) Yorkies are absolutely flawless, especially the big fella’s.
4) $1,500,000,000 profits per quarter.
5) Cheap 65nM for sale when and if the economy turns around as 45nM ramps to full production by December. (read Christmas)
6) Good margins and, from all indications, excellent yields at 45nM.
7) Everyone is clamoring over ATOM and they can’t crank out enough of the ugly little bastards. (forgive me, but you guys know I love the heavy hardware)
8) Q6600, as I predicted last year, is still selling like a give away item. The things are ridiculously popular, everywhere. (Buy a (GO), get 600 MHz free!)
9) INTC HAS EMBRACED the enthusiast segment with open arms practically ENCOURAGING OVERCLOCKING!!!!
10) Partners, vendors, OEM’s, are getting first rate attention and support.

The American Way, INTC got off the balls of their asses and made it happen, big time!

Let the race begin.
Only the strongest survive.
Slow and steady wins the race.
Let the best man win.
Performance is king!
Second place is the first loser.
If you find a better product, buy it.
Supply, demand, profit, REWARD.

There is one fly in the ointment however, to take all this excitement away. It’s an ugly little shit determined to bring the whole thing down because of entitlement. That’s right, entitlement.

Given the current economic woes, gloom and doom, ad nausea, a thought occurred!

Banks have subsidized losers who couldn’t afford to play/pay the game! Homeowners had their mortgages and AMD had ATI. Perfect! One group had a house they paid $550,000 they couldn’t afford, now worth $220,000, AMD has a company, admittedly with the same ratio, but 10000 times those numbers!!!

Wait, let’s stop everything! Let’s reward AMD’s failures and punish INTC’s successes. Let’s give AMD a free pass, because they are ENTITLED to manufacture inferior products and coexist with INTC. They deserve market share irregardless of how and what they manufacture. Perhaps AMD should be subsidized by INTC or UNCLE SAM because we like AMD! Therefore, if a manufacturer in this country/world is too successful they need to be punished.

Monopolistic Behavior, say it a dozen times like some Orange Clad, bald, 1980’s, chanting, wild-eyed Mooney, and you too can join the club

Let’s give the defaulting homeowners a break, since they have a house, let’s have them keep it at the expense of the more successful folks, taxpayers. After all, they’re entitled! Where does that damned SPARKS come off buying 1500 CPU’s? The NERVE! Let kick him up a tax notch, let’s see if we can tax him a little more.

SORRY, I DON”T BUY IT, AND I’M TIRED OF IT.

Sorry fella’s I don’t eat quiche.
I like steak, rare.
I’m a man, I have the hardware to prove it, and I am not cultivating my sensitive/feminine side.
I’ve won some, and lost quite a few. No one has been there to help me off the balls of my ass. I wouldn’t have taken it anyway.
I’m not some poor Mexican immigrant success story who made it big, only to beg for hand out in the end for sheer incompetence.
I’m of Italian decent, just like Big Paulie, we don’t take charity, and we don’t accept excuses for personal failures.


Crush them, Big Paulie, I’ll take a 10 dollar a share hit, any day, to put an end to the “Scrappy Little Maggots”, once and for all. AND, let the chips fall where they may.

SIMPLY CRUSH THEM!

Quite right Tonus, this is business; no one gets a pass, not with these monumental blunders.

OK, rant off------------Goddamn it.

SPARKS

Anonymous said...

Are you implying that Ruiz being mexican has to do anything with his failure? I can name several italians that were failures as well.

Anonymous said...

Only the strongest survive = INTEL

Slow and steady wins the race? You mean fast and steady. Slow and taking on debt is AMD's way

Wait, let’s stop everything! Let’s reward AMD’s failures and punish INTC’s successes = The EU way!

SPARKS said...
This comment has been removed by the author.
SPARKS said...
This comment has been removed by the author.
Anonymous said...

Let’s give the defaulting homeowners a break, since they have a house, let’s have them keep it at the expense of the more successful folks, taxpayers.

Well when the Senate chariman of the banking committee was a 'friend of Mozilo' it's hard to not want to bail people out and help the housing industry. Note - this is the name of a real program where people Mozilo (chairman of Country Wide Financial) knew/liked got a favorable interest rate on their mortgage.

Dodd of course claims ignorance - this seems quite plausible - it's not like the chairman of that committee he would know much mortgage rates!

This of course is also the same Congress that encouraged no money down loans as apparently everyone is entitled to a house even if they can't afford it which partially caused the bubble in house prices.

Big Oil is next... you know the companies who have a lower profit margin then some entertainment companies, but are clearly 'gouging' customers for record profit - despite the fact that no one asks how is this possible if they earn less profit on every dollar taken in then say Dreamworks entertainment. If we tax their profits they'll clearly lower prices and continue to want to grow right? Isn't that how economic works in bizarro world? Higher taxes = lower prices, more growth?!?

And to the point... after all this, the people will need a new enemy and AMD's ineptness will put Intel in the crosshairs. Big, successful, growing, main competitor failing... I see some deep pockets!

Of course with Obama having an upcoming campaign rally IN GERMANY (trying to secure the Americans overseas in Germany vote?), I'm sure Intel will be fine.

SPARKS said...

Someone posted as me with the last two posts, they were deleted.

The big one was mine.

SPARKS

SPARKS said...

“Are you implying that Ruiz being mexican has to do anything with his failure? I can name several italians that were failures as well.”

Not at all, what object to is this American success story coming from nation whose citizens weren’t blessed with Wrectors opportunities, rise to the CEO of a multinational corporation, fail, and then panders to foreign interests like a poor beggar with hat in hand, while making tens of millions despite 4 billion in losses.

When he fails the American success story, his slash a burn tactics and failures have not only run one AMERICAN company into the ground, but it has put another at jeopardy, worldwide.

On the contrary, I’m quite certain; his naturalized status and charming success story has given him an edge gaining sympathy and popularity against the big evil American Monopolistic Empire called INTEL Corporation overseas.

No other CEO in history has done with this man has done, ever.

He sold out AMD to foreign interests, and now he selling out INTC. I see it as selling out America and its workers, worldwide, twice. It has nothing to do with his nationality, it’s his poor me, beggar tactics, I object to. He’s playing this one low and dirty, right from the street.

In fact, it sounds as if he wants equal opportunity status, doesn’t it? Italian Americans NEVER got equal opportunity status, got it?

We’d shovel shit first, pal, and we have.

SPARKS

SPARKS said...

Sorry about those two posts gentlemen, my wife’s idea of having fun. I think she’s jealous of my “worship” of you all. :)

At times, I envy the fact you work with women in the field who love this stuff as much as I do. Indeed, a rare breed.

SPARKS

Tonus said...

You should be glad that you've got a wife that knows how to use a computer at all. :)

hyc said...

I think she's proved that she knows enough to be dangerous... ;)

SPARKS said...

HYC- We don’t agree on much, but this time I concur 100%.
Tonus- Yeah, she’s a trickster.

SPARKS

jumpingjack said...

MR report for Q2 came out yesterday.

AMD lost marketshare to Intel in Q2, this is not surprising considering the results.

What is sad though, they also lost share to Via :)

Anonymous said...

What is sad though, they also lost share to Via :)

This brings back fond memories of Scientia's ridiculous argument a while back when he tried to claim lost Via share meant AMD was gaining even more then indicated by the marketshare #'s and tried to do some voodoo economics and math by simply ignoring the via share.

If we were now to use that theory today it would mean AMD lost even more marketshare then the #'s indicate (this is of course an absurd argument as there is no way of knowing who Via is gaining or losing MS to) - but it would be interesting to see if Scientia would use consistent arguments!

Of course we will never know as the earnings and marketshare #'s are only blog-worthy when they are trending in AMD's favor... which is obviously consistent with the unbiased analysis over there :)

Anonymous said...

From amdzone, everyone's favorite blogger:

"5.) AMD's DTX motherboard standard seems to have gotten no real traction."

I think this is absolutely hilarious after all the noise he made about DTX. Is anyone else surprised at all?

Anonymous said...

Losing share to Via that is simply embarrasing.

What is Scientia and Sharikou blogging these days, nothing at all.

Poor AMD cum monkeys

Axel said...

The recent Mercury Research findings (HUGE swing to Intel btw) remind me that Scientia has been mum for quite a while on AMD taking market share from Intel. For nearly a year following the C2D performance previews in Spring 2006 Scientia kept insisting that due to the unstoppable ramp of almighty Fab 36, AMD would continue taking market volume share at the rate of 0.6% per quarter and there was nothing that Intel could do about it.

Scientia failed to realized that the ramp of a fab is largely driven by demand, and the demand for X2s dropped off a cliff even as AMD cut prices on their flagship products to embarassingly low levels. Fast forward to today and we have one fab idling and being converted to produce 45-nm Shanghai, while the other one (Fab 36) is underutilized producing undesired 65-nm product that AMD has to practically pay people to buy. What happened to taking 0.6% volume share per quarter?

Yet another massively flawed prediction swept under the rug and not mentioned anymore.

Tonus said...

Was the prediction of market share gains based on the assumption that Intel would be capacity-constrained? And is this the case at the present time? I think Ed at Overclocker's was saying that Intel was going to be behind schedule on their 45nm ramp, if I am remembering correctly (something about problems with either chips or a fab).

How does that play into the numbers? What will AMD have in the way of fab capacity when Shanghai is ready to ship, and how will that affect product availability?

Orthogonal said...

I think Ed at Overclocker's was saying that Intel was going to be behind schedule on their 45nm ramp, if I am remembering correctly (something about problems with either chips or a fab).

Without getting into specifics, the delay is solely due to F28 in Israel. F11X in New Mexico is, and always has been, scheduled for a Q4'08 Ramp and is still on target for it. F28 has had a lot of issues (Not process related) in ramping, and being built in the first place.

All I know is that I would not want to be a manager over there. Heads may roll!!!

Anonymous said...

Scientia what a joke

These days whines about unfair benchmarks. Benchmarks are one small part in the larger picture that makes a product successful, a product successfuly, and money for a company. In the case of AMD no amount of benchmark manipulation, PR BS, or fanboi commentary will change the fact their product sucks, their process sucks, their management hasn't a clue.

Yes it is funny that I've been reminded of Scientia's old prediction of AMD taking MS away due to great ramp in Dresden. He forgot the little detail called competition that had superior process, superior resources, and superior capacity.

AMD is finished, Hector dug most of the grave, Dirk will be laying the casket.

The clock has run out on them.

Roborat, Ph.D said...

http://www.fabtech.org/content/view/65676/

Fabtech is now making a u-turn and now thinks asset-smart = fabless.
We can expect Scientia to, yet again, accept he was wrong. That AMD will never go fabless and that SMART is some process improvement concept. Way wrong!

Anonymous said...

AMD fucked if it does, fucked if it doesn't

Ass smart.. Perhaps first they should have not bought ATI and instead spent 5 billion on fab technology and accelerated their 45nm and 32nm technology.

Then today they'd have 45nm in production and 32nm on the horizon.

Imagine Barcelona 50% smaller and 20% cooler.

Nah, it was more exciting to pile on the debt and try and grow versus focus on the basics. Invest in quality manufacturing process. It ain't glamourous, it ain't easy, takes billions of dollars and years to make happen. Nah.. lets go buy a company and make fancy powerpoint about how 1+1=3.

Turns out Hector's folly of 1+1 = 0 at the moment.

Ass smart my ass.

Anonymous said...

Lex, why not just login in and use your name?

Anonymous said...

Robo - The INQ had the same link on their site and AMD PR told them that Dirk was misquoted by that Austin link. If you read it you'll also see they have no direct quote from Dirk so this is quite possible.

All that said, I think fab-less is an eventuality, though it'll be dressed up in some majority owned subsidary or joint consortium to satisfy the x86 license terms and give AMD a PR attempt to save face.

I don't think the link to the Austin Statesmen is valid though.

InTheKnow said...

Without getting into specifics, the delay is solely due to F28 in Israel. ....

All I know is that I would not want to be a manager over there. Heads may roll!!!


This is a much bigger issue that it seems to be on the surface.

With F28 a quarter behind, where did Intel get the capacity they were planning on? I'm betting that they kept D1D on the 45nm process for a quarter longer than was originally planned.

If my speculation is right, that means that Intel is behind on the ramp of their 32nm process. They wouldn't be behind in development, but they would be behind in terms of volume learning.

Development is done on small volumes with a focus on hitting device parametrics. The big yield hitters will be shaken out, but there typically hasn't been enough volume to see everything that can go wrong. You need to crank up the line to production volumes to find the kinds of problems I'm talking about.

They still have over a year before 32nm goes out the door, so it isn't going to be a total catastrophe, there is time to catch up to where they should be.

But I expect Intel to have a very rough start up on 32nm because of this. They'll get it out the door, but I'll be pleasantly surprised if yields are not behind where the last several process nodes have started.

SPARKS said...

Orthogonal-

Correct me if I’m wrong, and I probably am, but I have read somewhere that the Israel Fab delays are due to INTC not securing adequate power for facility itself. Something about a power generation facility built by INTC. From where I sit, this power would need to be substantial and exceptionally reliable.

I think any thing regarding actual 45nM process is purely speculative and nonsense. QX9770 is running too well (so is everything else down the line).

After all, and I know very this well, NOTHING moves without plenty of reliable juice on tap.

SPARKS

SPARKS said...

Could this be true?!?

http://www.theinquirer.net/gb/inquirer/news/2008/07/23/amd-confirms-fab-spinoff

SPARKS

Orthogonal said...

Correct me if I’m wrong, and I probably am, but I have read somewhere that the Israel Fab delays are due to INTC not securing adequate power for facility itself. Something about a power generation facility built by INTC. From where I sit, this power would need to be substantial and exceptionally reliable.

Tip of the iceberg, something convenient to throw the blame at. I've never seen a fab ramp done so poorly on so many levels (granted I've been involved in very few, but from what everyone else says, it's true). It looks like it is going to be ok in the end, it's just that the fab will be more than 6 months late as a result.

jumpingjack said...

Sparks,

If one has followed the history of AMD, before Intel dissolved their Xerox machine and forced them to do their own design, AMD was nothing more than a clone maker.

AMD was, in the past, a good semiconductor maker of a variety of products (not of just of CPUs, during the time when they were not a CPU company).

Interestingly, over time they whittled themselves down, and dissolved or divested themselves of businesses (AGP chips, DSP, Flash) to the point they basically only knew CPUs ... until --- tada ---- ATI.

Assuming that asset smark is indeed what we are hearing... then AMD is now going to divest themselves of a semiconductors' companies greatest asset (and, if poorly managed greatest liability).

I have watched over the years AMD acquire, fail, divest, dissolve, shut down fabs, sell off buildings. The idea of splitting off a manufacturing only company and a design company, to me, is them choosing between the lesser of two evils. Ultimately, considering the complexity of design on a leading edge process technology (which is needed for competitive CPUs) amounts to the first step toward AMD ceasing to exist.

This is bad for the industry as a whole ... and we can only hope -- if the split it down the middle concept is true -- that Design AMD Inc, does not follow in the footsteps of Transmeta (another fabless CPU maker).

I am worried about AMD's viability.

Orthogonal said...

If my speculation is right, that means that Intel is behind on the ramp of their 32nm process. They wouldn't be behind in development, but they would be behind in terms of volume learning.

P1268 development (32nm node) is still on schedule and there is no delay in deployment to HVM. The only thing this has affected internally is which fabs ramp which process and on what timeframe. They've changed a couple of times in the last year and I'll bet they'll change again before finally set in stone and made public.

Axel said...

Anonymous

The INQ had the same link on their site and AMD PR told them that Dirk was misquoted by that Austin link. If you read it you'll also see they have no direct quote from Dirk so this is quite possible.

Well it's really a moot issue on whether the Austin Statesman misquoted Dirk or not. The bottom line is that Bob Rivet told us what Asset Smart is during the Q2 earnings CC, as Robo points out in this blog entry. Rivet categorically explained to the world that Asset Smart will remove two of the following three things from AMD: process development, factories, chip design. The only two that make sense are the first two. We don't need Dirk to figure out this riddle.

Anonymous said...

Well it's really a moot issue on whether the Austin Statesman misquoted Dirk or not.

I agree, and I thought that I indicated that in my post - it's just now that several sites seem to be take the Austin Statesman link as gospel. Given that it is Austin, and probably has some interest in AMD news (?)- they may have tried to creatively spin the news? Especially as the INQ got a statement from AMD PR saying it was incorrect.

So now you have a single source rumor that gets cut and paste on several tech sites (INQ, Fabtech and probably others) and suddenly it seems like there are multiple confirmations of it.

So while AMD will likely do this approach, the INQ and fabtech (both quoting this Austin website) really have no new information on the subject.

Folks need to be less anxious - initially any arrangement will have zero impact on AMD operations and products. Intermediate term it will help the cash flow (lower capital spends) but hurt margins - so it is not clear if there will be any impact on pricing (likely the competitive situation with Intel products will remain the key hurdle). And long term, barring all industries becoming commodity businesses; AMD will be purely a "value" player in the market (meaning low cost).

As you are now paying the foundry a margin on top of the cost to produce the wafer, this will be a difficult grind, unless the foundry has significantly lower production costs than Intel or AMD has a significant design advantage that enables smaller die sizes. With SOI on the roadmap for the forseeable future, Intel seemingly able to maintain at least a year lead on transitions to smaller nodes, I think it is unlikely for the foundry to be lower cost (heck, it'll probably be hard to even be even on cost). So it will come down to whether AMD can pull a design rabbit out of the hat again (or Intel pulls another design terd out of the hat)

Being tied to a foundry when your competitor is not, is a recipe for a commodity business, barring Intel falling down. With a move like this AMD places their long term success in the hands of Intel making a serious mistake (or multiple mistakes) or possibly governmental intervention - and this could be what AMD is banking on.

I'm sure some fans will try to spin working with a foundry as no big deal and throw around terms like DFM enabling this (you know who I mean) or say the model has worked well with graphics so they can just follow that model - but in the end they will just be flapping their gums about something they have no direct knowledge about, just a G.D (google doctorate in web searching)

Anonymous said...

Ortho have you become the unofficial offical leak spokesmole at INTEL? Process engineer in marciopa... Beware ip address make it easy to find ya!

Is that F28 that late, damm whomever is in charge should be fired for incompetence. That simply isn't like INTEL to slip schedules if indeed it is slipped 6 months. That is a lot of billions doing nothing!

Still waiting to hear about AMDs ass smart plan

TIck tock tick tock the clock has run out on AMD

Tonus said...

What would spinning off the fab business mean for AMD? Would it be the same as outsourcing chip production to (for example) TSMC, or do they gain some benefit from it (ie, have it both ways to some degree)?

Splitting one company that is in bad financial shape into two companies that are in bad financial shape doesn't sound very productive, so there is lots that I am missing here.

Orthogonal said...

Anonymous said...

Ortho have you become the unofficial offical leak spokesmole at INTEL? Process engineer in marciopa... Beware ip address make it easy to find ya!


It's been well known on this blog for quite some time that I work at F12 in Chandler, AZ. I'm not exactly spilling the beans here since it's known there are problems, just not very many public details, of which I'm not likely to disclose.

And you can drop the charade's Lex, you aren't fooling anyone.

InTheKnow said...

Orthogonal said...
P1268 development (32nm node) is still on schedule and there is no delay in deployment to HVM.

Let me clarify my statement. D1D has two groups working in the same fab. There is PTD (process development) and Ramp. What I'm saying that the delays at F28 must have done is keep the ramp group on the 45nm process longer than initially planned.

At the risk of sounding like another blogger, logic would seem to dictate this. The reasoning goes like this.

There are only 2 fabs making 45nm, F32 and D1D. At some point in time, D1Ds volumes will be taken over by F28. This will release D1Ds capacity for ramp of 32nm. But F28 is late. Therefore, it can be concluded the F28's delay has required that D1D's volume production is needed on 45nm for longer than initially anticipated.

Note that this does not imply that global schedules have slipped. All I'm saying is that D1Ds internal ramp to volume has probably been delayed. And that delay will mean a loss of high volume learning before moving the process to HVM.

Hence my expectation of a rough start for 32nm.

Orthogonal said...

Blogger InTheKnow said...

Let me clarify my statement. D1D has two groups working in the same fab. There is PTD (process development) and Ramp. What I'm saying that the delays at F28 must have done is keep the ramp group on the 45nm process longer than initially planned.


Ok, I see what you are saying. I honestly don't know the answer to your supposition. I would guess that D1D would not sacrifice any schedule for the 32nm ramp to support 45nm, but the business need could have trumped it in the short-term.

Anonymous said...

And you can drop the charade's Lex, you aren't fooling anyone.

Ortho - don't worry Lex will get real tricky on us, like going anonymous (wait he's already done that). Of course he is not intelligent enough to change his tagline.

He could get rid of the tagline or change it to something else real subtlety 'the clock has run out on them'.... real sneaky! Of course the dilemna is he wants to be identifiable so he can't change it too much, but apparently he got bored (or embarrassed?) with 'lex'.

Just ignore him and don't feed the troll. In the event he drops the stupid taglines and makes a semi-intelligent comment then it is worth replying to anyway. Give him the choice - he can remain effectively non-anonymous with his blather and stupid tagline and crudeness and be ignored or he can clean it up.

I guess the third option is he can get back in delusional world and think people are asking for his ban and attempt to play the victim. Heck I wouldn't be surprised if he posted comments against himself just to stir things up and further play the victim role.

(Though this is not one of those comments as Robo could confirm via ip addresses)

Anonymous said...

Splitting one company that is in bad financial shape into two companies that are in bad financial shape doesn't sound very productive, so there is lots that I am missing here.

They will almost certainly burden one of the companies with the majority of the debt (the foundry). This is a way of attempting to spinoff debt. Though in this case there are a few issues:

1) I don't see anyway around AMD avoiding having to take a majority stake in the spinoff (due to the x86 license terms). So the "healthy" AMD part will still own 51% of the debt.

Clearly the exit plan for AMD is to hold on long enough to after the x86 license is renegotiated in 2010 and then dump this ownership and write off the losses (like Spansion). I have a bit of a moral issue here as if the spinoff is public, AMD is just screwing the shareholders (again)

2) They need to find a sucker, ummm partner, for this. There clearly is no short term profitability / stability prospects so you need owners with ulterior motives.

3) You have potential IP issues. Remember much of AMD's process is licensed from IBM. If the partner is outside the IBM fab club you have to negotiate this piece, unless IBM is simply willing to donate its IP to a third party.

4) AMD has a market of 2.67Bil, assuming the design house is worth some part of that and keeping in mind AMD will only be getting an inufison of 49% of the value of the foundry (you also probably should take into account that they will probably get some premium for the spinoff, much like a takeover attempt).

But still even if you assume ALL of the value is in the foundry, you add a 30% premium, and then AMD gets 49% of that, you are talking ~1.7Bil (and again that assumes the AMD design spinoff is effectively worth $0 which is clearly not the case). So if I had to guess you are talking about ~1Bil infusion which will likely be used for cash flow to keep the foundry alive until AMD can dump it's stake in it when the x86 license is renegotiated.

Now I've heard some people thinking - state of the art fab is 4Bil, couple it with assembly fabs and other assets and AMD should get more than 2Bil no? Well this completely ignores that the buyer will also be likely assuming the debt obligations AMD currently has (49% of them)

So I read this as a cash infusion and road for AMD to become a design shop. While this might be the best for short term survivability, it does seem to hamstring them longterm if they manage to come out of this.

SPARKS said...

Whoa! Easy wrangler! Don’t be picking on Ortho!

I’m a know nothing enthusiast from Podunk and I read about the FAB 28 delays months ago. Anybody with an ear to the railroad tracks knew this. (Read: Stockholders) He ain’t given up nothin’.

Ortho-You just keep cranking out those BEAUTIFUL, LOVELY chipsets, bro! If anyone messes with you, GURU will eat them for lunch! I truly fear his wrath!

BTW: X48 is absolutely incredible. It has taken the place of my all time favorite chipset 850E w/RDRAM.

My X48 fired up1800 MHz native on the first boot and the son-a-bitch is stable as a brick! In fact, I’m looking at 2000 MHz memory so run the QX9770 @ 4.27 on water.

Well done, for the EOL FSB, it’s going out in style. It was the overclocker’s ticket to Valhalla!

HOO YA

SPARKS

Anonymous said...

My X48 fired up1800 MHz native on the first boot and the son-a-bitch is stable as a brick!

Yeah but who cares if it is stable or it can run 7x24 or under full load during any useful applications - can it run Prime95 on all cores? because that is the true measure of usefulness and stability! I mean if you can't be running 4 instance of Prime or using Prime in the background while running other applications, what good is it?

And water cooling - so what if it actually keeps the temp OK... that's cheating! You are not allowed to CHOOSE to do that to enable OC!

Remember the rules of overclocking: Don't tell anyone about overclocking! 2nd rule - only air cooling is allowed! 3rd rule - must be able to run a useless program just to prove something. 4th rule - if you achieve something better than I intuitively believe, then you must be doing something abnormal or cheating!

Oops sorry wrong site, :)

SPARKS said...

Woo! Ya had me going there for second! I was up on my haunches ready to fire back a reply with fissionable material!

Man, that guy is pathetic.

Anyway, my accolades regarding this fantastic combo may seem rather superfluous and redundant at times, perhaps even boring, at worst, bragging. But, this upgrade comes on the heels of a 975X/955EE combo. (With a Q6600 sandwiched in between to test the waters) This setup was and is worth wait, beyond words, worth the price, and without doubt an enthusiasts dream come true.

KUDOS to all at INTC

(Oh Larrabee, Larrabee for where art thou, Larrabee)

SPARKS

SPARKS said...

Hey GURU, what is this thing and is a RISC based processor. Frankly, I didn't have a clue.

What's INTC up to here?

http://anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3361

SPARKS

InTheKnow said...

2nd rule - only air cooling is allowed

That would be cooling by free convection. Using fans and forced convection is for gurly men!

jumpingjack said...

"What's INTC up to here?" ...

Sparks,

A few points -- the SOC, this particular chip, makes an interesting headline ... but for the consumer level user, this is not quite the product/visibility of the CPU. This chip is for embedded applications -- think of things like smart enviornment monitoring in your car, or a fancy network hub, or home automation controller... etc etc. ARM, powerPC, and other pretty much live here -- the deal -- they do not have the wide software base that x86 has. Imaging your refrigerator sending you a text page via the internet that it has problems and the food is reaching spoilage conditions :)

Arstechnica did a good piece on this: http://arstechnica.com/news.ars/post/20080723-intel-launches-new-tolapai-system-on-a-chip-design.html

The concept is to bring x86 into very specific, and power/space crucial applications, and with the software support makes integration into the common fabric easier.

Now, a commentary ... what has, historically, always separated the quality of Intel from AMD, in my opinion, is that Intel learns from it's mistakes and takes dramatic measures to avoid them..... AMD habitually repeats theirs...

Example --

- Copy exactly was born out of learning how to transfer technologies to other fabs as I read up on Intel's history of this philosophy....

- Pentium 1.13 Ghz chip that bellied up .. ever sense then the guard band has been 20%+.

- Netburst to Core transition, the right hand turn.

In this case, Intel learned that it is best to reinvent your core compentency rather than try to grow by simply engulfing existing technology -- i.e. they divested their RISC processor division, loped off other failing business groups.

The push for x86 into all facets is the right one for Intel -- whether they are successful remains to be seen... however, right now there is good indication they are heading in the right direction.

Anonymous said...

Ortho / intheknow.

From a PTD perspective:

1. So far, the delay to F28 has NOT created the supply shortages of the 45nm chips. Intel is ahead of their commit ship volumes, and on track to crossover 65nm earlier than expected. My supposition: What they have lost (perhaps) in volume they have more made up for in better than expected yields.

2. The 32nm ramp schedule has not been delayed or pushed back. The Westmere design, and process are approaching the committed sync point -- right around 2 years after Penryn A0. You'll see D1D significantly ramping volume *next* year, not this year. Gotta work out the process issues with real product first. And this is *not* out of the ordinary.

SPARKS said...

“Intel learns from it's mistakes and takes dramatic measures to avoid them.....”

“The concept is to bring x86 into very specific, and power/space crucial applications”


Ah, sounds like hype, years back, about PowerPC chips with RISC supplanting x86 in the desktop market, except this is in reverse, i.e., complex x86 will never hit the embedded market. It also sounds like INTC learns from its successes, too.

Nice, that’s all we need now, wise-ass refrigerators informing us that our ‘caloric intake has surpassed our suggested daily quota.’

Thanks Jack

SPARKS

hyc said...

Eh. Just make sure Intel doesn't bring Microsoft with them everywhere they go. I don't need to see a BSOD when I open my refrigerator door. Nor do I want to see that my refrigerator's antivirus database needs to be updated all the time.

There may well be a huge x86 software base to play with, but most of that software is crap. Most of the stuff that isn't crap deals with functionality that's not relevant for embedded processors. Outside of the base developer toolchain not much else matters.

SPARKS said...

Jack-

“I am worried about AMD's viability.”

I had this long drawn comment as a reply. I think the topic has been rendered academic by INTC, however. Obviously, they are going to put the kibosh on an AMD Deneb “launch”. It seems INTC is not going to let AMD up for air. They are going to stage their own “launch”. I seriously doubt they are concerned about AMD’s viability, irregardless of lawsuits and government(s) intervention at this juncture.


INTC just turned thirty. I suppose AMD is roughly the same age.

These are interesting times we live in.


http://www.digitimes.com/mobos/a20080724PD205.html

SPARKS

Anonymous said...

I am worried about AMD's viability.

Call me cold, call me heartless - but I don't really care. If they succeed great, if not someone needs to take the old horse out back and put it down and let someone new come along and give it a shot.

And for the constant whine of those who say we need AMD for competition and to drive innovation - I'm just not buying it. We need competition, but if this is the level AMD plans to compete at, then we don't need AMD as that competitor.

If Intel suddenly had control of the entire market and stopped innovating or making better products, then the market will stop buying upgrades and the market will slow down and Intel will be punished via slower revenue growth, lower margins, etc.

I think competition is good and it drives people (and companies), but weak competition is not competition. Time for AMD to step up or step aside. Frankly I don't care either way - I will buy the better product at my price point; I'm not rich enough to subsidize a company just to keep hope alive and honestly if I were rich I wouldn't do it anyway. So I appreciate those 'fighting the good fight', but keep in mind there is a chance that prolonging AMD may actually be working against the good fight by preventing a more able competitor from sprouting up.

jumpingjack said...

"And for the constant whine of those who say we need AMD for competition and to drive innovation - I'm just not buying it. We need competition, but if this is the level AMD plans to compete at, then we don't need AMD as that competitor."

:) :) Well, I do have to agree to an extent. Free enterprise and the capitalism from competition is Darwinian ... the idea that the strong survive, the competition for the best drives the best to come. I am totally not for rewarding mediocrity ... doing so kills the spirit of invention and slows or even haults progress.

I am rooting, frankly, for AMD's survival but not at the expense of progress.

Ho Ho said...

Scientia can be pretty funny some times. He first says that everyone who believes that 8/10 encoding will be used with QPI are delusional because that's what PCIe1/2 uses and what he thought PCIe3 would also use. When I pointed out that PCIe3 will most likely drop the encoding he answered by telling me why that kind of encoding was neccesary and what kinds of problems it causes.

Too bad he failed to show that PCIe3 and QPI would use 8/10 encoding. Of course there is no way he would fix his mistakes in the original article.

Also he hasn't allowed any of my ~5 comments up I've made during the last month. Though he did have two of them castrated and reposted by himself with cutting out a few crucial details.

SPARKS said...

Saturday, being my day off, gave me some time to scour the web, looking for tech tidbits. Then I ran into this Motley Fool piece. Naturally, I will not paste the whole mess here, for the fear of being severely beaten. But I will go over some key points. The article is titled:


Dreadful Stocks to Avoid


Here’s what they say to look out for in a nut. They elaborate on each point, however, since there are very smart people here, I think most folks here could, quite sucessfully, add more.

1 Businesses that bet the farm
2 Businesses dependent on research
3 Debt-burdened companies
4 Companies with questionable management
5 Companies that require continued capital investment

Uhhh, does any of this remind you of a particular company?


http://www.fool.com/investing/value/2008/06/06/ dreadful-stocks-to-avoid.aspx


SPARKS

A Nonny Moose said...

Sparks said...

"Nice, that’s all we need now, wise-ass refrigerators informing us that our ‘caloric intake has surpassed our suggested daily quota.’"

Heh, some posters here (not you of course :) need a 'wise-ass bar' to tell them not to drink & post.

I for one would like to see smart fridges that know when food contents are about to expire (courtesy of RFID packaging), smart washers & dryers that sense the amount of dirt in clothes or the humidity & water content and can adjust their cleaning/drying actions accordingly, a smart rear-view camera on my car than can detect toys or kids behind me before I back up, and a smart mailbox that automatically jettisons the junk mail directly into the garbage can :). Just kidding about the last. And if the price is cheap enough I'd actually put my $$ where my big mouth is and buy it!

Yes you can go to Expo (Home Depot's upscale showroom) and spend $3.5K on a smart toilet that will remember and recognize your arse signature and set its heat & perfume & cleansing jets accordingly, but this is just some personal Japanese cleanliness fetish that the average citizen will pass by, if s/he can buy scented Charmin double-ply :)

Well my smart monitor is buzzing me that my Saturday nite beer content is too high and that I am making stupid posts so I had better quit now :). But I'm married to her so I can't shut her off :).

But before I go, I should note that AMDZone is in terrible shape - apparently Sci is galloping around on his Shetland pony, shooting his popgun at Ah-Ben-Stoopid, m-m-m-marq (only poster who posts in ESL - English as a Stuttering Language) and various & sundry other Zonerz. Seems Sci is terribly bent out of shape since, as Robo pointed out, his "asset smart" blogging was revealed as totally bogus and incredibly stoopid. And this is the bozo who is throwing stones at Anandtech...

Anonymous said...

Seems Sci is terribly bent out of shape since, as Robo pointed out, his "asset smart" blogging was revealed as totally bogus and incredibly stoopid.

He'll get over it - when Asset light/smart/jettison is formally announced, I'm sure he will praise it and AMD, and say this is just what they needed and is the smart thing to do.

Remember closing the technology gap? How's that going?

Remember RDR's (restrictive design rules) being over-hyped as a reason Intel was doing well? Now all you hear is how 'DFM' is going to give AMD some great jump on 45nm to close the gap with Intel.

And on a side note - this DFM (design for manufacturing) crap is getting totally overblown and misunderstood. This is just another TLA (three letter acronym) to make something which has been done since the 8" days sound like something special. DFM has been around under many different names and incarnations... the key thing is the EXTENT and EFFECTIVENESS to which it is deployed (i.e. you don't want to implement design rules if they have no benefit - the key is understanding where the real lever bars are - if you go overboard you start to introduce significant cost and inefficiencies).

Scientia seems to think it is something new which is being implemented. As a process - - - - (Sparks can fill in the blanks), I find it as hilarious as his Asset Smart 'analysis' (and I use that term loosely)

Sorry for the rant - JJ has his TDP, and I have my DFM/RDR pet peeves.

Anonymous said...

the key thing is the EXTENT and EFFECTIVENESS to which it [design rules] is deployed

Spot on. As somebody who was present to suggest design rules for 45, 32, and 22nm, I don't think *anybody* (but a handful of some of you) know precisely how much of a balancing act you have to play.

Engineers think designers are a bunch of lazy sobs who just want the easy way out. Designers think that manufacturing is waay to conservative and they should just spend the money and make it right.

Both sides have their points. =)

Sci can bleat all he wants about DFM, but this is all predicated on designers and engineers sitting in the same room, looking at the data, and making the hard decisions. It doesn't work if you have designers in Dresden/Austin and process developers in Fishkill.

Anonymous said...

"As somebody who was present to suggest design rules for 45, 32, and 22nm"

Perhaps you can give a few examples as that would help people better understand what you are referring to? (don't need exact #'s as that would be proprietary).

SPARKS said...

“As a process - - - - (Sparks can fill in the blanks),”

Uhh? I’m still trying to figure out what the hell is going on!

DFM???? I have a TLA for you, WFB – Why F------ Bother?

Why the hell would you design ANYTHING if it were NOT going to be manufactured (en mass), design elegance, engineering excellence awards, save the plant, altruistic/religious doctrine?

Dementia is an idiot.

--------------------------------------------------------


“Engineers think designers are a bunch of lazy sobs who just want the easy way out. Designers think that manufacturing is waay to conservative and they should just spend the money and make it right.”

?????WHAT????? Here’s another acronym--- GMFB!

Hey, if they tell me I have to run 4-3in rigid conduits up a 20 story shaft and there ain’t enough room in the shaft, no money in the world is gonna get ‘em in there! If they reply with, “yeah but we need 1200A” my response would be, “lets see if ((((WE)))) can find a better way!”

It certainly wouldn’t be, “you lazy slob” or “your approach is too conservative”. They key word if you haven’t noticed is ‘WE’.

I realize this is an extremely simple analogy to a very complicated and coordinated effort. But, if I were a team leader (in my field I am) and someone had that attitude, they could count their time with me with an egg timer!

There’s no ‘I’ in TEAM.

Further, I’ll bet dollars to donuts no one would ever talk that bullshit in front of a guy like this:

http://www.intel.com/pressroom/kits/bios/mbohr.htm

You’d be selling pencils in the parking lot, right next to the coffee truck.


“Sorry for the rant - JJ has his TDP, and I have my DFM/RDR pet peeves.”

Sorry for the rant- you guys have yours, this one is mine.

SPARKS

A Nonny Moose said...

Anonymous said ...

"Remember closing the technology gap? How's that going?"

About the same as DTX?? Or maybe SSE5?? Weren't those the 2 most important developments in all of 2007? :) After pooh-poohing HKMG and stating AMD's 65nm process prior to the B3 stepping 'is not broken', the illustrious Scientia now thinks the first is a good idea and will enable Shanghai to quickly catch up with Intel at 45nm.

However it's not just Sci, it's all the bozos over at AMDZone. Yes, all 4 of them :).

It's amusing, on the same level as watching Jerry Springer or Maury Povich bring on some trash-talkin' 400lb beeyotches to go a-clawin' & a-slappin' at each other over which out-of-work boyfriend fathered their last 8 welfare children. Only here the 'host' is Ghost, who thinks Intel pays people to come onto AMDZone and do viral marketing for Intel.

Anonymous said...

Sparks:

I think my point got lost.. the fact that you've got designers and engineers in the same room .. that *means* they are on the same page. It opens the communication, so *slowly* designers understand why the process engineers demand more and more restrictions and the process engineeers *slowly* understand why restrictions are bad.

It truly is a team effort -- and it's the only way you can be successful.

So, as for examples, I can only safely speak of 45nm since it's the only thing in the wild, so there is no "news" here. But that was the first layer which really had a lot of restrictions. I'm going to try to keep it VERY generic.

1. Aligning the poly on a grid. Before that, you had gate poly, field poly, and all sorts of routing stuff. Designers like the flexibility to make different performing transistors, but they didn't like the variability. Litho was really having difficulties with that. They ended up with a gridded poly. Made litho *really* happy, and really brought the variability down. The designers were worried about the floorplan, but revisited their routing/layout tools and recovered efficiencies there. It did mean that the poly routing had to be done on different layers .. contact and M1.

2. Speaking of M1 and contact, because of the choice of *dry* litho at the small pitches, a significant amount of restrictions needed to be placed on pitches and there was not the same amount of shrink in 2D regions. The limited pitches/linewidths was cantankerous because, while patterning would struggle to support that, it really affected the size (the number of rows) of various logic building blocks. It was a pretty big battle -- what you gain in manufacturing, you lose in die space. In the end, you take a hit in both.

In all of this, it's *data driven*. Everybody is on the same page -- and when anybody tells the group "It's impossible"... they are expected to back it up with data. Both sides then go back and see what that means to their camps -- device performance, die space, manufacturing capability and variability. Once it's minimally acceptable to all, that's it.

Clair Webb gave a great talk on this at the SPIE advanced litho in the DFM session. (and I've vetted my comments with respect to that) If you can look at the paper, you'll get a better idea of what I'm talking about.

A Nonny Moose said...

Some of the more notoriously inept Scientia postings:

Wednesday, February 07, 2007
Intel's 2007 Processors: Mist In The Morning Sun

...So, 2007 is now shaping up to be pretty much even in terms of performance between Intel and AMD. And, 2008 shouldn't be much different since IBM/AMD also have high-K and will use it on 45nm just as Intel will. The talk about what Intel would offer in 2007 looked like giants in the mist. And, now that the mist is clearing we can see that Intel's processors don't stand any taller than AMD's.

Tuesday, January 30, 2007
Q4 06 Intel and AMD: Who Won?

Sometimes it is easy to pick a clear winner and sometimes it is more difficult. The fourth quarter of 2006 for Intel and AMD is hard to call because of conflicting numbers for revenue share and volume share.
...Overall, in 2006, there is no doubt which company is the winner. AMD added a second FAB and the ability to produce its own chipsets while gaining revenue and share while Intel lost both revenue and share in spite of releasing a new architecture. This puts AMD in a much stronger position in 2007 than it was in 2006 and it gains a new architecture on top of that. This makes for a very strong offering by AMD in 2007. However, if we
take Intel's bag of tricks for 2007, turn it upside down, and shake it hard, not a lot falls out.

Thursday, December 28, 2006
2007 -- A Year Of Promise

...As to the idea that AMD won't deliver K8L on the desktop until Q4 07, AMD has contracts with one supercomputer that will be running by end of Q2 07 so K8L production has to start in Q2. Desktop versions of K8L will come out in Q3. Apparently, the fact that the intitial supply of K8L is tight because of preordering has led some to believe that production won't start until Q3.

Thursday, October 19, 2006
A Real Forecast

Although there have been many projections lately that favor Intel, it is my guess that these are based largely on old biases towards Intel. I believe that a more careful look at AMD and Intel shows that AMD's performance and current position are better than Intel's. And, although I believe that Intel hit bottom in the last quarter and will now grow, it seems likely that AMD will grow faster during 2007 and 2008.

Saturday, October 14, 2006
What Quads May Come

I had a technical look at AMD's K8L in AMD Unveils Barcelona Quad-Core Details . There are improvements that I had not heard about before. There are several that should make a difference and boost both K8L's SSE and Integer performance. From this information it appears that 2007 will be much tougher for Intel and that it will lose most if not all of the ground it gained with the release of Core 2 Duo.

Thursday, September 28, 2006
IDF, AMD And Intel In Perspective

...It is clear that Intel will not abandon the FSB because there would be no point in licensing a FSB that was going to be dropped and there would be no reason to waste money developing products for a FSB that would be gone by 2008. Therefore, we must assume that Intel is not dropping the FSB. This also has to mean that Intel is not releasing CSI in 2008 and is not following AMD's lead to an onboard memory controller.

Anonymous said...

Got PTD prospective

Got HVM CE Monkey prospective

Only thing we need here is circuit annoucements here.

Keep sharing dudes!

Anonymous said...

What is fair?

Scientia these days can only find this topic as there is nothing else to talk about with AMD.

Fair is to let companies compete!
Compete includes things like subsidies. If a company can get kick backs from goverments for tax breaks or even free loans or such that is okay in my book. In principle this is still free market as goverments and states can weigh the benifets for the country versus the cost.

As to subsidies these exclusive deals are no different then what you find at fast food resturants where incentive pricing finds only one brand choice. Offering volume discounts or incentives is free market competition. Even at the direct requirements for limititation of the competition. This is free market. AS LONG AS NEITHER COMPANY is SELLING BELOW COST, the CONSUMER WINS. As long as consumer wins then its free compeition. Those in the EU don't care about free competition. They are a becoming more and more socialistics. They live on glories and wealth acculilated from century ago. These days Europe is in full decline as is the US. Only in the East do you see capitalism truely exist. It isn't for free, but in the end free enterprise is the best approach. It is what built all the great economies. In the end they will come around and put the right regulation in as they get rich enough. In the end they could end up stiffling just like the EU stiffling competition to protect lazy worker work hours and long slow decline.

Now as to AMD, they should compete on technology and products. Hector is nothing but a lazy lucky mexican. Since he has ruined INTEL his only hope for a legacy rightly or wrongly is to see INTEL dismantled. If successuful it will be like what happened to IBM and ATT. At the time it made the fanbois and leftest socialist happy, but in the long term was a huge mistake to the competition and progress. It castrated great companies made them pussies like that made the leftist fanbois happy as they ended up like themselves, pussies with no balls

Tonus said...

anon: "And for the constant whine of those who say we need AMD for competition and to drive innovation - I'm just not buying it. We need competition, but if this is the level AMD plans to compete at, then we don't need AMD as that competitor."

I am a capitalist at heart. My desire is for AMD to become a viable company on its own, not by any other means (lawsuits, government decrees, etc). I agree that in their present shape they are not providing that competition, and it seems as if they will not. If they die or become irrelevant because they couldn't hack it, I can live with that.

And no, I do not think that Intel would stop innovating or doing anything to harm itself if they were the only game in town, and I don't think anyone else does, either. But they would be in a much better position to dictate pricing and the pace of progress from top to bottom if there were no competition. I do not feel that Intel will slow progress to a crawl and price their CPUs out of everyone's reach-- I don't think anyone believes that, either (well, there are some that may).

I like competition because it forces the players to move faster than they'd like and to price items lower than they'd like. That benefits me, the consumer. The fact that one company may win out versus another doesn't bother me, that's capitalism and I believe in it. But I really do like those times when companies have to hustle for my business. The era when AMD and Intel were doing that is almost at an end, and that's a bit sad. But it's also the way things work.

SPARKS said...

"Clair Webb gave a great talk on this at the SPIE advanced litho in the DFM session."

Well, there you have it. Another dead end link where you have to pay 24 bucks to read an 8 page PDF. It seems anything this SUPER GURU says or publishes is guarded like gold in Fort Knox.

In any event, thank you. You’ve given me quite a bit to research and study for the next couple of months.

http://spie.org/x648.xml?product_id=772052


Holy cow, what a can of worms. Clair Webb in corroboration with Mark Bohr filed a Patent (6,762,464) which makes “N-p butting connections on SOI substrates”.

http://www.patentstorm.us/patents/6762464.html

Now if these guys invented it, made it happen, and then said “nah, it’s not good enough, yet” what was AMD thinking if THEY could make it happen? Further, if AMD used the process wouldn’t they have to license the salicide/titanium/nickel technique from INTC/Webb/ Bohr, etc.? Or perhaps they didn’t (in conjunction with IBM) which why SOI was, and has been, so difficult for them to successfully implement. I’ve read, however, INTC hasn’t really given up on (FD-SOI), they just put aside for now.

“There is another type, however, called fully-depleted SOI (FD-SOI) that is under investigation at Intel and is not being used by any chip makers today.”

http://www.intel.com/pressroom/kits/45nm/IntelHigh-K_metal_gate_glossary_FINAL.pdf

Any Way, Thanks Again.

SPARKS

Anonymous said...

INTC hasn’t really given up on (FD-SOI), they just put aside for now.

This is a bit of speculation coupled with some knowledge of the area.

First - SOI technology was (and some may say still is) not mature enough for a fully depleted implementation - the reason for this is that you need a rather thin upper active Si area and because it is so thin, the uniformity requirements are on the single Angstrom type range (essentially atomic level uniformity across the wafer).

Second - when Intel was migrating to 300mm, the substrate industry was having enough trouble just to get bare Si wafers ready. The cost of these wafers were rather steep in the early days, fuhgeddabout the cost and volume availability of an SOI wafer at that point in time (obviously as volumes have gone up and the tech has matured, costs have come down, though 300mm SOI adds quite a bit to the finished wafer cost).

Third - the benefit of SOI was demonstrated to lessen as you scale to smaller nodes. So as the technology matured and costs started coming down, that would be the point at which the technical benefits would be diminishing. As AMD lagged on both the 300mm and tech node transitions they had a bit bigger window. It can be argued on both sides whether SOI was worth it or not for AMD - there really is not enough public data to prove either side 'right' or 'wrong'.

Finally - I think Intel's look at SOI now has little to do with the technical improvements in transistor performance which folks try to spin, and more to do with the potential for memory applications. As usual, Intel is far more pragmatic, cost and manufacturing conscious about the technology decisions. They will not do this simply to eke out a slightly better transistor for a short period of time or to create some PR. If the new memory techniques are viable, this will cut down traditional SRAM cell sizes anywhere from 3X-5X and this might offset the added cost of wafer production with SOI. It is a question of whether or not the the potential die size reduction offsets the added cost of SOI (assuming the 2 options are comparable technology-wise and they both can scale)

I look at Intel's approach to SOI like IMC or FSB or monolithic vs MCM or whatever. You can argue IMC/HT was better than FSB on paper, but in reality there was not so much real tangible benefit in most applications (especially in 1P setups). You can argue monolithic quad is theoretcially better but is there any actual clean data on this? And at this point in time is it significant?

Bottom line - Intel will put away its pride and do what is the right technology, market, and business based decision. It is not about who is copying whom, it is about doing the right thing at the right time (and this doesn't mean technology only, there is a cost, sales and marketing piece to it as well).

The one major time Intel did not do this, they got burned (P4) and they seemed to have learned their lesson. So you can have all the academic arguments about what is theoretically or technically better or who came up with this idea or that...ultimately it is about who brings the product to market at the right time and at the right cost and with the right infrastructure support. History is littered with failed companies and products who were merely first to market.

Tonus said...

anon: "the reason for this is that you need a rather thin upper active Si area and because it is so thin, the uniformity requirements are on the single Angstrom type range (essentially atomic level uniformity across the wafer)."

This stuff never fails to amaze me. It's just incomprehensible, I think. We can talk about the numbers but they just don't give us a proper sense of scale. I guess it's the other extreme from being told how far away from us the next galaxy is.

A: They need to make that top layer very thin.

B: Thin like a sheet of paper?

A: Not quite.

B: Oh, so it needs to be thicker, then?

A: Uh, no. Thinner. A lot thinner.

B: Thinner than a sheet of paper?! That's impossible!

A: It's not impossible. Compared to what we need, paper is very thick.

B: Really?

A: Thousands of times thicker, actually.

B: Oh, the room... it's spinning, I think I need to sit down...

Anonymous said...

This stuff never fails to amaze me.

Q: So, looking at that TEM picture, how thick is that gate oxide layer?

A: Around 11 A (Angstroms)

Q: No, I think it looks more like 12A, don't you?

A: Well the layer is amorphous so it depends on where you take the measurement.

Q: So it's 12A?

A: Sure, it could also be 11, it could also be 13A... (Shakes head) What would you like it to be?

SPARKS said...

“This stuff never fails to amaze me. It's just incomprehensible,”

Whats incomprehensible is that they are doing this on 300mm wafers, probably the size of the Moon’s orbit in comparison, while their spraying silk jackets, tailored to fit all the little people (transistors), by the hundreds of billions, and keeping them all happy with a nice fit.

These guys' 3 dimensional conceptual visualization, layer by layer, trace and pathways, with chemical interactions, transistor by transistor, are what are truly amazing. This comes from the collective minds of tens of thousands, over decades.

Oh, keep the highways wide enough to handle all the traffic with no sharp turns or bottle necks! After all, it’s no good if the populace can’t eat and communicate very quickly. It gives a new meaning to ‘working well with others’.

It’s all brilliant, simply brilliant.

SPARKS

SPARKS said...

Oh, here's a reality check.

You got a guy in a bunny suit with clipboard saying, "We got to make our quota this week!"

SPARKS

Anonymous said...

Now here's some OC'ing:

http://www.xbitlabs.com/articles/cpu/display/core2duo-e8600_2.html#sect0

The new dual core 8600:
3.33GHz stock
3.9 GHz at STOCK voltage! (just FSB increase)
4.45GHz @1.4V (about as high as is probably reasonable)

This was done on an ES with the new 45nm E0 stepping (supposed to start shipping in Aug). Oh and it used air cooling and ran some version of Prime and OCCT for stability checks.

I wonder what the quads will do with the E0 stepping.

A Nonny Moose said...

Looks like the Zonerz are about the only ones sipping the green Koolaid nowadays - certainly anyone with actual cash in their pockets (aka investors) is walking away. Despite light sweet crude declining $2 and the stock market up $175, AMD's stock keeps dropping - down to $4.16. In contrast, INTC up 44 cents.

Anybody know how AMDs 'Puma' mobile sales are doing? One thing I noticed in all of Sci's blather was 'wait until Puma' this & 'Puma is going to eat into Intel's mobile share' that. So now I'm wondering if Sci managed to get one prediction right yet.

A quick Google search yields only a few glimpses - "AMD's Puma pounces; misses red-hot netbook market: Jun 4, 2008 ... AMD took the wraps off its new "Puma" laptop platform at Computex ... AMD's mobile market share has long been a weak spot for the company.." from arstechnica.com and "Jun 12, 2008 ... Intel: New Line Delayed, But Gaining Market Share - JP Morgan ... Micro Devices (AMD), despite AMD’s launch of its Puma laptop platform". But these are very early so it's hard to compare Puma with Centrino 2 sales. Guess we'll find out in September when the Q3 earnings reports come out.

Seems to me that the AMD fanbois are always 'waiting' on something good coming out to derail or dethrone Intel. First it was Barcie, and now Shanghai. Sci keeps referring to Intel paper- or soft-launching Nehalem in September/October but apparently has swallowed AMDs promise of Shanghai in Q4 without question (as he did with Barcie's 'for revenue' release in 2Q07. Geez, he was only about a year off with that one). You would think that an intelligent person would have learned by now, not to believe much of anything AMD says.

Anonymous said...

Nehalem performance ... Wow!!

http://www.xtremesystems.org/forums/showpost.php?p=3168831&postcount=816

Anonymous said...

"You would think that an intelligent person would have learned by now, not to believe much of anything AMD says."

I am a bit shocked there has not been a shareholder class action suit filed yet ... AMD has been consistently misleading investors (as well as customers) since the C2D launch.

The best was Dec 06 where they said they would exit 2007 accretive (yeah, right).

Then there was the Mid 2007 launch of the company savior Barcleona.

Then the wild arse suggestion they would reach break even in Q4 07, when that rolled around it was Q2 08 ... then in Q1 08 it was 2H 08.. blah blah blah.

Not to mention publishing faked benchmark data on a processor that did not exist.

Having their Spec scores pulled....

This is not the quality of output one would expect from an industry leader....

A Nonny Moose said...

Anonymous said...

"I am a bit shocked there has not been a shareholder class action suit filed yet ... AMD has been consistently misleading investors (as well as customers) since the C2D launch."

Maybe the stockholders are banking on AMD being successful in suing Intel :). However I think the CEO shuffle was partly done to head off some stockholder lawsuits over mismanagement, and partly done to show investors that AMD is making changes. Of course the Abu Dhabi investors probably have written off their investment already so I doubt they'll be filing a lawsuit - more likely, they'll declare a JIHAD against Ruiz and maybe Meyer :)

Anonymous said...

AMD has been consistently misleading investors

There's a difference between believing this and proving this. There is also a difference between incompetence vs intentionally misleading statements. If you prove the statements were INTENTIONALLY misleading then you need to go the extra step and prove that they were designed to mislead investors as opposed to market the product.

Many of the examples you give are unproveable - with a civil lawsuit you are talking a court of law and preponderance of the evidence, not belief or management incompetence. How are you going to prove what was in their mind unless you have emails or other docuements showing they honestly did not believe what they were saying and were doing so to simply fool investors.

This is not the quality of output one would expect from an industry leader....

First off last I checked Intel was the industry leader. Secondly how does the QUALITY of output have anything to do with a lawsuit.

If folks invested in AMD, quite frankly they deserved what they got - there were clear signs over the last 1-2 years, and to any knowledgeable investor the risks were clear - it's not like they were 'duped' by AMD. Some investors chose to take this risk knowing that if AMD could pull things off perfectly or got bought out then there would be good upside. They were wrong and that is life.

I'm sure there will eventually be lawsuits, but they should not go anywhere. If anyone should be targeted, it should be the board of directors whose job it is to oversee some of the strategic move and senior management - clearly they were a bunch of yes men and I see the board as the ones who truly failed the investors.

SPARKS said...

“If folks invested in AMD, quite frankly they deserved what they got - there were clear signs over the last 1-2 years, and to any knowledgeable investor the risks were clear”

Oh, how true. My morning commute buddy, or shall I say ex-buddy, didn’t listen to a damned thing I told him at the end of Sept 2006. This was AFTER Conroe was kicking the guts out of Opteron. Every time AMD stock took a hit thereafter, I practically begged him out. This was at 24 and change.

He would talk about encryption performance, Spec numbers, 4P scaling, “you’ll see Barcelona will kill Conroe. Native Quad Core, INTEL is too big. Their time has come and it’s over. We are headed for a 50, 50 processor market”, on and on he would rant.

Nothing, absolutely nothing was going to turn this guy around. They were all believers, including the professional Anal-ists (sic). !!!HALF of those jackasses were recommending AMD as a BUY up until JUNE 2007!!!!!

In fact, while we’re drawing up legal briefs, let’s sue those idiots, too!

The mantra for the time (1H 2007) was market share, market share, and more market share. The roof was caving in. but they increased market share. AMD diluted their shares, TWICE!!!!, bonds payable 2012 achieved “Junk” status BBB-, and “market share”, they cried! AMD even sold the backyard!!!!!

“If folks invested in AMD, quite frankly they deserved what they got - there were clear signs over the last 1-2 years, and to any knowledgeable investor the risks were clear”

My (ex) buddy lost over 50 large, we haven’t spoken about the whole ugly calamity to this day. The fact is, I was right; he was wrong, dead wrong.

I sat next to him this morning and he didn’t even raise his head from his NY Times.

SPARKS

Tonus said...

sparks: "Oh, how true. My morning commute buddy, or shall I say ex-buddy, didn’t listen to a damned thing I told him at the end of Sept 2006. This was AFTER Conroe was kicking the guts out of Opteron. Every time AMD stock took a hit thereafter, I practically begged him out. This was at 24 and change."

I guess he was reading the wrong blog. :)

SPARKS said...

Tonus-

“I guess he was reading the wrong blog. :)”

Honestly, being an informally accepted (maybe, more like tolerated) member of this blog is indeed humbling. The unbiased accuracy here has been exceptionally accurate during my tenure. All one needs to do is go back through all DOC’s posts, with their respective comments, to see this. The proof has been the uncensored timeline contained within.

Hell, look at the stuff GURU’s got me reading, just trying to keep up!

“The thermo-mechanical reliability of integrated circuits (ICs) gains importance due to the reducing feature sizes and the application of new materials. This paper focuses on the delamination in the stacked back-end structure underneath bond pads.”

And this is when he was being “VERY generic”

“But that was the first layer which really had a lot of restrictions. I'm going to try to keep it VERY generic.”

Despite AMD’s irrevocable decline, I have had fun, learned a lot about the industry and have become an exceptionally informed investor.

INTC 28 cents a share!

HOO YA!

SPARKS

SPARKS said...

Look at this spew from that idiot Fuddie.

“Deneb ends up a bit faster than Yorkfield quad core but it simply cannot compete with Nehalem, at least not clock to clock.”

Well, this I’ve got to see. It ain’t gonna be faster that MY Yorkie quad core!

http://www.fudzilla.com/index.php?option=com_content&task=view&id=8695&Itemid=1


SPARKS

Anonymous said...

Look at this spew from that idiot Fuddie

He also says it is 'on schedule' for Q4... that was the original schedule? When you publish/speak to about 30 different schedules over time, it is hard for something not to be considered 'on schedule'

Deneb in 2007 could reach 3.0GHz but not more than that

Does he mean 2008 or 2009?(clearly not 2007)

Sparks, with all due respect, Fuad is an idiot and I'm not sure why you bother with it (whether it is Intel, AMD or whatever news). When he can't even cut and paste correctly, how could expect decent analysis.

SPARKS said...

I’m sure you recall Faulkner’s ‘ The Sound And The Fury’ the title taken from a line in Shakespeare’s ‘Macbeth’---

“it is a tale told by an idiot, full of sound and fury, signifying nothing."- (Act V, Scene V).

True, I really shouldn’t bother, however, he does have inside sources. Occasionally, he does come up with some breaking and exclusive news. You need one of your angstrom sized screens to filter it, but he does bring ‘em in.

Dementia is quite another story, I think the line might suit him better, He truly signifies ‘nothing’, as he doesn’t have sources to support his ‘analyses’. I think the full quote is in order here.

“, a poor player that struts and frets his hour upon the stage and then is heard no more: it is a tale told by an idiot, full of sound and fury, signifying nothing."

Pure Dementia.

SPARKS

SPARKS said...

Perhaps this will be more interesting, Nehalem 4P up against 8P Denebs.

http://www.nehalemnews.com/

SPARKS

Orthogonal said...

“If folks invested in AMD, quite frankly they deserved what they got - there were clear signs over the last 1-2 years, and to any knowledgeable investor the risks were clear”

Since we're talking finances, might as well make a little observation with AMD's balance sheet. If you look near the bottom there is a section called "Retained Earnings" which is essentially money the company has made for it's shareholders in its business. When a company pays out a dividend or buys back shares the value goes down. Since AMD has never had a dividend and has never done a share repurchase, their "Retained Earnings" are essentially the Gross Profits of their business during their entire existence.

As of the end of Q2'08, AMD's retained earnings sit at a whopping -$4.647B.

http://finance.google.com/finance?fstype=bi&q=NYSE:AMD

Ho Ho said...

Hehe, this must be my new record: permabanned from amdzone in two days!

As someone there said:
"this is no safezone, this is amdzone". Too bad they forgot to say that only AMD praizing is allowed there.

Anonymous said...

I noticed that hoho's reply to Abinstein - "Someone ought to tell that to Scientia." was deleted. Now who could have done that?

Scientia is worse than the Chinese government when it comes to censorship!

Anonymous said...

As of the end of Q2'08, AMD's retained earnings sit at a whopping -$4.647B.

Well, this makes sense because it's not like AMD's primary goal is to return value to their shareholders. Their primary goal has been to beat Intel, break the monopoly, and get to >30% market share at ANY cost.

Now if their primary goal was to be profitable then they might be in a position to capitilize if/when they get a good product design. Are you better off profitable at 15-18% market share or non-profitable at >20%? AMD in the recent past (especially Ruiz) believed market share was the most important metric. While it is important in order to get to a critical mass where you have developers and OEM's believing in you, AMD had achieved this with K8. You cannot keep going for market share at sustained net losses - you might be able to live this way for a few quarters but then free cash flow starts getting impacted and your ability to spend and ramp and do R&D starts getting impacted.

Just think what would have happened if AMD was able to make $5 more (on average) on every chip they sold over the last 2-3 years. (AMD is doing , what, ~50mil CPU's per year now?) Would this have hurt AMD market share? Probably a bit - but it certainly wouldn't have crashed their market share. The fundamental problem/oversight is - all AMD did by cutting prices was force Intel to keep prices low as well - while this hurt Intel, it didn't get AMD the market share growth they wanted. Seeing this, they should have given up on this foolish strategy - had they raised prices (or preferably slowed the pace at which prices were cut); I would bet you Intel would have slowed their cuts in turn to recover their margins.

The flawed underlying assumption was that AMD's transition to 300mm would significantly lessen Intel's manufacturing advantage and help more dramatically on costs and margins (especially as Intel was already on 300mm) - remember all those armchair technologists calculating how much theoretical market share a single 300mm fab should be able to support? And the capacity benefit going from 90nm to 65nm? And all this spin about APM 3.0 and flexibility and world class manufacturing based on some non-descript Sematech benchmark on their 200mm fab?

By most benchmarks, the 300mm transition yielded about a 30% cost improvement - then factor in that aggregate die size got worse as the transition to dual core and eventually quad cores kicked in. While 65nm scaling helped, again there is a 10-15% cost impact with a 30-40% die size shrink. Sure this is a net positive, but again you have migration to quads and more cache and suddenly you are no longer getting a 30-40% die size reduction and as much a cost benefit as you are lead to believe.

It does seem AMD is now more focused on the bottom line (earnings) than the top line (revenue) but I wonder now at what long term strategic cost. The asset light will certainly help short term on the cost side, but long term I think it is an unrecoverable disadvantage. AMD let the problem get out of hand where they were forced into drastic actions and had few viable options. For this I think Hector (primarily) and the board (secondarily) are to blame.

Anonymous said...

Intel's Barrett sees no slowdown in world PC market

http://www.eetimes.com/news/semi/showArticle.jhtml;jsessionid=FJUWNM2BFJL3MQSNDLRSKH0CJUNN2JVN?articleID=209900716

Additionally I found this interesting:
"It looks as the market is functioning as it should, because every year consumers are getting more for less. We continue to say that, please just look at the facts, don't just listen to a competitor complaint," he said.

And herein lies the rub with the EU complaint... even if some of the allegations are proven true, how exactly did Intel's action HURT the EU consumer? You can argue it hurt AMD - but that is AMD's problem and civil suit, the EU's job should be (?) to protect the consumer, not one American company vs another American company. It is not the EU's job to social engineer the world markets - it is to protect the EU people... so exactly how were they hurt again?

Intel lawyers have previously said that that new charges filed against the company by the European Commission could lead to higher prices for consumers.

And wouldn't this be ironic - by imposing a fine against the big bad monopoly Intel, all the EU will ultimately do is hurt their own consumers in terms of higher prices down the line. Of course conveniently enough the entire world will be paying higher prices - so the EU collects the money but spreads the pain! With government institutions like these 'protecting' the consumers (did I mention the EU consumer will not see a single penny from any fine collected), who needs enemies? I feel so unfortunate not to be living in such an enlightened environment.

This is simply another potential revenue stream to fund more European socialist programs.

Khorgano said...

This is simply another potential revenue stream to fund more European socialist programs.

Unfortunately, European history is rife with Marxist big government interventionism. Whether from the big Fascist states to the hardline Commies, they have attempted to nationalize or control just about every industry in their history at some time or another. Their extreme socialist agenda's, however, will soon (read: next few decades) come crumbling down as they topple under their own weight. Many of the countries are effectively insolvent and have no way of making due on their insane entitlement programs. The US isn't far behind with SS and Medicare...

The EU charges against Intel have nothing to do with protecting the consumers, it just wreaks of a money grab at an easy target.

Tonus said...

anon: "remember all those armchair technologists calculating how much theoretical market share a single 300mm fab should be able to support? And the capacity benefit going from 90nm to 65nm?"

This is one reason that I regard a lot of the forum/blog talk as idle speculation and an entertaining diversion. One thing I like about Rob's site is that there is more talk about how real world metrics (markets, production capability, economic factors) affect the situation in the CPU business.

You can spend all day talking about how certain technological advances give company A a big advantage and act as if this is enough to override market realities, but in the end those realities cannot be wished away. Telling me that processor A is better because it has some fancy technology that really shines when you use synthetic_benchmark_B is all well and good, but don't be so surprised when processor A doesn't sell very well. It turns out that companies looking to upgrade computers aren't running synthetic_benchmark_A.

There is a lot more to the picture than just pure performance in one specific area that can be shown with a single benchmark. Just ask NVIDIA how well their $700 video card is selling these days. That's right, they're not selling them for $700 anymore...

A Nonny Moose said...

Orthogonal said...

"As of the end of Q2'08, AMD's retained earnings sit at a whopping -$4.647B."

I'm sure that is one "earnings" that AMD would have loved to distribute to its shareholders.

AMD: "We are distributing our quarterly earnings of -$1.00 per share. Since you own 10,000 shares, please send us a check immediately for $10,000. As always, we appreciate your continued support and, more importantly, your wallet remaining open for the foreseeable future"

A Nonny Moose said...

Ho Ho said...

"Hehe, this must be my new record: permabanned from amdzone in two days!"

Didn't you get "perma-banned" before?? What exactly is a "perma-ban" if not permanent?? :)

Yeah, I noticed AssMountie calling for the moderators to ban all Intel-loving trolls, which is basically anybody there who doesn't swallow gallons of green Koolaid every day. No deviation from the party line of AMD being superior in all regards, or that Intel is the evil scum-sucking lying benchmarketing scourge of the 8 planets (plus 3 Plutoids :).

Just let AMDZone wither away, much like the company they are so enamored of. By discouraging any real discussion, making everyone dance in the same chorus line, it will become bland & monotonous and will lose lots of eyeballs and any remaining advertising. I seriously doubt those Zonerz will want to prop it up with actual cash (which on average would probably put a serious dent in their fry cook wages I'm guessing).

A Nonny Moose said...

Tonus said...

"Just ask NVIDIA how well their $700 video card is selling these days. That's right, they're not selling them for $700 anymore..."

For a contrarian viewpoint (not my own), check out Motley Fools posting from 7/30:

'The bull case for NVIDIA

All-Star tely1 gets right to the point: "Best product in the GPU market. [Advanced Micro Devices (NYSE: AMD)] finally made the 4870 to compete with them but it only competes cause of price. I say bet on both, everyone wants their computers graphics to be nice. This will lead to both company making profit."So how much of a threat is AMD? According to Sisyphos, not much: "As a gamer I also know that Nvidia cards are currently technologically the better ones. As such, I presume that ATI will be stretched to get to par technologically with their current profit levels and I see Nvidia to get a sustainable lead in this market." Sisyphos adds: "Nvidia is also working on mobile graphic chips. ... 10 years from now, mobile devices will definitely have extensive graphical elements to them and that market (counted in pieces) is massive (total of over 3 billion mobile phone subscriptions globally, with around 1 billion mobile handsets sold annually)."In the opinion of CAPS All-Star Perkeo, this all adds up to: "a great value/growth opportunity."Let's run a few numbers and see if that's so. Right now, NVIDIA sells for just 8.4 times its trailing-12-month profits, and 8.5 times trailing free cash flow. Yet once it gets over the hump of reduced short-term earnings guidance, most analysts expect the company to proceed to post five years' worth of 15% earnings growth.

So in a nutshell, what we have here is the leader in graphics chips, selling for a 0.56 PEG based on short-term earnings jitters. Forgive the hyperbole, but that price is just insane. My advice: Take advantage of Mr. Market before he wises up.'

PS - don't shoot the messenger :)

Tonus said...

Hehe.

What I was referring to was that if we went by the assumptions that you often get on messageboards (which are frequented by people who don't know nearly as much as they'd like you to believe) NVIDIA had nothing to fear from AMD/ATI. And in fact they still do have the top-performing video card... but they are selling them for $400-450 now, instead of the $650-700 that they were selling for earlier.

Reality has a way of interfering with the expectations of the non-expert. I'm sure that the advent of the 4850/4870 line has those same people telling us that AMD/ATI are back in the driver's seat and that NVIDIA is on the way down. I wouldn't bet too much money on those predictions either.

Anonymous said...

For fun, I've been reading AMDZone the past few days.

General gist of the impression I am getting from most of the board members:

Question all Nehalem benchmark numbers EXCEPT the one Cinebench benchmark from Anandtech which shows Nehalem being no improvement over Penryn in single-core benchmarks. This benchmark is absolute FACT that Nehalem will not be faster than Penryn. Try disputing this and you will be labeled a viral marketer from Intel.

Take AMD's word as gospel that Shanghai will be 15% faster in IPC than Barcelona and will reach equal clocks as Nehalem. This is now FACT and cannot be disputed or you will be labeled a troll and banned.

10 out of 10 hardware review sites are in fact paid by Intel. Likely that 9 out of 10 directly run their servers straight from Intel Corporation.

A Nonny Moose said...

Tonus said...

"Reality has a way of interfering with the expectations of the non-expert."

Amen to that. Take my wife. Please :) She foresaw the stock market crashing last fall so she moved most of her 401K holdings to the bond fund, banking on the old adage 'when stocks move one way, bonds move the opposite'. Well her bond fund has farted around the same or a couple percent lower ever since then. She would have been better off in the T-bill fund. At least she was smart to get out of the stock fund - the S&P500 index has dropped around 20% since October - one of the indicia of a recession I think...

Anyway if AMD can move their GPUs to 40 nm quickly, NVidea might be in real trouble, rather than just embarrassed with 'cans of whoopass' turning out to be a whoopee-cushion instead.

I might go nutty (like Sparks :) and jump on a new system next spring - Nehalem of course - and will see who has the best GPU at that time. Hopefully NVidea will be back into their game plan by then and quit resting their butts on the bench. I ain't spending over $600 on a single GPU like I did last time with the 8800GTX...

Anonymous said...

"So in a nutshell, what we have here is the leader in graphics chips, selling for a 0.56 PEG based on short-term earnings jitters."

PEG or Price to Earnings ("P/E") ratio / growth rate is often the most misunderstood and misused metric for so called value investors, especially in the technology space

The rationale here is anything under 1.0 means that earnings are expected to grow at a rate faster than the price/earnings ratio. (Which implies the stock is undervalued)

Couple of problems with relying too heavily on just the PEG ratio:
- PEG goes down if stock price goes down (even if earnings is flat or falling at a slightly lower rate than the price). This could mean structural issues within the company even as the PEG gets better and the stock appears 'cheaper'. Do you really want to buy stock of a company seeing decreased earnings growth? this is the classic "value trap" scenario where the earnings growth #'s eventually get revised downward and all of a sudden the PEG is no longer so attractive.
- Earnings estimates change and could be very far off in the technology space. Do you really believe 5 year earnings estimates in the technology space, and specifically Nvidia, are to be relied on? This is why I don't put a lot of weight on PEG in the tech sector (it should be looked at, but should not be the primary factor). In industries where earnings are very stable and predictable, then PEG is a much better indicator.

hyc said...

Frankly I've found the discussions of economics, socialism vs capitalism, and electric vehicles here to be more enlightening than the CPU talk. Guess I have to agree with Tonus there.

Blogger Tonus said...

This is one reason that I regard a lot of the forum/blog talk as idle speculation and an entertaining diversion. One thing I like about Rob's site is that there is more talk about how real world metrics (markets, production capability, economic factors) affect the situation in the CPU business.

You can spend all day talking about how certain technological advances give company A a big advantage and act as if this is enough to override market realities, but in the end those realities cannot be wished away. Telling me that processor A is better because it has some fancy technology that really shines when you use synthetic_benchmark_B is all well and good, but don't be so surprised when processor A doesn't sell very well. It turns out that companies looking to upgrade computers aren't running synthetic_benchmark_A.

There is a lot more to the picture than just pure performance in one specific area that can be shown with a single benchmark. Just ask NVIDIA how well their $700 video card is selling these days. That's right, they're not selling them for $700 anymore...


Of course, there are benchmark trolls on both sides, but it feels to me like there are more on the Intel side. Why do people still post SuperPi results all the time, when everyone knows that it's utterly meaningless?

Anyway, it's obviously *fun* for a lot of these folks to endlessly rehash their thoughts on various technological features, otherwise why would they keep doing it...

A Nonny Moose said...

Hyc said...

"Of course, there are benchmark trolls on both sides, but it feels to me like there are more on the Intel side."

Probably because there are more Intel users than AMD users, would be my guess.

"Why do people still post SuperPi results all the time, when everyone knows that it's utterly meaningless?"

I don't see a lot of benchmark-touting spouting going on here at Robo's blog, or maybe I tend to skip over it as unless it's related to some technical feature such as number and mix of complex/simple decoders. The SuperPi discussion was mainly a result of the quick & dirty Nehalem test Anand did a while ago. When all you have is a grapefruit, you talk about grapefruit. I think most of us would like to see a thorough & comprehensive comparison of Nehalem to C2D as well as K10 or K10.5 when its released.

"Anyway, it's obviously *fun* for a lot of these folks to endlessly rehash their thoughts on various technological features, otherwise why would they keep doing it..."

Just human tribalism & aggression. Everybody tends to have favorite groups they identify with, to some extent. When your group does well, you get gratification and some (probably undeserved) sense of superiority & dominance. When your group doesn't do too well, then such predictable behavior as retreating to your 'safe zone' and consoling with your fellow groupies, denial, shoot-the-messenger & beating up on saps from the winning group who wander into your midst, occur. This pretty much describes AMDZone at the moment :).

Seriously, what most of us here find objectionable are fanbois such as Sharikook (PhD), Scientia & Abinstein who hold themselves out as some sort of all-knowing experts, when it was obvious years ago that they are merely amateur enthusiasts who don't work in the industry. Well maybe Sharikook worked for Intel, but not likely in process engineering or even CPU design. I myself am no expert and I don't work in the industry; however I make no pretense of doing so either. It's highly unlikely you'll ever see Scientia or Abinstein make a similar admission.

What I like about Robo's blog is that any FUD usually gets called out pretty quickly. At Sci's blog or at AMDZone, mostly Intel FUD gets called out - the AMD variety is permitted or even encouraged.

This gives me an idea - I thnk Robo should start his own website - Robozone :).

SPARKS said...

Hmmm, let me say first and foremost I am no bargain hunter when it comes to CPU’s. However, for me to spend, the timing must be right.

For example, my Crossfire 1900 XTX have held up quite well considering AMD’s failed 27XX series, and the partial recovery 38XX series. Factor in the performance difference on NVDA single slot solutions, the increase was in benchmarks. The performance improvement wasn’t WORTH the price of entry.

Speaking of NVDA single slot solutions, the 9800X2 has been trumped so badly by AMD; NVDA had to lower their prices dramatically.

(As you all know I absolutely adore INTC chipsets, NVDA is simply not an option.)

Crysis has been quite playable on medium settings, with the QX9770 a few of those kicked up to high, not a lot, but the increase was noticeable, nonetheless. The point is, I got a lot of mileage on the XTX’s at ~500 bucks a piece. TWO YEARS AGO! Further more, I got a lot of milage on the 975X INTC “Bad Axe”, a great board that took Q6600 to 3 Gig, no sweat.

The crazy money HAD to go to the X48/QX9770/1800 FSB/1800 DDR combo. Why? TIMING WAS RIGHT. This combo will be a force to be reckoned with, for a considerable length of time, even from an enthusiast lunatic fringe perspective, up to and including Nehalem. I believe the ‘older’ FSB deeply entrenched infrastructure, will coexist quite well with new X58 series.

At this juncture, two Crossfired 4870’s or one 4870 X2 will indeed double my performance. That’s what I’m looking for, that’s why I’m waiting. And at ~600 bucks, I consider it a bargain. It renders all previous NVDA cards (sans 280’s and 260’s in SLI) redundant, including the mighty $2000 2-8800 Ultras people paid so dearly for.

I may be nutty, quite true. But I’m not stupid. Further, I’ve built quite a foundation for 2 4870’s or 4870 X2. This IS the last wait until a Nehalem upgrade ~3Q, 4Q 2009. Again that’s a lot of mileage on my current setup. At less than 3 grand, sweet! Think Moore's Law.

“Of course, there are benchmark trolls on both sides, but it feels to me like there are more on the Intel side. Why do people still post SuperPi results all the time, when everyone knows that it's utterly meaningless?”

hyc- It seems you and I are in agreement once again. (I’m getting scared :) ) As for the benchmarks the QX9770/48X numbers are off the wall, BUT, they will be very meaningful when coupled with the new generation graphics cards I mentioned above. The combo’s performance will only help the card(s) performance. I’m thinking bandwidth here.

BTW: 10 sec SuperPi @ 4.06 :)

SPARKS

SPARKS said...

"Just human tribalism & aggression"

HOO YA!

SPARKS

Anonymous said...

Anyway, it's obviously *fun* for a lot of these folks to endlessly rehash their thoughts on various technological features, otherwise why would they keep doing it...

A theory? When you can't speak to actual technical performance, you can't speak to a credible roadmap, can't speak to financial results or the competitive environment... there is little left to talk about (at the moment) if you are an AMD fan. So you are left with a few options:

A) Talk about theory and academic questions like clock for clock comparisons (and choose to ignore the clock mismatches) or the minutia about scaling, cache size or whatever detail as opposed to end user performance. Just continue to rehash specific architectural details and who's copying who debates.
B) Live in denial and blame others for AMD issues. This has grown quite a bit as AMD has fallen on hard times.
- If it wasn't for Intel's monopoly abuse...
- If it wasn't for all of these corrupt and ignorant reviewers...
- If it wasn't for bad benchmarks like SuperPi and idiot fans... or misuse of compilers or...
C) Simply make stuff up to rally the masses and convince yourself things will turn around. On AMDZone, I think I saw someone saying a 45nm K10 should be able to get to 3.9GHz in 2009! I wonder what would have happened if someone made a claim about Intel being able to do this in 2009? On another distinguished website, there was a 'novel' interpretation of Asset Smart.

The reason things seem like a rehash of the same things over and over is because there is little else positive to talk about. You don't see discussions of financials or manufacturing or process technology or even future products. It's all become about bringing others down and why 'the man' is keeping us down or why things 'should' be better than they actually are.

I would think AMD fans would dig into the details on why the 4850/70 is so competitive with an Nvidia part that is much bigger physically and has many more transistors. I do not know much about graphics architecture but to me this would seem to be a positive and enlightening discussion as oppose to just yelling troll or fanboy or paid Intel viral marketer to anyone who presents an opposing point of view on CPU's.

pointer said...

A) Talk about theory and academic questions like clock for clock comparisons (and choose to ignore the clock mismatches) or the minutia about scaling, cache size or whatever detail as opposed to end user performance.

can't agree more with you on this, just look at the below quote from some zone:

Postby abinstein on Sat Aug 02, 2008 5:27 am from Amdzome said ...>

Phenom is superior to Penryn precisely because Phenom is a better quad-core design.


A better design that lose to the inferior one in most of the computing tasks? :)

Abinstein said...

Another paid Intel troll I see. Barcelona is a superior quad core CPU because it is a native design, and it is not limited by a front-side bus bottleneck architecture. This is why AMD has superior scaling and performance in real workloads like SpecFP. You Intel trolls will never understand that, since all you seem to do is run pointless benchmarks such as superpi over and over.

a nonny moose said...

Abinstein said...

"Barcelona is a superior quad core CPU because it is a native design, and it is not limited by a front-side bus bottleneck architecture. This is why AMD has superior scaling and performance in real workloads like SpecFP. You Intel trolls will never understand that"

Hmm, so you are the arbiter of "real" vs. "fake" workloads? Your credentials are in doubt, not only here but back on the ranch at AMDZone. Remember the thread about Dreamworks dumping AMD and going with Intel, and your continuing to argue with Cat Barf (who works in the industry) about rendering, long past the point where you were revealed as ignorant and foolish?

You know, it is possible to have meaningful discussions without all the ad hominem attacks. However you get your shorts in a wad real quick it seems - from AMDZone's poll on Shanghai performance, some of your more illuminating remarks:

"Intel is known to slap on the faces of its fanboys numerous times in the history. What amazes me is people's willingness to keep believing what the company's marketing department has to say."

- This after AMD promised "40% faster across a wide variety of workloads", then couldn't even deliver bugfree Barcelona until a year later.

"How can HyperThreading be good technology when most people who demand performance knowingly turn it off? They just turn it off."

This has been discussed both here and on AMDZone I believe. First of all, just how many people "who demand performance" do you actually know? Got any statistically signifant evidence here? Didn't think so. Second, you can't base SMT performance on the P4 with its long pipeline to that of Nehalem, at least until the latter is benched with SMT turned on and off. Third, SMT makes sense for the average desktop workload - keep the core busy with another thread (OS, background services, etc) when it would otherwise be idle. A 10 - 20% throughput increase is better than none, right?

"everyone should carefully looking at the stuff he buys from Intel - there's high chance that it's just crap. :p"

I guess that is your tongue-in-cheek humor showing through, but it probably is close to your actual opinion. You seem to think that because Barcie excels in 4-way that it is the superior choice for every application - desktop, gaming, render farms(?), etc. However 'most people who demand performance' for what they do (see above) disagree with you.

The only good thing about AMD going BK in the near future will be the disappearance of trolls like you & Scientia and Sharikook.

SPARKS said...

“Another paid Intel troll I see. Barcelona is a superior quad core CPU because it is a native design, and it is not limited by a front-side bus bottleneck architecture.”

How do you know this well you’ve undoubtedly never worked on an INTC chipset with an FSB over 800MHz? They are quite fast at 1600 Mhz in today’s top end machines. Never mind the 1800 MHz I have been living with for the past couple of months.

Frankly, for a man who says his interests are in physics, how do you support your claims without ----------like---------ummm------DATA?

Jeese, just tell everyone you’ve found a Higgs Boson, and you’ve completed the Standard Model in your basement.

Look, if Pointer is a paid troll, I must be one too! We are both very satisfied customers.

Last quarter we have +10 billion data points to show you why.
And an additional -1.2 Billion data points to show you why.
You’re a mathematician, you do the math.

Do you still think AMD has a superior product for the bulk of the worlds PC’s?

WHERE? I’d like to buy one.

“This is why AMD has superior scaling and performance in real workloads like SpecFP. You Intel trolls will never understand that, since all you seem to do is run pointless benchmarks such as superpi over and over.”

Your argument is contradictory and myopic. You claim superiority, and then site a pointless benchmark. How do you define real workloads, for which market?

SPARKS

pointer said...

Anonymous Abinstein said...

Another paid Intel troll I see. Barcelona is a superior quad core CPU because it is a native design, and it is not limited by a front-side bus bottleneck architecture. This is why AMD has superior scaling and performance in real workloads like SpecFP. You Intel trolls will never understand that, since all you seem to do is run pointless benchmarks such as superpi over and over.


you seems to like to attack the Superpi (and others) lately ... as your argument of AMD QC is better than Intel QC if not running 'useless' benhcmark like Superpi. You do realize some other test were run right? when you talk about phenom, you are talking about the desktop side, which in most 'real life' application, AMD's Phenom are crushed? are those games, media encoder/decoder, etc are fake applications??

2ndly, talk about the Barcelona ... I have never claim that its scalability is not good. It is still lose mostly to the Intel's QC at UP and DP though .. even the Specfp that you mention. It did beat Intel's at 4P. Again, this is one set of benchmarks. There are others too, which again Intel beat AMD's mostly at UP and DP, and AMD also beat Intel's mostly at 4P, but either way are not absolute.

Attacking Superpi as useless benchmarks, and never mention about what others were run ... a cheap tactic for a fanboy.

btw, i do have some quotes of yours in my blog, you might wanna visit it and have a look:)

Axel said...

pointer

"Phenom is superior to Penryn precisely because Phenom is a better quad-core design."

Though we saw comments like this in many hardware forums back in 2007 before the K10 launch, now that the truth is out there AMDZone is the only forum you'll still see comments like this coming up today. This is because most of the posters there like Abinstein and Scientia are in denial and will not accept reality. Virtually everyone else in the world readily acknowledges that compared to Kentsfield & Yorkfield, Phenom is a steaming pile of horse manure. It's generally much slower per clock (for both single threaded and multi-threaded apps), consumes much more power, and has a larger die to boot. AMD understand this too and have consequently already dropped the tray price below $200 for most of the X4 models.

Together with the massive strategic blunder of overpaying for ATI, the poor K10 design is the reason that AMD will likely soon be divesting themselves from CPU manufacturing. Since mid-2006, the progression of AMD towards this transformation has been correctly predicted ahead of time on this blog and in the accompanying comments, step-by-step, earnings CC by earnings CC. Since the very first K10 benchmarks started coming out in 2007, we knew that it wasn't going to catch Conroe/Kentsfield, and that AMD's ASPs would remain in the gutter for the forseeable future. We predicted that the continually depressed top line revenue would inevitably result in the closure of one of the two fabs and delayed re-tooling, as AMD could no longer afford to operate both fabs.

The "other" blog, sadly, has struck out on practically every prediction of significance since mid-2006. Market share gains, K10 performance, significance of DTX, potential of QuadFX, importance of ATI acquisition, you name it. This blog made contrary predictions on every one of those items and we were correct on EVERY one. I have to admit that I expected Hector to spill the beans on Asset Light / Smart a while ago. But in the end we will be proved right again, Asset Smart has nothing to do with the SMART fabrication efficiency project. Rivet has explained exactly what it is and we know that it involves outsourcing the manufacture of future AMD CPUs.

Seven straight quarters of losses at AMD is all the testament we need to know which blog / forum to pay attention to...

I hope everyone realizes why AMD actually have a prayer of breaking even on the operating line in Q3 or Q4 this year. It's because they are bringing their breakeven point down to $1.5 B per quarter. How are they managing this? Simple, they have cut way back on CAPEX and R&D... In other words, they have severely reduced investments into CPU manufacturing. 45-nm is the end of the road for AMD, future process nodes will almost certainly be outsourced.

Anonymous said...

You Intel trolls will never understand that, since all you seem to do is run pointless benchmarks such as superpi over and over.

I run one socket on my computer as does >90% of all computer buyers. If I was building supercomputers, I might care about scaling - but I'm not going to buy something because it thoeretically is a superior architecture when it has WORSE performance in the vast majority of applications.

Quite frankly if there were rubber bands holding Intel's quad together as long as it was good performance and reliable for what I use it for. To hear the same old Yahoos just yell troll and spout on about some NICHE application and superior arhcitecture, is pretty comical.

Hey Abinstein - tell us about your yield calculations again....dazzle us with your intellect... you remember - AMD had 50% better yield then Intel by virtue of your insightful analysis? You'd think with with differentials in yield like that the gross margin picture would be a bit different.

Anonymous said...

Couple of things (for some balance):

A 10 - 20% throughput increase is better than none, right?

Generally yes, but HT does mean more transistors, which means some increase in power consumption. Assuming this increase is not significant or the transistors couldn't have been 'spent' better elsewhere (more cache, etc, then yes 10-20% is nothing to sneeze at.

Together with the massive strategic blunder of overpaying for ATI,the poor K10 design is the reason that AMD will likely soon be divesting themselves from CPU manufacturing.

I don't think K10 design is the source of the issues. Yes they have has some execution issues but I don't think the design is fundamentally poor or flawed. AMD's 65nm process, and specifically the power, is more the issue - if K10 was able to clock up in the 2.8-3.0 regime last year/early this year (as many thought) it would have been competitive and allowed AMD some better ASP's and maybe stem (or take back) some of the server market erosion.

What really did in AMD (in addition to the ATI purchase) was the ASP erosion, which initiated with Ruiz's "30% market share at all cost" and was catalyzed by the introduction of core 2 (which pushed all AMD dual cores down the pricing chain).

K10, in my view, is a microcosm of AMD - a good idea with poor execution and timing.

InTheKnow said...

Abinestein said..,
Barcelona is a superior quad core CPU because it is a native design, and it is not limited by a front-side bus bottleneck architecture. This is why AMD has superior scaling and performance in real workloads like SpecFP.

No one in their right mind is saying that AMD isn't competitive or even faster on 4P systems. All the data says they are. But it is the result of AMD's HT that provides that boost, not the "magic" of native quad core.

The problem is most "real workloads" aren't being run by people who have 4P systems. They are being run on 2P systems and below where the data seems to favor Intel over AMD.

As to what the future holds, I would refer you to Cray, who should know something about high end supercomputing. They chose to go with Intel and Nehalem going forward. That fact doesn't bode well for AMD moving forward.

InTheKnow said...

Sparks said....
Frankly, for a man who says his interests are in physics, how do you support your claims without ----------like---------ummm------DATA?

What you have here is the difference between a theoretician and an engineer. On paper, everything about Barcelona looks great. So the theoretician says it must be better. The engineer looks a the data and says Barcelona doesn't measure up. At that point a good engineer will ask why?

I would propose the why is that AMD hit the thermal wall with this beast and it's killing them. Until they get the thermals under control, they won't be competitive. And for that they need HK/MG. I'll be surprised if you see that any time before Q4'10 (assuming AMD actually moves to 45nm in Q4'08).

SPARKS said...

“What you have here is the difference between a theoretician and an engineer.”

ITK- I respect your, and others’ expertise in the field. However, there is something I don’t understand, why doesn’t he with his ‘superior intellect’?

Despite AMD severe 7 consecutive quarterly losses, their obvious thermal problems, their unsuccessful attempts to correct the problem with subsequent steppings, he still holds fast that Barcelona is a superior product. Huh?!?

Cray and others, as we all know, are abandoning AMD. He knows this. I’m sure he is well aware of Penryns quantum lead in performance sans one very specific niche market. “And this too shall pass”, quite soon.

http://www.digitimes.com/mobos/a20080724PD205.html

http://www.hexus.net/content/item.php?item=14614

http://www.tomshardware.com/news/intel-nehalem-cpu-bloomfield,5968.html


Maybe, he’s shooting his last wad before the final nail is hammered in the Barcelona coffin come 4th quarter. It’s going to be a first round knockout at less than 300 bucks a pop. His last remaining benchmark in support of AMD superiority will be history. Then what? Hell, he’s out of a job.

Really, what was he trying to do by coming here posting such nonsense? Perhaps he’s lonely? No one has posted ‘over there’ in quite some time. Maybe, he needed to pick a fight with his former (banned/deleted/edited) adversaries.

After all, personally, I would be very unhappy if I couldn’t talk computer with SOMEONE, even if they didn’t agree with me on anything. The sad thing is he’s has angered and deleted, banned, whatever, so many people he had to come here to troll for an argument.

Honestly, it’s weak, and pretty pathetic.

SPARKS

Anonymous said...

Talking about trolls

One says "You Intel trolls will never understand that, since all you seem to do is run pointless benchmarks such as superpi over and over."

There is only one benchmark that matters! That is the proft loss column. AMD rules that one. The second is Market share. AMD rules in that one to.

The cum lackys can argue about benchmarks, FSB, scalability. But in the end he who makes money wins as he who makes money has money to build more factories, invest in technology and designs.

One company invests in the fundamental infrastructure required. The other looks to court rooms for bailouts from poor execution flawed business models.

Sure benchmarks matter, but when you are close other things matter more. These days AMD needs far more then just competitive or a bit better benchmarks to survive. Those that dont' see this are simply morons lost in their own rear

Anonymous said...

Lex, you really should just use your name... it'll make it easier to ignore you.

Anonymous said...

Is Lex really causing an itch that far up your A$$?

LOL

Anonymous said...

I'd say nothing would get Lex excited if AMD could show any signs of life. Or if any AMD cumm monkey had a proposal that might make AMD competitive. I'm not saying how AMD could get better, but how AMD could actually BE COMPETITIVE. That is two very different things, lose less money or really give INTEL a run for the money

InTheKnow said...

Sparks said...
His last remaining benchmark in support of AMD superiority will be history.

Sadly, this isn't true. We will have to live with this argument for another year until the 2nd Gen Nehalem comes out. Until then you will only have 2QP interconnects on each Nehalem package. So AMD will still hold an advantage in 4P systems.

a nonny moose said...

Sparks said...

"Look, if Pointer is a paid troll, I must be one too! We are both very satisfied customers."

Geez - where do I sign up for this paid 'Intel Troll' position? :P Here I've been posting for free, based on my own humble opinions, and now I find I can make some money off this??

Too bad AMD doesn't have any $$ to spend on its trolls. In fact, I hear they actually have to pay AMD in order to spread their BS, plus take a 2-week Koolaid Kourse of brainwashing, roadmap interpretation, selective benchmark-picking, etc etc. I believe Sharikook was so good at it, AMD gave him an honorary PhD. Unfortunately, Sci & Abi are still pursuing their post-grad AMDegrees.

I bet Ah-Ben-Stoopid won't be back - just some drive-by trolling I suppose...

Tonus said...

"Really, what was he trying to do by coming here posting such nonsense?"

Was that actually him? He has a Blogger account, but it looks like a message posted from an anonymous account (see the icon next to his name). It was probably someone else making a parody post.

a nonny moose said...

Anonymous said...

"Generally yes, but HT does mean more transistors, which means some increase in power consumption. Assuming this increase is not significant or the transistors couldn't have been 'spent' better elsewhere (more cache, etc, then yes 10-20% is nothing to sneeze at."

Presumably Intel with its large, power-efficient transistor budget on 45nm HKMG and its experience with P4 SMT, optimized Nehalem to get the most bang for the budget, under present and near-future workloads. AFAIK AMD does not possess any SMT design experience; hence AMD would start from scratch if they went with SMT at some time. Much easier to pooh-pooh it than try to design for it.

Just my opinion with no facts to back it up, but I think Intel's compiler experience also helps them figure out exactly what today's and near-future workloads are. In other words, it's not just a one-way street with optimizing the compiled code to run on Intel processors, but looking for trends in the code to help otpimize future processor designs.

IMO, where AMD goes a bit off is not knowing exactly how to optimize for the majority of today's and tomorrow's applications. Sure they can design a highly-scaling CPU for HPC, but as has been pointed out many times, exactly how big a market is that?

So Intel designs a mainstream processor - large and fast cache with superior prediction, 3 simple & 1 complex decoder mix that is a good match for today's software (largely single-threaded still) - and glues them together via the FSB which they ramp up the clock so that it can scale with bandwidth demands. As the industry (grudgingly) moves to multithreaded apps, Intel comes out with Nehalem that is a better fit.

In contrast, AMD doesn't have the $$ to design and produce CPUs as often as Intel's ticks. Hence they have to aim at some point in the not-so-near future. Plus their tocks are not too good either :)

SPARKS said...

“Until then you will only have 2QP interconnects on each Nehalem package. So AMD will still hold an advantage in 4P systems.”

ITK- Good for them, I split my fingers, raise my arm and say, “live long and prosper”.
More importantly, however, is the in the gut feeling I’m getting about this Nehalem launch.

"As the industry (grudgingly) moves to multithreaded apps, Intel comes out with Nehalem that is a better fit."

You industry boys keep chugging along, designing, shrinking, mixing chemicals, blasting, spraying, baking, and windup selling it to guys like me. To the average guy ‘2P’ would mean hitting the head at Yankee Stadium after a couple of beers. To the lunatic fringe, to which I am a full subscriber, this scaling thing is actually much ado about nothing.

I thought about the Skulltrail setup, initially, but even computer crazies get reality checks. It wouldn’t have given me that much of an improvement over the QX9770. True, I still lust for all that juice, but my virtual Lear Jet wouldn’t have landed any better at Long Island Macarthur Airport in MS FS X. But, if I worked for Rocketdyne building/designing these lovelies:

http://www.pw.utc.com/vgn-ext-templating/v/index.jsp?vgnextoid=579d34890cb06110VgnVCM1000004601000aRCRD

You can bet your life I would own one.

But, here’s the thing and there’s no getting around this. All the fanfare concerning this launch (not the Shuttle, Nehalem) is the memory controller. Yes, new instructions, new architecture, yada, yada, what does it mean to me? After all, QX97XX does a fairly good job of wiping the floor with everything else, including the vaunted IMC AMD copied from INTC’s Timna Series back in ’99.

http://www.cpu-world.com/CPUs/Timna/

Further, will the native 4 core Nehalem; give us some “magic” in the performance area, as opposed to the “rubber bands”? Will there be a clock for clock improvement over Penryn? I’m reading the preliminary benchmarks and frankly, I’m not that impressed-----yet.

Basically, we the fringe, don’t know much of the details surrounding this Nehalem/X58 launch, and it starting to eat my guts. As a share holder, I realize how important it is not to compromise your current inventory and partners with intimate details too early, thereby disturbing current sales.

What bothers me is we don’t yet know how or where this thing in going to fit into the market. Is this a desktop upgrade, or Xeon upgrade? 2P scaling, 4P scaling, who GAF!

Example, 2-QX9775's are lovely, but for the desktop, QX9770 is the weapon of choice.

Where will Nehalem fit in?
Will this thing rock??
Will it CLOCK!

The suspense is killing me.

SPARKS

pointer said...

Blogger Tonus said...

"Really, what was he trying to do by coming here posting such nonsense?"

Was that actually him? He has a Blogger account, but it looks like a message posted from an anonymous account (see the icon next to his name). It was probably someone else making a parody post.


yup, agree.

hyc said...

If AMD really were to fold up and disappear, what makes you think the trolls would just go away? They'd just disperse to other products. Maybe a year from now you'll be fending off the "Penryn Rulz Nehalem" trolls...

hyc said...

On a completely different topic... Saw a comment about how improving overclocking for the Phenom may make it more difficult to underclock.

What determines the slowest workable speed for a processor? E.g., if I have a CPU spec'd at 3GHz, using a x15 multiplier on a 200MHz reference clock, why wouldn't the CPU still run correctly on an x1 multiplier? (Aside from the fact that there's currently no way to select such a low multiplier...) Would running at such a slow speed actually yield a big power savings over say, 1GHz, or 2GHz? We used to get a lot of useful work done on 33MHz 486s. Even my Pentium-M Dothan only downclocks to 600MHz, why doesn't it go any slower?

Anonymous said...

Will there be a clock for clock improvement over Penryn?

I suspect the improvement will be VERY application specific (more so than previous architecture transitions).

Quite frankly, I'm not one of the idiots whose going to separate out the SMT gains as a trick or consider it extra cores. If this yields a gain, then it it really no different that adding more cache or another FP unit or whatever - I find it insane that people want to separate it out. It may be application specific but it should be considered part of the overall aggregate improvement (Or else why stop there let's filter out impact of L3? Or shared cache? or new instructions? Or....)

If I had to speculate I would say 10-20% across most applications with large jumps on specific applications (and large jumps in server and MP specific situations). This is at same clock - it has yet to be seen how clockable this architecture is.

If it can scale to higher clocks better, than I'm not going to simply ignore the higher clocks and only consider clock for clock comparisons - you should compare best product on Penryn vs best product on Nehalem, or products at similar price points, or max OC's if OC'ing is your thing.

All this clock for clock, turn off SMT crap for a "true" comparison is a nice academic study, but when you plug these chips into your computer most of us don't turn off new architectural features or underclock the chip.

BTW - I was trying to compare the 0-60 speed of a Corvette to a Tercel - does anyone know how to turn off some of the cylinders in the Corvette engine and increase the drag, so I can get an "accurate" comparison. The Corvette may be faster but it's important to know how much of that is due to the trick Chevy does by using an engine with extra cylinders as who knows maybe Toyota will beef up the Tercel engine in the future! I was also trying to measure braking power - can I just disable some of the 'tricks' some car manufacturers do
to improve braking power?

I'm also doing a study on frequency response in subwoofers - there is this one subwoofer which goes to lower frequencies but it has a bigger driver, different amp, and a different shape. Can I just dremel or mask off part of the driver and adapt the shape to get a true driver for driver, amp for amp, volume for volume comparison and filter out the 'tricks' some of the subwoofer manufacturers are using? This really is the right way to compare, correct?

pointer said...

Blogger hyc said...

...

What determines the slowest workable speed for a processor? E.g., if I have a CPU spec'd at 3GHz, using a x15 multiplier on a 200MHz reference clock, why wouldn't the CPU still run correctly on an x1 multiplier? (Aside from the fact that there's currently no way to select such a low multiplier...) Would running at such a slow speed actually yield a big power savings over say, 1GHz, or 2GHz? We used to get a lot of useful work done on 33MHz 486s. Even my Pentium-M Dothan only downclocks to 600MHz, why doesn't it go any slower?


Due to unknown reason to me , Intel lowest multiplier has been always(?) 6. On your dothan, I believe its FSB is 100MHz, and hence lowest you get is 600MHz.

I believe selecting any other lower values (assuming you can), is not guarantee to work.

hyc said...

Right, the lowest multiplier on my Dothan is 6.

My question was - is the lower bound set arbitrarily, or is there a physical reason why the CPU would not function at an even slower speed? Or, is it an issue of diminishing returns, where going slower just doesn't yield any further power savings?

Ho Ho said...

"Didn't you get "perma-banned" before?? What exactly is a "perma-ban" if not permanent?? :)"


Yes, I've been banned from there before a few times. This time I noticed the word "permanently", I don't think it was there before. I usually to go there once in a couple of months and try to use their logic against them. So far I've been banned for doing it in a week or two :)

As for low clock speeds, I've underclocked my 3.2GHz P4 to 400MHz in Linux. The CPU did just fine but some capacitors on the motherboard started to make funky noises so I clocked it back up.

Ho Ho said...

Just great, now Scientia is making up stuff on amdzone that I've never said. It's funny how he keeps on responding me there but doesn't say a word about the posts I've done on his blog that were talking about his mistakes.

Tonus said...

anon: "Quite frankly, I'm not one of the idiots whose going to separate out the SMT gains as a trick or consider it extra cores. If this yields a gain, then it it really no different that adding more cache or another FP unit or whatever - I find it insane that people want to separate it out."

I think it is a case of working backwards. Often if you are comparing different CPUs or architectures, you can look at benchmark scores and then try to figure out how the various features affect the scores. It can help you to determine what applications the CPUs might be better for (or worse for).

This looks like a case of looking at the various CPU features and then determining whether or not they provide an advantage, unfair or not, and how that makes the CPU --as a whole-- better or worse. It can lead to an interesting discussion on comparison of features and how those features may develop over time, but there is not nearly as much practical application for the user looking to make a buying decision.

A Nonny Moose said...

Hoho said...

Yes, I've been banned from there before a few times. This time I noticed the word "permanently", I don't think it was there before. I usually to go there once in a couple of months and try to use their logic against them. So far I've been banned for doing it in a week or two :)

I usually visit a couple times a week, just for the amusement factor - as when Sci & Abi accuse each other of lacking reading or comprehension skills. However I haven't noticed any flaming or name-calling from you or the other non-Koolaid-guzzing posters there. M-m-m-marq called somebody "a stupid git" but since he's pro-AMD he obviously gets a pass from the admins. Besides, just trying to read his posts likely induces headaches if not brain cancer, so more power to him posting on the Zone :).

hyc said...

Yeah, I wish I could put mmarq in a kill-file like good ol' usenet. I can feel my IQ dropping whenever I try to parse one of his posts...

SPARKS said...

Here’s a bit of advice. Don’t find yourself under this piano, or god forbid, try to catch it.

http://finance.yahoo.com/q?s=NVDA

What is incredible is Jin Junk Sung didn’t see this coming when he was screaming ‘can of whoop-ass’ at NVDA press conferences. In fact, he obviously didn’t have a clue given he acted so arrogantly immediately before the train wreck. He has been totally blindsided, this is clear.

Charlie at INQ has what seems to be a pretty accurate report of the recent chain of events and some future predictions concerning NVDA position in the market. In the interest of objectivity, he absolutely loathes NVDA. Still, I think he’s got a handle on this one,------this time, and time will tell. Corporate arrogance run amuck is a very scary thing.

AMD
Yahoo
NVDA

http://www.theinquirer.net/gb/inquirer/news/2008/08/02/nvidia-chipsets-dead

SPARKS

SPARKS said...

Here ya go fella’s, you won’t need to know much a graphics architecture. INTC is going to use x86 core logic to run the world!

http://news.cnet.com/8301-13924_3-10005391-64.html?part=rss&subj=news&tag=2547-1_3-0-5

Imagine, my new computer in 2010, INTC CPU, INTC CHIPSET, AND INTC GRAPHICS. CALL ME A FANBOY! F—K IT, so be it!

I can’t wait.

HOO YA!

SPARKS

SPARKS said...

"Oooh this is dangerous"

http://anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3367

SPARKS

Giant said...

Oh would you look at that? Over at the illustrious AMDZONE Scientia is making me laugh even more.

He is insisting that no Wolfdale or Yorkfield is capable of running at past 4GHz 24/7 stable with air cooling. What a load of nonsense.

What makes it so damned hilarious that on his own blog, he made the following claim:

This is a strange claim indeed because everyone else seems to be able to clock these chips to 3.2Ghz with no trouble, and I've seen claims as high as 3.6Ghz

This is on air cooling and he is making reference to a 3.6GHz Phenom. Where is his proof, four copies of Prime95 as he demands as proof from others on AMDZONE. Hasn't Scientia heard of using a single program to load all four cores? It's much easier.

He claims that no-one can do it, that a 4GHz Yorkfield on air just can't be done. Hah. With the 790i board (best motherboard I've ever used, TBH) I can easily do 4.05GHz on my QX9650. That's a 9X multiplier with 450 (1800) on the FSB and 1.38V in BIOS. That was in the Silverstone TJ09 case, with no less than seven 120mm Silverstone case fans, and two S-Flex fans on a TRUE 120 for CPU cooling. Such a setup was stable running ORTHOS for 15 hours across all four cores before I stopped it; I don't see any reason why it couldn't have continued on for longer than that. Now I'm running the Koolance water cooling on the QX9650 and 4.275GHz is my 24/7 overclock (9.5 x 450).

The E8400 is even easier to overclock to 4GHz. With the P5B Deluxe board (this is a 2yr old P965 board) the same 4.05GHz (450x9) is easy to achieve. I put the TRUE 120 on the E8400 after I water cooled the QX, so it hums along all day long and passes ORTHOS easily. The E8400 only takes 1.28V to do 4GHz, it seems I got one of the better Wolfdales. :)

Of course, these are all C stepping 45nm CPUs, not the newer E stepping CPUs that will allow for even higher clockspeeds.

SPARKS, have you decided on your new GPUs yet? Far too many choices. They've got the Radeon 4870s and the GTX 260s for well under $300 now, and the GTX 280 for under $450 now. I'm still running the 8800 GT SLI setup for now, but I'll upgrade soon enough. The 4870 dual GPU card and the die shrink GTX 280 card are due this month, I think I'll wait for them. :-)

-GIANT

Anonymous said...

Please forgive me for asking a stupid question, but how is Prime95 on all cores the 'standard' for overclocking goodness?

I realize this puts a huge stress on all cores, but if you can run 7x24 for weeks on end, wouldn't that be a more meaningful measure of a stable overclock? (I don't care if that means stability or not for other users who may use different applications, for me it means stable!) I frankly could care less whether I can run Prime95 on all my cores simultaneously because I don't think Prime95 simulates gaming or encoding or whatever normal application people use computers for and it's not something I tend to multitask with :)

So who anointed Scientia the arbiter of what is and is not a 'true' overclock? Am I missing the link to his hardware review site where he does this on a regular basis and has vast experience in this area?

Frankly I tend to put a little more stock into X-Bit labs as opposed to some yahoo with a blog... but that's just me and my craziness! I'm not saying they or any other review site is perfect or doesn't make mistakes, but it is rather amusing how Scientia views himself as the overclock police (and judge and jury). The value of his knowledge and $3 will buy me a small cup of coffee at Starbucks (maybe).

Ho Ho said...

"So who anointed Scientia the arbiter of what is and is not a 'true' overclock?"

He did it all by himself.

SPARKS said...

Giant-
This a copy from an above comment:

"At this juncture, two Crossfired 4870’s or one 4870 X2 will indeed double my performance. That’s what I’m looking for, that’s why I’m waiting. And at ~600 bucks, I consider it a bargain. It renders all previous NVDA cards (sans 280’s and 260’s in SLI) redundant, including the mighty $2000 2-8800 Ultras people paid so dearly for.

I may be nutty, quite true. But I’m not stupid. Further, I’ve built quite a foundation for 2 4870’s or 4870 X2. This IS the last wait until a Nehalem upgrade ~3Q, 4Q 2009."

SPARKS


PS-Dementia is trying to bait me to come back. I refuse to give to give him the satisfaction.

He's a total write off, as far as I'm conserned. F--K him.

Tonus said...

Meh, when it comes to overclocking, isn't everyone an authority? My "ultimate test" for an overclocked CPU used to be installing Windows NT Workstation 4.0. It would crash on me even when everything else would run without a hitch. I guess I could say that a CPU isn't really overclocked until you can install WinNT Workstation 4.0.

Overclocking and stability is like anything else in computing-- what matters is what you plan to do with the component(s) in question. Running game demos does me no good if my intention is to run 3D applications. And in some cases, if you're not running a specific 3D application, it doesn't do me much good either. You can't judge components based on some universal standard, because it may not apply to you or your needs.

I'm not surprised that Futuremark's benchmark programs favor Intel CPUs, though. Their benchmark programs were always pretty useless, IMO. You could run it all day on different platforms, and your results would often be completely at odds with the results of actual game benchmarks. Complete crap.

hyc said...

Well... When you buy a CPU and run it at its rated speed, you expect to be able to run any program(s) on it without any instability.

I think the point is that an overclock isn't "stable" unless it can also run any program(s) on it without crashing or producing incorrect results.

If it's only stable enough to boot and get SuperPi running, but no other apps work, then it's pointless, let alone stable...

Anonymous said...

I think the point is that an overclock isn't "stable" unless it can also run any program(s) on it without crashing or producing incorrect results.

If it's only stable enough to boot and get SuperPi running, but no other apps work, then it's pointless, let alone stable...


And how is prime 95 across all cores the arbiter of that? You keep changing the arguments around... where did anyone state superpi should be used to measure a stable overclock, or simply the ability to boot? You keep trying to spin/reinterpret these arguments and it is quite honestly getting frustrating.

So I'll go back to my original point if I'm OC'ing and haven't had a single issue for weeks, does it really matter (to me) if it can or can't run Prime95? Should I suddenly lower my OC if I find out Prime95 crashes (even though I don't see any other effect on anything else I do)? My point (and I think other's as well) is that there is no single OC benchmark goodness barometer and Dementia is just trying to invent one to take Intel, or various reviewers he has some personal vendetta against, down a few pegs.

What is lost in his flaming about x-bit labs is that the overclock in question was in the neighborhood of 4.5 - 4.6GHz! And he's nitpicking about small FTT vs large FTT's - even if that were an issue (and there is no data either way to suggest it is) is that really going to bring the overclock down 600MHz?!? He is fighting the typical guerrilla war - he knows he won't flat out win the argument so he attacks small points hoping this keeps people from remembering the bigger picture?

Quite honestly Dementia is starting to remind me of a certain Captain chasing around a certain whale.

SPARKS said...

“And how is prime 95 across all cores the arbiter of that? You keep changing the arguments around... where did anyone state superpi should be used to measure a stable overclock, or simply the ability to boot?”

Quite right.

You made some poignant analogies in a previous post. Subwoofers as related to design parameters, Xmax, Fs, etc, please, allow mw to make one.

Lets take, say, a Bugatti Venron, for example. Would you say the engine is a dog or it shouldn’t be rated at 600 or 700 HP because the engine couldn’t hold up running a week straight at top RPM and HP? No the Venron’s motor was designed for dynamic loads in real world operating conditions. Taking my QX9770 to full load in Prime95 for an hour, a week, whatever, is like putting a Venron on a chassis dyno and running in at full bore 24/7 for a month. Of course the million dollar son-of-bitch is going to blowup. Does this mean the car sucks? Nonsense.

Look, this is fact, I own a QX9770, and have been running in for months now at 4.06 GHz on air, with nary a glitch, on every game and every program sans one, Prime95. I clocked it down to 3.85, and it would run Prime95 all friggen day. It simply overwhelmed the Zalman CPS9700LED’s ability to cool the chip. There’s no INTC conspiracy theory here. In real world usage the 4 cores are NEVER, HELLO!!!, loaded 100%! It’s running so well DYNAMICALLY, I DON’T NEED WATER!

They can speculate and criticize, theorize, and compromise all they want. They can speculate about the HOTWOMEN, THE HOT CARS, THE HOT BOATS, AND THE HOT CHIPS, until their nuts turn blue. To these guys, who don’t have a pot to piss in or a window to throw it out of, they will never know. They just pick on some small detail and say sour grapes; it sucks, and they’ll derive some SMALL inner satisfaction!

It a losers way out, if they can’t have it, afford it, or have a personal grudge, they’ll knock it. They’ll criticize the hot women, and then they’ll go home and rub one out just thinking about her, then tell you later what a bitch she is.

I’ve seen it my entire life.

QX9770, the best desktop chip in the world, @ 4.06 on air 24/7, the rest is just plain horseshit.

SPARKS

pointer said...

Blogger hyc said...

Well... When you buy a CPU and run it at its rated speed, you expect to be able to run any program(s) on it without any instability.

I think the point is that an overclock isn't "stable" unless it can also run any program(s) on it without crashing or producing incorrect results.


This is true from intel's standpoint, but not from the user's stand point. As long as whatever a user do with his system doesn't crash due to OC, it is perfectly fine for the particular users. Intel on the other hand, cannot sell such high freq rated part of it could crash or the parameters exceed certain spec such as TDP, etc.

If it's only stable enough to boot and get SuperPi running, but no other apps work, then it's pointless, let alone stable...

This is the exact problem with people in AMDZone. Attacking SuperPi, misleading people as if this is the only benchmark run that intel has a lead ...

and in this context, misleading people as if those people do the OC claim stability with the superPi only.

This used to be Scientia commenting tactic which i described in some posting long ago. Mix an unrelated but correct statement, to imply a wrong statement true.

Anonymous said...

A bit offtopic, but management that makes AMD's look like shining stars.

http://www.cnbc.com/id/26046479/site/14081545?__source=yahoo|headline|quote|text|&par=yahoo

Part of me wishes I still had a few shares of that stock (so I could launch a variety of lawsuits). Unfortunately (?), I was smart enough to dump it a long time ago as it is generally unwise to hold shares in a company that has such poor management.

The bottom line - Yahoo surprised everyone when they announced there was only 15% opposed to Jerry Yang in the recent shareholder vote, a number which confused many analysts....

Well, apparently there were some "tabulation" errors and a few, ummm a minor....200 MILLION (out of 1.4Bil) votes were not counted and the disapproval % went from 15% to just under 34% (which is rather high given the support Icahn gave at the last minute in exchange for seats on the board).

A mistake, some in the Microsoft camp tell me, that speaks directly to Yahoo's credibility and why Microsoft ultimately walked from the deal. One source close to Microsoft tells me the company never trusted the numbers it was getting from Yahoo. That was a fact brought home by Yahoo's refusal to acknowledge Microsoft's higher offer, that would take the bid ultimately to $47.5 billion.

It is management like this that makes you scratch your head. Here's hoping the true details are uncovered and someone is actually serving some hard time or at least time at one of the country club federal prisons (yeah there's no chance in hell, but I can dream can't I?).

The stock is now $19.82, and I'm guessing Yahoo is sticking by the belief that the $33 (or more) that MS offered is "undervalued". This my friends, rivals the arrogance of Ruiz.

Anonymous said...

My bad it was 335Mil (and change) that wasn't counted.

* 335Mil votes not counted
* Yang's against vote went up by ~312Mil(based on diappproval % before and after)

So in what clearly must just be a remarkable coincidence the vote that wasn't counted just happened to be going 93.3% against Yang. Yeah, I would trust these "yahoos" (pun intended) to run my company.

SPARKS said...

“This my friends, rivals the arrogance of Ruiz.”

I must apologize, but I will have to take exception to this, despite that human compiler mind of yours.

NOTHING, rivals the arrogance of Wrector Ruinz.

SPARKS

SPARKS said...

Naysayer and doubters alike, here's what an obscenely cooled QX9770 by the extremely talented enthusiast lunatic fringe can do.

http://www.nordichardware.com/Articles/?skrivelse=538

I'm very comfortable at 4 Gig, thank you.

SPARKS

Khorgano said...

Scientia's blog has become the poster child of logical fallacy. (Sharikou forgoes all pretense of being a serious blog). Just read any of the blog entries and comments over the last 2 years. When addressing any topic that he deems Pro-Intel or Anti-AMD he begins with:

1. Create Straw Man or Red Herring.
2. When people call him out on his misinformation, resort to ad-hominems.
3. Move onto new topic, rinse/repeat.

pointer said...

Deneb looks promising.

http://www.hardspell.com/doc/hard/79579.htm

AMD might have chance to compete with Intel's YFD/NHM after all.

hyc said...

Blogger pointer said...

I think the point is that an overclock isn't "stable" unless it can also run any program(s) on it without crashing or producing incorrect results.

This is true from intel's standpoint, but not from the user's stand point. As long as whatever a user do with his system doesn't crash due to OC, it is perfectly fine for the particular users. Intel on the other hand, cannot sell such high freq rated part of it could crash or the parameters exceed certain spec such as TDP, etc.

Right, fair enough.


If it's only stable enough to boot and get SuperPi running, but no other apps work, then it's pointless, let alone stable...

This is the exact problem with people in AMDZone. Attacking SuperPi, misleading people as if this is the only benchmark run that intel has a lead ...

and in this context, misleading people as if those people do the OC claim stability with the superPi only.

I was just using SuperPi as an example to illustrate the position, don't take it too literally. I understand your point; I've seen plenty of posts in overclocking forums where SuperPi is provided as *one of* the results. Occasionally you'll see a post like "I got it to XX speed and SuperPi, but nothing else worked so I had to back it down" so I think most people know it's not a sufficient test in itself.

I recall that Prime95 actually heats a processor more than Orthos, which is probably why Scientia harps on Prime95 so much. I'm a little skeptical of its relevance since in my own day-to-day work, I run pretty much 100% integer code. I might be curious to see how well Fractint runs as a stability test instead, using pure integer/fixed-point arithmetic to heat the cores.

Anonymous said...

I've seen plenty of posts in overclocking forums where SuperPi is provided as *one of* the results.

Not to belabor the point, but there is a HUGE difference between someone using SuperPi to show the result of an overclock and the person using SuperPi to demonstrate stability (which is what you insinuated)

Quite frankly a portion of AMD fans have got hung up on this and need to stop whining about the efficiency and usefulness of this. SuperPi is an arbitrary benchmark, and is a quick and dirty of comparing OC's. Is it ideal? Of course not. Is it a useful program? Of course not. Is it something I would use? No, but I'm not going to complain about others wanting to use it.

Do people claim stability based solely on running SuperPi? I'd be curious to see a few links where this is explicitly stated. As you have done in the past, you are starting to evolve your statements over the course of a few posts.

So HYC, if you have problems with these other forums - post your specific concerns there. If you've seen similar posts here, point them out - but stop with the broad generalizations.

I've seen a forum of AMD fans where they compared Intel to Nazi's... should I go post over at AMDzone and ask why are some AMD fans doing that? Of course not, as the folks at AMDZone were not the ones saying it, and are not responsible or representative of all fans.

SPARKS said...

Uhh, Doc, there are 10 comments missing. Was it something I said?

SPARKS

Ho Ho said...

Wasnt' scientia supposed to be a programmer? If he is then he must be pretty bad one or retired for more than 15 years. Pretty much every programmer who knows anything knows that Java isn't interpreted language. He also seems to lack a lot of knowledge about JIT compiling and real-world experience on the field.

Abinstein seems to be worried that Larrabee can run OS as fast as 1GHz Atom. Well, I've yet to see OS taking significant part of my CPU resources. It's the GUI rendering and user applications that does it. Also he seems to think that if you put OS on Larrabee it will be downgraded to single-core Pentium because old pentiums didn't support any of the Larrabee features. I wonder why is it so hard for him to understand that OS'es aren't programmed for specific CPUs and they use whatever features they expose like multiple cores and HT.

Tonus said...

anon: "What is lost in his flaming about x-bit labs is that the overclock in question was in the neighborhood of 4.5 - 4.6GHz!"

I saw the review he linked to. They also use something called the OCCT stability test, which is designed specifically to test the stability of overclocked CPUs. No idea how effective the application is, though. They also did state that they used small FFTs mode for the Prime95 test, for what it is worth.

He mentions running OCCT on the CPU when it was OC'ed to 4570MHz and at 4275MHz when testing power draw. The E8600 they got was pretty impressive. It clocked from 3333MHz to 3900MHz at default voltage (1.2), 4450MHz at 1.4v, and 4570MHz at 1.5v. Looks like they got this while using a slightly fancy HSF, but it seems as if they did it with air cooling.

Apparently Intel will be releasing the E8400 and E8500 in E0 stepping versions soon as well (if they haven't already). That might make for some interesting overclocks.

Roborat, Ph.D said...

test....

Roborat, Ph.D said...

test2

Roborat, Ph.D said...

test3

Roborat, Ph.D said...

test4

Roborat, Ph.D said...

test5

Roborat, Ph.D said...

test7

Roborat, Ph.D said...

test8

«Oldest ‹Older   1 – 200 of 204   Newer› Newest»