3.03.2009

"At heart, we're a reverse engineering design company" - AMD

"Advanced Micro Devices Inc closed a deal to spin off its manufacturing operations on Monday, and said it expects the new company to assume responsibility for paying off about $1.1 billion of debt.
The plants which make AMD's chips are now part of a $5 billion joint venture with Advanced Technology Investment Co, of Abu Dhabi, temporarily called The Foundry Co."
- Reuters

In a very creative way, AMD has ridden itself of its crippling debt and the massive burden of capital investment going forward. While AMD may sound as if this strategic move brings them closer to their core expertise, it is without a doubt that this back-to-the-corner decision was the only way for AMD to remain viable. This new lease allows AMD, maybe for a few more product lifecycles to continue and remain as the only challenger to Intel.

At the bleeding edge of semiconductor technology, it has yet to be seen whether a fabless company can challenge one with a foundry. In the not so bleeding edge such as memory products companies with their own foundry like Samsung are dominating over the rest of the industry but competition remains vibrant. But in the x86 space where process leadership creates cost and performance advantages, history isn't kind to fabless companies. Starting this week, AMD is effectively what Transmeta was back in 2000. The difference is Transmeta had a lot of hype going for them and probably with a more compelling product offering in the mobile space.

966 comments:

«Oldest   ‹Older   601 – 800 of 966   Newer›   Newest»
SPARKS said...

"Ortho why don't you just post the schedule."

You must be joking. There are hints all over the web that 32nM will be ready before the 'official' 1Q 2010 time table. In fact, many are saying 4Q 2009 is more likely. Ortho is well in line with the company's publicly released schedules you can get anywhere on the web.

Besides, you've probably got guy's staked out at the fabs counting the number of trucks LEAVING DID.

Further, INTC has been sandbagging release schedules on regular basis. They did it with Penryn. My guess, they'll do it here too.

Hey, if it weren't for a WEB 'leak' regarding a certain chip with a 1600 MHz FSB, I wouldn't have waited to spend $1500 on the FABULOUS QX9770! Thank God they did. Thank God I did.

In fact, if 32nM by Christmas, and double the performance over my lovely beasty, I'm in.

SPARKS

Orthogonal said...

Ortho why don't you just post the schedule.

BTW make sure you copy carefully, don't think to much just copy!


I don't know if you're trying to be snarky or just an ass, but if you have even casually followed the news you would have known Intel's 32nm high level ramp schedule which has been publicly available for the last 6 months.

http://www.anandtech.com/cpuchipsets/
intel/showdoc.aspx?i=3513

Tonus said...

I think that a lot of the "catching up" talk is because it's really better to look forward when you are getting your ass beat. Sure, Intel is feeding you your own liver right now, but dammit, in just six months you'll be on even footing and then it's a whole new ballgame!!!

That's one reason why I prefer to look at what has happened, as opposed to what may happen*. You can talk about process leads and new technologies all day long, or tell me how spinning off half of your business is a good thing... but it's the end result that counts.

*The other reason, of course, is that I have no background and little understanding of process technology. However, I would note that a lot of people in my situation don't let that stop them from making big pronouncements about what will happen in the future. It does make for some entertaining reading, at least!

Tonus said...

AMD loses market share. The good news for AMD is that unit sales were up, but Intel picked up 0.5% share and AMD lost 2.2% share (ouch). And it's getting harder to use the 'antitrust' excuse, since these totals come after many of the investigations into anti-competitive behavior in various markets around the world.

Anonymous said...

I believe I mentioned that the Q1 #'s were inventory related...

(excuse me while I rest for a while as my back hurts from patting it so hard!)

Anyway... Intel said inventories were correcting starting to correct in Q4 and Q1 and while many just assumed this had equivalent impact on both companies (and thus AMD fans rejoiced PII was having an impact after the Q1 #'s were announced) - there was a subtle point missed in all of this.

Intel mentioned that there was better visibility and they sor more of an impact in desktop, while mobile inventories lagged as these inventories tend to have a longer cycle in the channel. So, it would be expected desktop would correct quicker and have a greater impact on the earlier (Q1) #'s, while notebook recovery would have more of an impact on the later #'s (Q2). It's not like there was a black and white shift based on the quarters - but the inventory corrections in notebook and desktop definitely had a different time component in them.

When you look at relative AMD/Intel market shares in desktop and mobile, Intel has a higher MS in notebook... so AMD saw a bit more (relatively speaking) of the recovery/flattening out of demand in Q1 while Intel saw more of an impact in Q2.

The other clue (in hindsight) was the relative forecasts after the Q1 earnings - I think AMD realized they saw most of their inventory correction impact in Q1 (and thus guided flat for Q2) while Intel saw the notebook inventories still to recover (and guided up in Q2).

So basically neither the Q1 or Q2 share gains or losses were "real" - it was a matter of two different segments having their inventory correct at different times. Now that they have, I think you could expect future market share gains/losses to be more reflective of "real" gains or losses.

Bottom line - If you want to see if Intel or AMD is gaining or losing share you will have to wait for the Q3 and Q4 #'s.

It's similar to the pump and dump AMD did (was that Q4'07? maybe Q4'06?) where they blew out a bunch of sales in Q4 which made the #'s look good (and probably helped on the finances at the time - they were either trying to secure the Morgan Stanley loan or do the convertible notes around then - I forget). They saw a huge hit in the following quarter as a result, warned right before the earnings release and had the stock get hammered, but they had secured the financing and you were left with a situation where you basically just had to average the 2 quarters to get the real story.

Anonymous said...

^ forgive the typos and bad editing ... man that was poor!

Anonymous said...

Can't wait to see AMD's 32nm HighK.

Anonymous said...

AMD doesn't do process technology anymore, you may have heard about it in the news? :)

Tonus said...

So Westmere will be a 10 core CPU?

Oh, sorry, didn't mean to interrupt the 32nm HKMG love fest. :)

Orthogonal said...

Tonus,

That link is referring to Westmere-EX, which is the MP server processor. It will be the Nehalem-EX succssor (AKA Beckton 8C\16T). I'm not sure on the actual core count but I suppose it's plausible.

Anonymous said...

That will be some darn hot dorito chip. What in the world could cool that?

Maybe they'll have to ask GlobalFoundries to make it for them. :)

Anonymous said...

GF is nothing but a shell game. AMD and hector were very clever in getting some dumb fuck arabs to pop billions into a new holding company to make AMDs balance sheet look neat and tidy.

It doesn't change the fact AMD process is no good. That they are behind. That their market share is too little to utilize the arab cash. Nothing really changes, except there will be more leading edge capacity possibly for some other company. The issue is that beside CPUs nobody really needs that leading edge capacity. So the suckers are the dumb ass arabs.

Tonus said...

"Maybe they'll have to ask GlobalFoundries to make it for them. :)"

But wouldn't it be worse for thermals if they use a less advanced process?

Tonus said...

PS, Am I the only one getting a Network Solutions placeholder page for AMDZone?

pointer said...

to Tonus,

No, you are not. I missed my jokes of the day.

Anonymous said...

"But wouldn't it be worse for thermals if they use a less advanced process?"

LOL. AMD's or GlobalFoundries 45nm process is more advanced than Intel's 45nm counterpart. Don't believe me?

Intel needs High-K interconnects to achieve the same level of performance that AMD/GlobalFoundries offers with their 45nm process. Also, (as someone else pointed out), AMD's (or globalfoundries) 45nm process runs cooler than intels. I really don't know if this is the reason of superb overclocking without coldbugs, but it's a hint.

Also, I want to make it clear that wWhen it comes to process technology, nothing beats the IBM/GlobalFoundires alliance. It's only a matter of time when we will see intel opting for an asset light strategy in the long term, because I'm darn sure those cheap-crap atoms can't/won't fund enough R&D bucks for intel's fabs, and each new node/process is double the price of the previous one.

All I want is a serious and unbiased debate on this, not the typical "AMD sucks" fanboy crap.

Anonymous said...

Intel needs High-K interconnects to achieve the same level of performance that AMD/GlobalFoundries offers with their 45nm process. Also, (as someone else pointed out), AMD's (or globalfoundries) 45nm process runs cooler than intels.

Performance measured how? Let me guess clockrates and TDP's on 2 different architectures?

Don't understand the logic above.

AMD/IBM needs an extra metal layer on their process to get the same performance as Intel.... given IBM's announcements on low K (which should help in terms of interconnect density) - doesn't that seem odd?

We could also apply this to lithography... is AMD more advanced or less advanced on litho? Intel got by on 45nm with an 'old' litho solution while AMD needed new tech to print the features on 45nm.

If you want some serious debate come with some actual info and not the lack of background to distinguish architectural variables and process variables. "running cooler" is not merely reflective of process capability - for the 100th time look at Prescott and Dothan on 90nm (they ran 'slightly' different thermals, yet were on the same process tech).

Tonus said...

"All I want is a serious and unbiased debate on this, not the typical "AMD sucks" fanboy crap."

That's an odd statement, considering your previous comments. I was tweaking you back with my process comment, but I suppose you'll get your chance to prove your comment above, now that you have a reply from someone who understand process technology.

SPARKS said...

If you want some serious debate come with some actual info and not the lack of background to distinguish architectural variables and process variables. "running cooler" is not merely reflective of process capability - for the 100th time look at Prescott and Dothan on 90nm (they ran 'slightly' different thermals, yet were on the same process tech)."

I get positively giddy when you write like that.


"Sparks, here are two good reads for you."

Excellent link/report. Maybe the joker (read: flamebait) should read the link, although I don't believe the individual would have the capacity to comprehend its' significance.

If I could make a comparison between a movie and that wonderful second link--

"It's the difference between TRIAL LAW and PAPER LAW!"

Consortiums and Quorums who have different needs and agendas are no match for a focused and determined Intel Corporation. How long are these guys going to eat AMD"s/IBM's rubbish on the so called "new and improved"?

It's 11 quarters later and $7B in the toilet, and it's still-- 'Wait till next quarter,' "It's only a matter of time when we will see Intel opting for an asset light strategy in the long term."

Perhaps you're impressed by 100 overclocking chips that are not for sale or that no one will ever see?

Buy a clue, will ya! It's game over. AMD is a shell of a company it once was, and IBM's diversity dictates it cannot focus ALL it's efforts (money) on CPU production the way INTC can.

And that is "World Class" LAW.

SPARKS

Anonymous said...

LOL. AMD's or GlobalFoundries 45nm process is more advanced than Intel's 45nm counterpart. Don't believe me?


Actually everything you type shows you are fucking ignorant. Why don't you go
study this


Intel needs High-K interconnects to achieve the same level of performance that

Dude hate to inform you HighK isn't used for the interconnect wiring its used to replace the gate oxide to allow further scalling of electrical oxide thickness without exponential increase in gate leakage. Traditional oxide scaling stopped at 65nm and IBM/AMD have been penalized with far inferior leakage and short channel control at 45nm then INTEL who moved to HighK. Jury is still out if the claims by AMD/IBM on their HighK will make production on schedule for 32nm. I have more confidence with INTEL as they already did it on 45nm and have appeared to be making incremental improvements and shrinks. Call it there process "tick."


Also, I want to make it clear that wWhen it comes to process technology, nothing beats the IBM/GlobalFoundires alliance.
Did you read the provide link? AMD/IBM alliance process is inferior to intel by a long shot. They are getting close but still behind in schedule. An alliance by the way is a sign of weakness. Any company that can afford to go it alone would naturally do it. Like political alliances in war its because the individual players can't compete seperately but have to join forces. Have you ever had to join in anything. You know you compromise as everyone has different ideas and opinions and agendas. Beteween debate about priorities, debate about who's in charge, debate about who's more important and the resulting compromise you always end up slowing the process and making compromises. Anyone who goes it alone will always be faster and better. The only caveat is that he who goes along must have the money and resources. If you didn't notice INTEL has both. They have 4 factories I believe coming on line for 32nm to IBMs/AMDs one. That tells you who has the money and resources and must do it right.


No go study the link and learn before you post stupid assertions that are filled with incorrect delusions.

Its a seperate question what the CPU design team does with the process. In past INTEL design teams did stupid things like deep pipelined fast clock space heaters on 90nm with the likes Prescott, or misguided new architectures like Itanium. Having a superior process is no gurantee of product performance leadership, but its a lot harder the other way around to have superior product with inferior process. If you are doing that, which is AMD's position you are counting on INTEL to screw up big time on the design. Today the 45nm products from INTEL all look very good. The jury is out on 32nm Tick and Tock, but we'll have our first look very soon. The process technology appears to be there, the factories appear to be there, lets see how the design ends up. The arabs would have been better off investing in dog food. At the end of the day rich people always spend quality money on feeding their dogs, instead they have wasted billions on a consortium that will surely provide lower ROI then if they had invested in Purina and making dog food then mariginal crappy chips. The foundry business isn't all that its cracked up to be either. You wonder what them dumbfuck arabs were thinking. I guess wasting billions on hotels, skyscrapers, art, airline companies wasn't enough they wanted to piss money away in high tech too


Sparks, you are getting good at telling them the way it is!

SPARKS said...

Hey everyone, here's a nice picture of AMD's idea of "groundbreaking" in my home state New York, a tent, a bunch of lights, a podium, and five shovels, no less.

I'm not sure, but I think the shovels are inscribed, "Thank You NYS Taxpayers For Bailing Our Asses Out with 1.2B"

You don't need to make money in the CPU business, really. All you need are savvy deal cutters like Henry Richard and Wreaktor Ruinz, and suckers like Sen. Charles Schulman and Andrew Como.

F**K technological superiority! And Dirk, pickup a shovel!

HOO YA!

http://www.anandtech.com/cpuchipsets/
showdoc.aspx?i=3614

SPARKS

Anonymous said...

Sorry folks.. have to repond to the ignorance... bear with me.

Actually everything you type shows you are fucking ignorant. Why don't you go study this

My guess is you went to the modern technologies process chart and saw a point with IBM 45nm up and to the right of Intel's 45nm data point (which would seemingly indicate better).

My next guess is you did not look down and to the left and see another "IBM 45nm SOI" data point which would show it as far WORSE than Intel.

My next guess is you did not take the time to actual READ the summary below the table which indicates the 45nm IBM SOI data point was based on a research paper and USES HIGH K AND IS NOT THE PROCESS IBM AND AMD ARE USING IN PRODUCTION! (oops!)

My guess is you will bury your head in the sand now that you realize the 45nm process being used in production is the one down and to the left (meaning inferior) to Intel's 45nm data point.

And the final oops... that same chart on drive currents (actually Ion/Ioff ratios) shows Intel's 65nm process NEARLY MATCHING the 45nm SOI process IBM and AMD is using in production.... I'm sorry is that the point you were trying to make and just got Intel and AMD mixed up? :)

So to use your "colorful" language who's the FUCKING IGNORANT PERSON NOW!

Anonymous said...

In case you can't follow (this is from the last page before the summary in the link you provided)

http://www.realworldtech.com/page.cfm?ArticleID=RWT072109003617&p=11

IBM’s 45nm SOI process with high-k dielectrics and metal gates is very impressive, nearly matching the performance for the high performance 32nm processes. However from the description was more research oriented and will probably never go into production.

Surprisingly, Intel’s older 45nm high-k/metal gate process stands the test of time relatively well. It is unexpected that Intel's performance at 45nm would roughly match the two more foundry oriented 32nm high-k/metal gate processes for NFET performance and exceed both for PFET speed.

See the problem here is the typical IBM problem - Intel reports what it is actually using in production, IBM tends to throw up a bunch of data (inlcuding stuff in reserach which doesn't go into production - SiLK, air gap anyone?)- as evidenced by 2 separate 45nm SOi process points.


And if you actually have a clue what you are talking about, you may also find it amusing that Intel's 45nm process (using dry litho) has a significantly tighter contacted gate pitch over the IBM process (which I believe uses "advanced" immersion litho). I would caution people to not read too much into the SRAM cell size as there are considerations in that area which are traded off against trying to minimize cell size (I think the article talks about this a bit). Intel is marginally better for SRAM size but nothing significant. (Ignore the Toshiba et al SRAM cell sizes as they have different design considerations).

Anonymous said...

sorry - think the comment was directed at the wrong anonymous idiot... either way the info is what is (my apologies)

Anonymous said...

Any team can publish a very small SRAM cell. The devil is in the details of how good the yield it has, how much redundancy it uses, how much ECC it needs, how fast it runs, how much periphery circuits it needs. Lets wait and see the final product cache sizes and actual density. You'll likely find out that what is publish often doesn't show up in the product. Those that do the reverse engineering know the real story

SPARKS said...

Regarding "the link" which I regard as close to any single article can come to the "Rosetta Stone" for a non professional.

Highlights:

Eleven Layer Miniature Motherboards!!!
That cross section was absolutely beautiful. The traces were nearly 90 degrees vertical! (It's a pity they didn't throw a little scale division in the photo.) I'd like to calculate the number of copper atoms across those traces purely for shits and giggles.

Extremely Low Drive currents, 0.9 V!!!
And from what I can read between the lines, INTC is confident they can go lower!!!
I've seen SCR's the size of a fist that would laugh at 0.9 V.

Gate Last.
The 'consortium' goes one way INTC goes 180 degrees in direct opposition. Yet, the consortium is still dragging their feet at Gate First! (Thank you ITK, GURU for the annealing discussion. The article was quite clear on those sensitive gates!)

SRAM this SRAM that???
Why the hell do they always "demonstrate" (brag) functional SRAM at every node, then fall on their asses with logic???? It's like some boring rerun one gets sick of. I'm no industry expert, not by a friggen long shot, but I does anyone else here see this monotonous pattern, or WHAT???

FinFET Anybody for a game of horse shoes?
Sure, it was mentioned at the end of the article. But no one is wrapping their love around those gates just yet. (Hey, even in my business, the more contact area you get the better!) The silence is deafening.

SPARKS

Tonus said...

So... my joke about AMD's process being less advanced isn't entirely a joke, then? :)

InTheKnow said...

Why the hell do they always "demonstrate" (brag) functional SRAM at every node, then fall on their asses with logic????

Sparks, SRAM is used as a test vehicle throughout the industry. Everyone uses it during process development. A fab will make hundreds if not thousands of wafers with nothing but SRAM chips during development and start up.

It is a simple structure and easy to debug. So demonstrating SRAM is a natural first step. And you want to show off your progress as early as you can, so you show SRAM.

Search the press and you will find that Intel has shown off SRAM wafers first as well. What you need to look for when you read between the lines is whether they are showing off wafers or SRAM cells. If someone is showing off a functional wafer, then you are probably getting close to prime time. Otherwise it is a "research project".

InTheKnow said...

AMD's or GlobalFoundries 45nm process is more advanced than Intel's 45nm counterpart.

In case you haven't figured it out yet, process tech is about transistor performance. Since there is very little public information regarding IBM's transistor performance, it will be hard for you to support your claim. The link provided by Guru has more data than I've seen anywhere else. Good luck finding a public source with more data, I've tried.

Anonymous said...

Why the hell do they always "demonstrate" (brag) functional SRAM at every node, then fall on their asses with logic

In addition to what ITK said:

- It is a very dense structure (logic is typically less dense)... so it's a good test for some of the process areas that might have trouble with node scaling (litho, etch, gapfill, etc)... you are essentially pushing these steps (especially litho) as hard as they will need to go by doing the dense SRAM cells. Well that's not quite 100% true, but close enough:).

Moving from SRAM to logic as you mention is not a no-brainer...as the pattern changes, you can suddenly start running into process & manufacturing issues on steps that are pattern sensitive. And this is one of the major reasons for using the whole design rule or "restrictive design rule" concept...basically the more you can make the whole wafer look similar from a pattern perspective, the less process variation you will have on the pattern sensitive steps. This would make it easier (but not necessarily easy) to transition from a 'simple' structure to a 'complex' one.

And ITK brings out a real good point - you have to distinguish between a functional SRAM chip (100's of millions to billions of transistors) working and a working SRAM cell (6 transistors). IBM made a big deal about a working 22nm node SRAM cell working a while back - but an announcement ona single cell working should be taken with a grain of salt and not be confused with the first step of getting to a manufacturable process (which would be a full SRAM chip)

SPARKS said...

Well, I guess that's where the rubber meats the road. Gentilemen, for your pleasure at 34nM.

http://channel.hexus.net/content/item.php?item=19707

Anonymous said...

If you look at IBM look at all the stuff they got out of because it took world class efficient manufacturing; PCs, laptops, memories, disk drives.

Pretty much everything IBM invented and made when margins were high and it didn't matter if they were efficient or good at it. Once it become a commodity of sorts and htey had competition they couldn't compete anymore.

If you look at silicon they were never a big time manufacture that competed against others. Their internal server, mainframe business don't have to stand alone and make money, they make money on software and services.

I suspect that their semiconductor paper publishing is top notch, but when it comes to actually making hundreds of millions of chips at world class yields and prices they aren't so good.

IBM publishes before they make stuff and often when it goes to production it isn't what they published at all. For people that notice INTEL publishes what they make.

Anonymous said...

IBM publishes before they make stuff and often when it goes to production it isn't what they published at all. For people that notice INTEL publishes what they make.

Actually, intel publishes junk they *don't* make. What they do make, they tend to keep quite secret.

Anonymous said...

Now that previous poster is one who doesn't know WTF he is talking about.

So their 90nm, 65nm, and 45nm are double secret process. I guess their 32nm isn't HighK/metal gate but really III-IV but the don't tell anyone.

Have you seen any of the 45nm reverse engineered devices?

Its one thing not to tell you everything, its another to tell you someonething missleading and never put it into production for reasons like that other company, wink wink

Anonymous said...

Funny....

That previous poster works for intel and knows first hand exactly what they can and can not publish.

Anonymous said...

Actually I agree - generally speaking they do not publish the stuff they are using, preferring to keep it as trade secret. Once published, even if it is patented - it can be reverse engineered or engineered around.

As a result they will allow papers to be published on stuff that runs out not to be implemented in produtcion. Intel is well known for this (and not in a bad way) in the industry. This is not meant to be done in a tricky fashion or to fool others - it is a nod to the researchers who often times are not allowed to write papers about what they are doing. The pos(t)er above obviously not aware.

And once devices can be analyzed (samples out int the wild) - that is generally when they will announce or publish a paper on the sensitive stuff.

SPARKS said...

What secrets? Sure, proprietary production methodology may be in the 'black,' but EVERYONE knows where INTC is going.

"Now, Intel is planning to use this transistor architecture as basic building block for future microprocessors beyond the 45-nm node, which would mean that the upcoming 65-nm and 45-nm nodes will be implemented in conventional way. Intel expects volume production for 32-nm devices sometime in 2009 and for 22-nm devices in 2011, Mayberry said Monday (June 12) in a phone conference."

No, it wasn't Orthogonal quoted here for those who think he's giving away the company's production schedules. Orthogonal is the guy who humps his ass out of bed at 3 in the morning.

This article was written in 2006.

http://www.eetimes.com/showArticle.jhtml?articleID=189400035

SPARKS

SPARKS said...

TRI-GATE!

WOOOOOOOO!


SPARKS

Anonymous said...

WTF


they do not publish the stuff they are using, preferring to keep it as trade secret. Once published, even if it is patented - it can be reverse engineered or engineered around.

I don't know what you fools are talking about, but I'm refering to the INTEL IEDM and VLIS publications on their major logic process releases for the past few nodes. Things like embedded SiGe strain, LowK Dielectrics, number of metal layers, pitches for gate, M1, M2 etc, SRAM cell size and of course transistor drive currents, gate leakage, and such. They annouce and show what they are going to make. In the case of the HighK/Metal gate they waited a bit before they showed their hand and rightly so as why give IBM/AMD a tip as to what you are doing early. As you can see today they are the only ones doing HighK at 45nm a full 2 and half years ahead of everyone else. I guess I'm lost as to how they kept it secret. They've published and even claimed the superiority of gate last, showed cartoons of the flow and order for processing. Of course they don't tell you the exact metals and thickness but reverse engineering has alreadly given enough clues to the composition.

As a prior poster noted, some things they keep secret.

But I think its very incorrect to say they don't publish what is in production they actuall do that but not with much lead time.

This is different then many other companies who publish extermely small SRAM cells, talk about air gap, ultra low K like SILK, or biaxle strain but never put them into a product.

As to SRAMs they are important not only for demonstration. Last I checked almost every fucking product logic design has a data path, cache memory sometimes megabytes of them. Did you monkeys know they are SRAM cells?

Anonymous said...

Calling SPARKS you have one to many? 45nm and 32nm are planar transistors using HighK/Metal gate and strain engineering. Already annouced at IEDM this past december. Trigate is someone's wet dream, maybe orthos or ITK or little dick anonmous posters like me

Anonymous said...

I don't know what you fools are talking about, but I'm refering to the INTEL IEDM and VLIS publications on their major logic process releases for the past few nodes. Things like embedded SiGe strain, LowK Dielectrics, number of metal layers, pitches for gate, M1, M2 etc, SRAM cell size and of course transistor drive currents, gate leakage, and such. They annouce and show what they are going to make.

And again all of this stuff is after the technology is out (via engineering samples)... how many Intel 22nm papers were at the VLSI conf this time around? With Intel it's either stuff where early samples of the tech are already out, stuff way out in the future (III-V), or stuff that probably is not going into production (FUSI comes to mine). Obviously when the tech is out (or being heavily sampled like 32nm is now), more papers will come out - especially on stuff that can be determined through engineering analysis of a sample.

There is nothing wrong with this, but Intel just doesn't publish a whole lot and if they do it is essentially after the fact (well at least from their technology node perspective). This is obviously a different mentality then IBM which has a far more researched based mentality and my dealings with some of the folks there - well they like to look at how many and what types of papers you've written and they value that. (not just from a general public perception perspective, but also perception within the scientific community).

And I'm laughing at the claim of what Intel published on gate last... how much of that picture do you really think you got from what was out in the public? Again, this is not a criticism of the approach, it is smart. If you only knew the myriad of options and flavors of gate last and gate first processing flows...

Last I checked almost every fucking product logic design has a data path, cache memory sometimes megabytes of them

(chuckles) Yoo do know that the SRAM cells published (and sometimes even the ones in the workiing SRAM chips) are not the same SRAM cells that are in chips... due to speed, power, etc they are tweaked... for example the L1 cache SRAM cells are typically larger than what is published.

Or even more basic - how many transistors are in the Nehalem cache (SRAM cells)? How many are in the cells that are announced?

The SRAM cell announcements are nice and can give you a qualitative feel of scaling (for example a less than 2X shrink on SRAM might tell you if a company is having scaling problems - whether it's litho, ILD, metallization) but the absolute values need to be taken with a large grain of salt as these cells are tweaked for speed (L1 vs L2 vs L3), power and latency. Don't get me wrong it is an important first step and building block but the SRAM cells that are announced are typically not what goes into the actual product.

Tonus said...

So... we have another Intel engineer or two posting here on the sly? Excellent...

A Nonny Moose said...

Anybody know why the Phenom 2 can't overclock well with a 64-bit OS, compared to Intel CPUs? Anandtech made a reference to this in their review of the P2 965:

A 625MHz overclock is nothing to sneeze at, but it certainly does not come near those 6GHz~7GHz clock speeds we see on a regular basis with the Phenom II series. Granted, reaching those high clock speeds requires LN2 cooling but there is another important reason. We have discussed it several times and still do not have an acceptable answer from AMD about the inability of the Phenom II to clock much past 4GHz with a 64-bit operating system. Even with LN2 cooling we have not successfully benched past 4.4GHz with a 64-bit OS.

Once again, we tried XP 64-bit, Vista 64-bit, and Windows 7 64-bit and the results are always the same. As we near 4GHz, the voltage requirements increase dramatically and the clocking ability of the processor decreases in much the same manner. This does not occur in a 32-bit operating system, which happens to be the recommendation for any sort of benchmarking activities with the Phenom II.


I find this rather peculiar, seeing as how AMD loves to crow about inventing X64 and Intel having to copy it :).

Anonymous said...

So.... just went thru a slew of the links at the INQ on the "new" PhenomII 965, and I haven't hit one that's been able to OC the part to 4GHZ. AMDzone showed a shot of CPUz at just slightly above the 4GHZ mark, but no actual loading was shown (not sure if it was just one of those boot/CPUz screenshot type OC's)

Given the Vcore steps that appear to be needed to get this thing to 3.6-3.8 GHZ, a couple of things come to mind:

- Those TWKR chips must have been really cherrypicked, and it is clear why they are not selling them

- This part looks like an OC'd 955 part. When you look at the actual CPU power consumption it looks like a 25-30Watt jump between the 3.2GHz part and the 3.4GHz part and the top OC seems to be realatively close.... might as well just buy a cheaper 955?

I wonder if these are pretty much just the same chips put into 2 different bins. It's almost as if this is just a 955 with a higher stock Vcore setting (1.4V)

SPARKS said...

"Calling SPARKS you have one to many?"


Ok, OK, so I'm an Intel, shout it out from the cheap seats, junkie. So I'm smoking indium antimide in my free time. Beat deal, It sure beats those idiots on my morning commute who read the sports section of the New York Post, or lust after the glamor slugs contained in the pages within.

Tri-Gate
3D and III-V
FinFET
Semiconductor Nanowire
Carbon Nanotube FET

Sure as hell beats,

A-ROD
Madonna
Brad + Angelina
Brad+Jennifer
Lindsey Lohan
Paris
Brittany Spears

And last but not least M. J.

I heard there's speculation he died of food poisioning.





Yeah, he was eating thirteen year old nuts.

SPARKS

SPARKS said...

"So... we have another Intel engineer or two posting here on the sly? Excellent..."

Don't cha love it!

SPARKS

SPARKS said...

"- Those TWKR chips must have been really cherrypicked, and it is clear why they are not selling them."

I can see it now. Dirk and Wreaktor are sitting on folding chairs at the automated binning machines, pulling all nighters. Dirks eating two day old White Castles (shipped in from New York), while Wreaktor has a can of sardines and a bag of chili peppers.

After a thousand or so chips pass by, alarms scream, red lights flash, and half the factory comes down to take a look.

"LOOK!!!! WE'VE GOT ANOTHER ONE!!!"

SPARKS

Anonymous said...

AMD and GF business is a lot like the DRAM or Flash Memory. The companies got little choice to continue, but if you look carefully they got nothing but red to look at as there really is no realistic chance they will make money.

Like the DRAM and Flash companies they hope for the impossible and wank on. But the reality is they have a broken business.

AMD is finished. 32nm will be the turning point. Combination of impossibly difficult technology change, sky high investments, complex silicon/design interaction going thru the roof requiring the utmost coupling of process and design means all firms and especially AMD are going to go BK on 32nm.

I predicted it alllll along

A Nonny Moose said...

Been a while since I visited AMDZone, but I went there today and found this thread:
http://www.amdzone.com/phpbb3/viewtopic.php?f=52&t=136603

Pretty funny rewriting of history - apparently Intel's 65nm was really bad & leaky, whereas AMD's was great! :) I guess that explains the delays & low clocks Barcelona had when it debuted...

pointer said...

to A Nonny Moose:

yeah, and according to the AMD marketing guy, their server products executed flawlessly resulted a pull in ... wait, that is based on their latest updated roadmap ... :)

similarly, there is actually no delay and migth as well could be a pull in for Intel's Tukwila, based on Intel latest updated roadmap :)

Anonymous said...

Pretty funny rewriting of history - apparently Intel's 65nm was really bad & leaky, whereas AMD's was great! :) I guess that explains the delays & low clocks Barcelona had when it debuted...

The problem was not AMD's 65nm process, it was the first iteration of Barcelona/Phenom that didn't cut it for this process. If you look at the performance of dual core K8 offerings made on this process, you'll notice that power consumption was greatly reduced compared to their 90nm siblings.

I definitely think that Barcelona was not well optimized to take advantage of the 65nm process.

Anonymous said...

Define "greatly"... one power bin does not seem great...seems about average. You also saw AMD drop the top clocks on 90nm when they made the switch to 65nm - considering they were having trouble competing with Core2 at the time this seems like an odd move on a straight shrink and a mature architecture.

If the power was "greatly" improved, why would you drop the top clock bins on 65nm when you are having trouble competing on performance?

This smelled a lot like a leakage issue - which would be seen when you try to push clocks... if you cut down on the clocks you would not see the issue. It is also consistent with Barcelona being released at lower clocks and higher power. This could also be a design issue, but the early forecasts of higher clocks (remember the INQ dancing in the aisle story?), makes it appear to be more of a process window problem and not a purely a design issue.

SPARKS said...

" If you look at the performance of dual core K8 offerings made on this process, you'll notice that power consumption was greatly reduced compared to their 90nm siblings."

Oh the memories, and how we remember! Case in point, I call your attention to AMD's infamous dismal failure to counter Conroe, the venerable FX-74, AKA/ 'Quadfather' @ 90nM.

It quietly dropped out of sight after a failed attempt at competing with Sears portable space heaters. You needed two of them by the way (they weren't cheap). QX6700 was kicking their guts out for nearly the same price, stock clock, with 80%!!!!! less power.

Way too much and far too late.

http://www.techpowerup.com/?21066

This one say's it all.

http://xbitlabs.com/articles/cpu/display/amd-quad-fx_14.html#sect0


SPARKS

InTheKnow said...

I have just set up a blog to allow me to write more detailed posts. Don't bother responding to my posts there, just post your comments here. You can see my reasons for running the blog that way at the site.

Tonus said...

HardOCP overclocked a Core i5-750 (2.66GHz stock speed) to 4.0GHz and more. They had stability issues at 4.2 and 4.4GHz due to a problem with their CPU cooler that didn't allow it to apply sufficient pressure.

They feel that the CPU should handle 4GHz and higher speeds on air pretty easily.

Anonymous said...

"They had stability issues at 4.2 and 4.4GHz due to a problem with their CPU cooler that didn't allow it to apply sufficient pressure."

Funny how you fanboys make up any excuses for intel to always look good. :D

Tonus said...

Funny how you obsessive haters think that a budget class CPU that OC's better than 50% on air (and with a reasonable voltage level) needs an excuse to look good. :)

InTheKnow said...

All I want is a serious and unbiased debate on this, not the typical "AMD sucks" fanboy crap.

Now let's put the shoe on the other foot. Wander over to the link I posted with all the comparisons between Istanbul and Nehalem from SPEC.org. Then bring your "serious and unbiased" discussion to the table. I'm dying to hear you try and poke holes in the SPEC benchmarks or explain how their methods are biased towards Intel. :)

Anonymous said...

Funny how you obsessive haters think that a budget class CPU that OC's better than 50% on air (and with a reasonable voltage level) needs an excuse to look good

I think the need for excuses arises over the expectation that the chip has lots of upside and a 50% OC, which to most would be fantastic is considered "OK" thanks to the last couple of generations. As in "what....only 4.0GHz on a chip that runs 2.67GHz at stock?"

Now if you had a PhenomII 965 (or should I just say a 955 with Vcore effectively turned up?), you'd get a roughly 14% OC (most site have trouble going over 3.9GHz). Given the OC of the 3.2GHz is also in that range the expectations are kept fairly low so no excuses are needed when the high jump bar is set at 3'6" and it's still somehow an AMD career best by stepping over it...

It's also funny to watch folks over at CANzone talking out of both sides of their mouths regarding HT, tri-channel memory and Turbo... when it comes to benches it's how Intel gets artificially high #'s... when the benches start coming out on the dual channel, no HT chips... it's "look see that stuff doesn't do anything it's just blue crystals"

Well does it pump up the benches or doesn't it? The answer to that seems to depend on the story that is being told (it's one story when it is an i5 vs i7 comparison and another story when it is i7 vs PhenII)

Tonus said...

HardOCP compared the i5-750 running at 3.2GHz to the i7-965. The 965 was faster throughout, but the 750 was close in pretty much all but the bandwidth benchmark. They left out actual numbers and a lot of info because they're still under NDA (though amusingly, they acquired the CPU online from Fry's).

They paid $205 for it, so look to pay around $400-500 for a 750 with a motherboard, depending on the features you are after.

As for excuses about cooling, here was HardOCP's description of their problem:
"For cooling we used Thermalright’s TRUE Copper air cooling unit with a 72 CFM fan, which might be considered somewhat extreme at first glance. Keep in mind that we do NOT have mounting brackets yet for this cooler, so it was simply setting on top of the processor. Since it was just setting on top, we did not want to introduce a fan being directly attached to the heatsink causing a lot of vibration, so we simply set the fan off to the side blowing across the fins. Hardly an ideal setup, but it worked rather well."

In other words, with a heatsink balanced on top of the CPU and a fan that was sitting nearby, they got the CPU to 4.0GHz stable, 4.2GHz mostly stable, and were able to boot to the desktop at 4.4GHz. Some people apparently consider this to be an 'excuse!'

Anonymous said...

,i.In other words, with a heatsink balanced on top of the CPU and a fan that was sitting nearby, they got the CPU to 4.0GHz stable, 4.2GHz mostly stable, and were able to boot to the desktop at 4.4GHz. Some people apparently consider this to be an 'excuse!',/i>

Trust me, this is just a lame excuse. Even if they had the mounting brackets and the fan propperly mounted, the wouldn't achieve more than what they got. ;)

Now wonder why many warned me not to post in this blog: "it's infested with fanboys" :D

Anonymous said...

Trust me, this is just a lame excuse. Even if they had the mounting brackets and the fan propperly mounted, the wouldn't achieve more than what they got

Yeah I mean come on they could have pressed down on it with their hand and used their mouths to blow on the heatsink.

It should be obvious that pressure on a heatsink doesn't matter and mere gravity is sufficient. I thought every one knew the only reason to screw these things down is because when you stand the MOBO up vertically the damn cooler keeps sliding off. I ran into that problem several times myself! ;)

In fact once mounted properly it will probably LOWER the maximum OC as the lower temperature will probably just cause the electrons to move slower!

Tonus said...

"Trust me, this is just a lame excuse. Even if they had the mounting brackets and the fan propperly mounted, the wouldn't achieve more than what they got. ;)"

Ah, I see. You decry the "fanboys" here, but your response to the article's explanation is based on "trust me."

It's no wonder that you trolls don't last long when you try to stir up trouble here. :)

SPARKS said...

"Now wonder why many warned me not to post in this blog: "it's infested with fanboys"

No, no, no, you've got it all wrong. I'm one of two of the only self confessed fanboy's here. (The other one is quite smarter than I, but he 's not as nice, especially to AMD fanboys.) The rest are simply brilliant engineers/scientists who base their posts on data though a thorough understanding of process engineering down to the atom.

While you regurgitate some ridiculous nonsense about floating heatsinks, they have accurately predicted the performance and power requirements of AMD's lineup since I've been here, and that goes back to mid 2007.

Let me fill you in, junior, just in case you've forgotten.

AMD is billions in debt, they have had serious techical issues since 2006. They are a year, perhaps two, behind Intel in development and research.

They are a mere shell of a company they were three years ago.

Wake up, they've gotten their asses handed to them on a silver platter because they can't compete without taking a loss.

GOT IT??

(How's that, LEX?)

SPARKS

SPARKS said...

In the interest of setting the score straight, let's take a look at an eighteen month old chip.

Shall we?

This "fanboy" as had the aboslute good fortune of owning a QX9770 which has NEVER been run at stock speeds for nearly a year and a half (April 2008).

Air cooled Zalman 9700LED. (The LED's have since crapped out)

SuperTalent 1800 MHz Mem @ 2.1V. (Never more, never less @ 7-7-7-21 2T.)

P5E3 Premium (X48) (More features than you can shake a stick at)

1KW PC Power+Cooling SSI rated P.S. (A gift from GOD)

QX9770 @ 8.5 X 450 MHz FSB= 3825 MHz, @ 1.447V (I haven't the cooling for 4 gig, I will in November)

GURU, thou shall not speak about electron tunneling.


Sandra:
Mem. Bandwidth-10248 MB/s
Memory Latency- 57ns!!!! (Still the best!!!)
Cache and Memory-59.60GB/s (3 percent higher than a stocked clocked I 920.)

Processor Arithmetic-Dhrystone 59.78 GIPS
-Whetstone 54.70 GFLOPS
(Gets it's ass handed to it by every i7, 35% more for a i965 XE, ouch!)

Processor Multimedia-141.80MPixels/s
(Ever so slightly below an i965 @ 142.99!!! Any speculations/explainations from you genius engineers are welcome. I haven't a clue.)

Power Efficiency went right down the toilet as the dead lowest, barr none!(Ask me if I care)

Multicore Efficiency-Second best @ 22895. (Absolutely trounced by an i920 with block sizes smaller than 128kB. The combined score for an i920 was 33395 MB/s!!!)

SUPER Pi- 1M-12 Sec.

Fanboy? I've got my reasons for a year and a half. The best money I ever spent since my 486DX-33.

SPARKS

BTW: I'm heading back down to 3.6 (8x multiplier @ 1.41V) so the old gal can last another eighteen months, perhaps forever.

InTheKnow said...

In the interest of setting the score straight, let's take a look at an eighteen month old chip.

But it can't run prime95 at 4GHz 24/7, so we all know it isn't a real overclock.

I wonder if those new AMD overclocking Wunderchips can pass that test? Odd how that never seems to come up.

SPARKS said...

"But it can't run Prime95 at 4GHz 24/7, so we all know it isn't a real overclock."

Ah, ITK, your memory serves you well.

If I'm not mistaken you're referring to Dementia's challenge to me last year. Technically, he was correct, I could run every benchmark with complete stability with the exception of Prime95. The Zalman 9700 couldn't dissipate heat for longer than 5 minutes before some of the cores started to throttle. As you are well aware, the program loads all four cores 100%. (I wish there were some practical applications that would utilize my hardware the way that thing does. I would gladly throttle back to 3.6 GHz for complete utilization.)

However, you make a point that completely escaped me. The AMD boys dare not push an overclocked AMD processor that far, especially on the thermal precipice from which it currently sits. Actually, I have never seen the benchmark run on any of the newer AMD chips yet. In fact, I'd love to see the results, even at 3.4 GHz., on air of course.

Well done, I muust be getting Alzheimer's

SPARKS

InTheKnow said...

I found one of the statements in this article on AMD's Mangy-Cours to processor to be interesting.

Of late, AMD seems inclined to design new chips that fit within the thermal and power constraints of their predecessors. If this continues to be true, the individual clock speeds of those cores will have to fall. "Doubling the number of available cores generally means halving the power," Conway said. That generally means a 25 percent frequency reduction, he said.

So they are providing somewhat less than 2x the processor power (less than 2x because this won't give 100% scaling) for a 25% reduction in clock speed. If we assume the scaling to be 98%, which I think is fair, then you will see a 48.5% increase in processing power by moving to this processor.

Since AMD has seen the light and is using the bogus Intel MCM approach (who is copying whom here?) they won't suffer the huge yield hit that moving to a monolithic 12 core die would bring. So I think this is a pretty good move for AMD.

An increase in processing power of 48.5% will give Mangy-Cours comparable baseline performance and superior peak performance in the SPEC scores. With the power envelope being kept the same, they should once again be able to claim power/performance superiority. I'll work up the graphs and put a new post on my blog in a couple of days for those that might be interested in the details.

Of course there are also some improvements on the hypert-transport front and I can't speculate on how those will improve the scores, so my estimate here is probably worst case for the next gen AMD product.

SPARKS said...

"I think this is a pretty good move for AMD."

I agree. They've had an excellent product in the HPC and multicore server class(s). A good majority of the super computer class are designed around AMD processors. Additionally, if I'm not mistaken, operating system software licenses charge by the socket for the lower end stuff.

If the can get the same performance by scaling these multicore units, they may be negate the gains INTC has made with, say, a very expensive W5580. Further, lower clocked dies would allow them more binning options for the multicore package. Depending on price and power these units could be very completive, perhaps even a win.

The quote you posted is the absolute truth. My apprentice has an overclocked 8600. He runs the little bastard at 4.4 GHz all day long. If I did that with the QX9770, I'd need the cockpit/avionics cooling system out of an SR-71.

(If it weren't there the pilot could have baked a cake in his lap)

SPARKS

InTheKnow said...

Here is what you get when you let a marketing guy explain how the process works.

We don't "fight over die", server dies are different from desktop dies. With Istanbul this is really the case, with Shanghai the changes were far more subtle. Istanbul is a server die from the moment the first layer goes down. I would guess with quad core (though I don't know for sure) there is probably a point in the process where it could go either way. Emphasis added.

The process flow is determined from the moment the Si is issued from the starts room. The Si is assigned a "route" that defines the process flow. It's not like at some point 3 weeks after you start Si some bean counter is going to say "Quick put that lot on a server processor flow". The decision was made before the Si was ever released to the factory.

And the first layer comment was pretty amusing too, at least for Bulk Si. The very first step for any semiconductor device is surface oxidation of the wafer. I suspect it would be the same for SOI.

You don't want to start processing on a native oxide film. It is not very uniform and isn't the same from wafer to wafer. When uniformity is the name of the game the last thing you want to do is start with a built in source of variability.

So the idea that somehow there is something definitive that is being put down on that "first layer" shows a pretty clear lack of understanding of process flow.

Obviously, this was an AMD marketing guy. I'm sure it wouldn't be hard to find equally clueless remarks by Intel marketing types. Marketing types in general are so tied up in spinning things, I'm not sure they really know where fact meets fiction anymore.

Anonymous said...

Hi Guys:

I've found this interesting deal for those who're interested:

Overclockers Dream

Anonymous said...

Actually the 955 OC's just as well - not sure why you would spend the extra money. In fact the 965 is probably just a 955 with a higher stock Vcore.

Could you also recommend the 1000W power supply and the 200mm fan to cool that 140Watt beast down? Cause when you try to puch that thing just 15% (3.4 --> 3.9GHz) you are talking a nice Vcore bump.

I guess 15% is "serious overclocking"? ;)

Anonymous said...

Could you also recommend the 1000W power supply and the 200mm fan to cool that 140Watt beast down? Cause when you try to puch that thing just 15% (3.4 --> 3.9GHz) you are talking a nice Vcore bump.

Please take a thorough read of this:
Phenom II X4 965 - Can it be Undervolted?

At a stock core speed of 3.4GHz, we were able to reduce the load core voltage from 1.392V to 1.240V and retain 100% stability with the system. The idle voltage with CnQ enabled dropped from 1.00V to 0.832V. At idle there is a slight difference in power consumption and temperatures, but nothing to get excited about yet.

However, in our CPU load test the system power consumption drops 16% and temperatures 19%. In the gaming test, system power consumption is reduced 11% and temperatures 15%. Of note is that we were able to run the Northbridge speed at 2.4GHz without increasing NB voltages at the 1.240V (1.224 actual) core setting.


Why would AMD launch this processor with a high Vcore when it's not needed is beyond my imagination. You might have to ask AMD yourself.

And the quote posted above is coming from a pro-intel site, so there's no way there could be any bias.

Anonymous said...

ITK wrote..
Since AMD has seen the light and is using the bogus Intel MCM approach (who is copying whom here?) they won't suffer the huge yield hit that moving to a monolithic 12 core die would bring. So I think this is a pretty good move for AMD.

Obviously having two independent memory controllers and on-package interdie HyperTransport is the same as routing everything through a front-side bus. But yes, agreed.

Anonymous said...

"I've found this interesting deal for those who're interested:

Overclockers Dream"


AMD can't even get their promotion advertisement correct. The description of the MSI video card is talking about the ASUS motherboard. :rolls eyes.

InTheKnow said...

You might have to ask AMD yourself.

Why? I'm rabidly pro-intel, so why should I care why they do what AMD does?

Seriously, you are the one that brought the whole thing up. It seems to me the burden of explanation lies with you. Trying to foist that burden off on someone else doesn't cut it.

Anonymous said...

Why would AMD launch this processor with a high Vcore when it's not needed is beyond my imagination. You might have to ask AMD yourself.

Margin... plain and simple... if every chip could be undervolted as the ONE test demonstrated they would drop Vcore and likely fit this thing into a 125 Watt bin. The fact that they are putting this (at least initially) into a 140Watt bin means either the bin splits suck or they are right on the edge of the 125W/140Watt bin or they are simply "factory Oc'ing" a 955 to make a new SKU.

And what of course you omit is the question of whether the 955 can be scaled down similarly as well - you make it sound like AMD is intentionally putting this into a 140Watt bin just for kicks.... I suppose you know the answer to that and just want to bury your head in the sand.

Take a look at the top OC's for the 955 and 965 and tell us these are not the same chip with different Vcore and multiplier (due to the Vcore). I would not be surprised if these are all coming out of the same bin split.

If you want a "homemade" 965 laser etch a 6 over the 5 in the 955 and presto... same chip or you could pull out one of your crayons (though that might not help the thermals).

You don't know why AMD is running such a high Vcore...that's cute...
(The must be doing it for no reason whatsoever and simply to make the power look bad by creating a new TDP bin... for variety maybe?

Anonymous said...

Bottom line 15% OC is shit.. that is not an overclocker's dream, that is an overclocker's terd... why even bother OC'ing - you are pushing the Vcore up to a point where you could be limiting the lifetime of the chip, for what a 10-15% improvement? This chip is pretty much the anti-overclocker.

If you are going to OC on AMD, might as well start with one of the lower bins as you'll get to the same 4GHz wall anyway...

Remember all thos 'oh these things are going to hit 4GHz on air EASY!' predictions... well not so much. Though with a slight mod and hooking up a taser to your computer you might get the enough voltage to get this thing past 4GHz on air.

Anonymous said...

No prime95 on all cores... hard to say the undervolt was stable! :) Or is that standard applied by the UAEzone folks only when convenient? (that'd be one of 'dem rhetorical questions)

I undervolt my CPU and after dropping way down (nearly 20% reduction), I noticed on a couple of rare occasions (and seemingly application specific), the application would crash or the system would lock up. It's easy to undervolt and get most things stable. Eventually I bumped my Vcore up a bit for margin (down 15%).

Given the hatred of Anand when reviewing an AMD CPU, it is rather ironic how it is quoted as hard data when the results tell the story AMD fanboys want to tell... if this same undervolt test were done to an Intel CPU, it would be trashed as biased, incomplete, misleading and lacking a full load test.

Anonymous said...

Intel raises Q3 guidance.

http://www.eetimes.com/news/semi/showArticle.jhtml;jsessionid=X05EQPKFBCO5JQE1GHPSKH4ATMY32JVN?articleID=219500528

Chipmaker Intel Corp. raised its guidance Friday (Aug. 28) for revenue for the third quarter to $9.0 billion, plus or minus $200 million, citing stronger demand for its microprocessors and chipsets as the reason.

The previous guidance was for the Q3 revenue to be between $8.1 billion and $8.9 billion.

At the same time Intel said its gross margin percentage for the third quarter is expected to be in the upper half of the previous range of 53 percent, plus or minus two percentage points.


So the midpoint went from ~8.5 to 9.0Bil... that would also be a healthy increase as well from Q2 (+1Bil)...this would be a 12% increase (following Q2's 12% increase over Q1).

With the tighter range (they took it from .8Bil to .4Bil), I'd be surprised if they missed that #. The stock is back over $20, if you believe in the economic recovery (not sure I do) this stock should see $25 within the next 12 months (while paying a 2.8% dividend).

Anonymous said...

AMD prepares six-core desktop CPU

Quote:
ACCORDING TO A REPORT at Xbit Labs, AMD plans to launch a six-core desktop processor dubbed "Thuban" next year.

Xbit Labs claims the processor will be released in Spring 2010 and share much the same features of current Phenom IIs - except for the six-core die, of course - that is, a dual channel DDR3-1333 memory controller, 512KB L2 cache per core and a crossbarred 6MB cache. According to the site, it will still be a 45nm SOI part and will fit an AM3 motherboard (or AM2+, depending on the power specs).

Anonymous said...

6 x crap = more crap

SPARKS said...

It seems IBM is at again with another threat to INTC's dominance. True to form this report on the Power7 comes with a few caveats, of course.

http://www.theinquirer.net/inquirer/review/1532674/power7-vs-nehalem-ex

Incidentally, the INQ is calling this "The 8-core battle at the high end." Hmmm, it seems IBM and INTC are the only two players in the battle, and AMD is not mentioned. In fact, from what I read, there seems to be a hardware divergence between AMD and IBM.

SPARKS

SPARKS said...

Wow. If you thought AMD was in it deep, read this. Is it any wonder why they've hyping everything up lately? Anand's article reads like 'the quiet before the storm.' I think he's tired of being bashed. His article is, therefore, unabashed.

"What do you do with your 45nm fabs when you start moving volume away from them? Make really cheap quad-core Nehalems of course:"

"I'm talking $196. I'm talking faster than AMD's entire lineup. I'm talking about arguably the best processor of 2009. I'm talking about Lynnfield..."

"The pink block to the right of the die is the PCIe controller, that's 16 PCIe 2.0 lanes coming right off the chip. Say hello to ultra low latency GPU communication. You'd think that Intel was about to enter the graphics market or something with a design like this."

The die is cast, literally.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634

SPARKS

Orthogonal said...

I liked this one-liner at Hardocp:

"As for the Phenom II, AMD might as well nickname the Core i5-750, the 'Coffin Nailer.' "

:)

http://hardocp.com/article/2009/09/07/intel_lynnfield_core_i5_i7_processors/

A Nonny Moose said...

IMHO instead of killing AMD quickly with an axe strike, Intel is bleeding them to death with knife cuts. The i5 is yet another cut - could have been much more serious if Intel had included two PCIe-16 lanes in the onboard controller and clocked it independently so that the core could get 5GHz overclocks :). Anandtech hints at the latter on maybe a later stepping, certainly by the time the 32nm version is released.

100% agreement with Anand stating now AMD will have to drop the P2-965 price to equal or lower than the i5-750's $196. Yet more compression of AMD's pricing levels, and bye-bye any hope of increased margins since P2's die size is about equal to Nehalem's.

All AMD has in response is the cacheless Pro-pus :).

I'm gonna go out on a limb here and predict AMD either goes belly-up, bought out or leaves the CPU biz by the end of 2010.

Tonus said...

Nonny: "IMHO instead of killing AMD quickly with an axe strike, Intel is bleeding them to death with knife cuts."

Let's face it, they don't have to, and I think that trying to bury AMD all at once would be detrimental to Intel's bottom line, particularly in the short term. Long term, AMD doesn't seem to be a threat, so there's no reason for Intel to really crater their own ASPs.

Aside from one recent quarter, Intel has continued to record large profits regularly, while AMD has continued to bleed profusely even as they run through every accounting trick they can find. At some point you run out of ways to postpone the reckoning.

Maybe the "M" in AMD stands for Medicare?

SPARKS said...

"IMHO instead of killing AMD quickly with an axe strike, Intel is bleeding them to death with knife cuts."

Sorry Moose, Guru said it over a year ago. (Another one of his "speculations" that has come to fruition.)

He said 'Death by a thousand paper cuts' to be exact.

No matter, just Like Hartcourt Fenton Mudd observed, the key word is DEATH!

SPARKS

Anonymous said...

No AMD = even more scrutiny/cash grabs by gov'ts around the world.

Intel's growth is dependent on overall market growth not taking AMD's market share. If the overall market growth is 8-10%/year that means 2-3years of growth is the equivalent of taking AMD out of the game... (obviously that growth is not happening right now, but that is not a crazy growth rate in normal times).

It is far better for Intel to keep AMD limping along then to be alone. They still set the pace on technology standards for the most part (as volume typically drives development and acceptance of standards). And they pretty much can determine their own pricing at this point as they simply cap AMD's pricing with a mid-range chip.

Intel's challenge will be continuing to grow the market - computing power has far outstripped the SW/application needs and is becoming commoditized for a large chunk of the market.

Intel's challenge with AMD is to not kill them off but keep them limping badly and probably in the 12-25% market share range. One of the issues from an Intel perspective is that by offloading the factories to Abu Dhabi, AMD can limp along and keep swinging for a homerun as much of the debt/cash burn is on the GF side of things - the problem is until the x86 license is re-negotiated AMD has to continue to play the shell game of we still own a significant stake in GF (wink, wink) and thus are tied to it's health. Once those terms are eliminated (which they will as the current US administration will step in if needed), AMD will dump the GF shares quicker then Spansion. This is also why you see Abu Dhabi taking a run at Chartered - they know once AMD dumps their stake GF will fold without a larger scale and hence the Chartered move.

Anonymous said...

What a day today is.

Start of another school year and AMD is totally schooled again.

INTEL launches their lyfield platform and AMD's current lineup pretty much gets pushed down to loss leader segment. The whole line is now sorry consumer enty line. How many segments can AMD maintain? 50 buck processor, 60 buck processor, 70 buck processor, 80 buck processor. The frightening ramifications of their strategic blunders from years ago is now apparent.

Intel is readying 32nm and now selling 45nm for next to nothing. INTEL still makes a pretty penny, but AMD is yet to make any money and is sucking billions in cost in development and deprecation of the hardware.

Some smartass AMD fanboi will point out that Global foundries is doing the manufacturing, but the bottom line doesn't change the problem. Global foundry has invested a billions in 45nm only to now use that very expensive capital to make 100 dollar entry level CPUs.

Whats worst todays acquistion of Charter reconfirmed my original premise that Global Foundry has NO foundry skills. They made a lot of hype but they were nothing but AMDs factory. They had none of the stuff required to be a world class foundry NOTHING. Now they have to go spend a billion + to pick up a second tier foundry that at least has some design kits and customers. What a disaster of a business plan those arabs and dum AMDers are. Expect to see them ask Obama for a bailout in a couple years.

AMD and IBM are so behind when it comes to silicon it ain't even funny these days. Before people could make up FUD today even the most creative person can't create FUD. AMD is nothing but a total also ran there place for the future is making entry money losing CPUs for the most basic and cheap computers. This is like the Hyndai's of the auto world. There is a need for such cheap and crappy CPUs/computers/cars so I will sadly have to admit AMD won't die they are just relegated forever to making the cheap low performing stuff. Not a problem for me as I always buy the best and it ain't AMD.

SPARKS said...

Folks, here's an interesting read for your pleasure. There's a lot of dynamics going on here which, needless to say, I will never fully understand. This guy has got a handle on most of the key elements, however.

The plot thickens.

http://www.edn.com/blog/1690000169/post
/380048638.html

SPARKS

Anonymous said...

Global foundry has invested a billions in 45nm only to now use that very expensive capital to make 100 dollar entry level CPUs.

Intel's Atom is on what process node?

It's the use of capital and cost structure of the part that's the key... what's crushing AMD is the fab utilization (more so then the price). How many EE, BE, top of the line chips you think get sold?

It's amazing how people get so focused on desktop retail, they have NO FREAKIN CLUE what % of the business that actually represents. Not saying it's a healthy thing for AMD, but what % of AMD's (or Intel's) business is retail quad core desktops? (Yet it gets probably 90% of the press)

Anonymous said...

Wonder what impact the Chartered deal would have on the x86 license.

AMD is arguing that a 34% equity stake is sufficient (along with a 50% voting stake) to avoid the outsourcing language in the x86 license terms... this deal would certainly dilute AMD's equity stake in GF (unless they magically found some cash to put up 600+Mil) and I wonder if ATIC will continue to allow them a 50% voting stake?

I wonder how that part plays out; I haven't been following Intel's objection to the original deal - but I guess this only strengthens Intel's argument that this is clearly not an AMD controlled company and this is effectively outsourcing chip production.

I still don't get how the US trade regulators let the original deal go so quickly (not to mention the idiots in NY handing 1.2+Bil over to Abu Dhabi). After this it is clear that AMD will be divesting themselves completely once the x86 license is renegotiated and this will be a wholly owned foreign company country with access to technology that is normally under strict export controls. Once AMD divests from GF, the US will have no oversight of the technology.... unless they do an EU and just assume oversight control over any company they feel like.

A Nonny Moose said...

Sparks: Sorry Moose, Guru said it over a year ago. (Another one of his "speculations" that has come to fruition.)

He said 'Death by a thousand paper cuts' to be exact.

No matter, just Like Hartcourt Fenton Mudd observed, the key word is DEATH!


LOL - yes I remember that Startrek episode.

Anyway, death by a thousand paper cuts sounds so -- accountant-like! :). I much prefer knife cuts. Or maybe Ah-nold unloading on AMD with a couple bazookas.

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

Updated my blog again, sharing some information on Turbo Boost, and poking fun on claims on why one should turn off turbo boost with their CPU that support Turbo Boost ... (yeah, its from AMFUDZone again:))

btw, back to the Lynnfield release. It's kinda nice to see those good review feedbacks :)

Tonus said...

Here is your reading for the day. No comment needed, but you will want to make sure you have a towel to wipe your monitor with if you happen to be drinking anything at the time. :)

SPARKS said...

From all evidence, it looks as if the cash rich Middle East money backed Global Foundries is determined to be a significant player in the worldwide Semi conductor industry.

At first, what seemed to be a simple purchase and bailout of AMD, indicates GF has a larger, more dynamic plan for the future.

Obviously, they have PLENTY of money. This is not a factor.

The $3.9B, 14.5% stake in State owned Chartered (undoubtedly with more money to follow) puts them a serious position with both SOI, AND now, bulk process.

I believe the Mideast oil folks have looked at the world stage and realized the tremendous potential that micro electronics has now, but more importantly, in the future. (We all know HERE the future IS process.)

After all, I'm sure they realize the world will react significantly if oil pricers reach a threshold where alternative solutions become economically feasible. We are seeing it now in some small part, more so, if prices get out of control. 'Market Law' dictates this dynamic, oil is a finite resource, and the West is crafting new solutions that will come to technological fruition. It's just a matter of time and price. Therefore, they have found/need a new venue. One that's future proof.

With this move on Chartered I now feel AMD was merely a front door entrance way (INTC, License or not) to enter the market during a world wide recession where companies are weak and cash strapped at bargain basement prices Hence the business adage "Strike when your still hammering."

No one thought a few years back that a single entity could come up with enough cash to deliberately consolidate the weaker semi's into a larger Global Conglomerate. AMD is their crown jewel, and I believe this was only the opening in very large, well conceived plan.

I think AMD was a good start. This Chartered 'second move' was a good one, too. One more move like this, just like a Queens Gambit Opening in chess, in three moves you'll know you're in for a good game, they're in it for the long haul, and they mean business.

(Hey Lex, them Arabs ain't so "fucking stupid," eh?)

SPARKS

Anonymous said...

Sparks,

If you look across the landscape of mergers you see what kind of success record?

If you look across the landscape of mergers of marginal or money losing business what do you see then?

Money thrown at a bad business plan or a bad businesses ( IE ones missing key things for sucess ) have an even worst record.

Now you got people with money throwing money around like druken sailors in port. There is no shortage of desperate business looking for these handouts. Money alone isn't the answer. They need customers, they need market share they need technology.

Pss the combined new GF+Charter don't actually doesn't result in 1+1=3 its more like 1+1 = 2 which means the sum is not greater then themslves seperate.

They have radically different cultures. They have radically different technology expertise. Some might ague that the technology are complementary, but they aren't. Charter's bulk is of no measurably help to GF logic SOI and vice versa. Matter of fact what is the implication to ATI's business to TSMC? Will ATI really move its volume and GPU to Charter. Isn't that a huge step back for AMD's only money making competitive business?

And what can the money people bring in terms of expertise. You are talking people with no understanding of the two very different culturals or business. Do they really know how to mediate when disputes arrise. How will they force the right compromise if they have no clue. How can Gross manage a business that he tried to do at IBM but couldn't be successful?

Its desperate companies looking for anyone with money. Looks good on paper but its going to fail. Only loser is the arabs. The good is that some oil money is going to comeback to the USA in the form on construction and a few thousand jobs in Luther Forrest. But longterm AMD will continue to lose money as will GF. You know what I think the letters stand for

SPARKS said...

Good points all, no doubt.

However, for the sake of argument, let's leave INTC out of this, as they are in a league of their own. Obviously, going toe to toe with them is just plain stupid, if not suicidal. Sure, perhaps the lower end, and I do mean the lowest end of the market, the previously mentioned 12 to 20 percent gulag in which AMD has squeezed out a living for twenty some odd years, may be enough for GF as they grow and gain experience. Perhaps they can swing away until they hit that homerun/hailmary, who knows?

Fact is, they're still in the game with plenty of money behind them.

In the mean time, what I'm talking about, is a major player in the rest of the chip business, especially graphics, phones, microcontrollers, memory chips, etc. Up until now, INTC has not been a major player here. In fact, given the diversity of cultures, what would seem a detriment to the high end spectrum of logic, may be an asset in the remaining less sophisticated markets. Personally, I believe it's too early to say the whole GF approach toward chip consolidation is an exercise in futility. After all, they just entered the game, they are making some interesting moves, and they will be something to watch. I'm sure my beloved INTC is, too.

That's a fact.

SPARKS

SPARKS said...

I'm compelled to respond about the excellent IBM point you made. Christ, they couldn't keep focus on a what initially was a great hard drive (DeathStar), or a line of goddamned laptops. Maybe this guy Gross, running unimpeded by IBM corporate bureaucracy, can pull this thing together.

In any case, this is an interesting turn of events and a nice twist in the plot of our very unique daytime drama.

SPARKS

Anonymous said...

The merger is smart if they want any chance of long term survival... the fab business is all about economies of scale so you can keep the fab utilizations at a significant level and 1-2 fabs would not have been enough for GF to be viable long term. With that small a volume any appreciable hit in utilization by say a loss of a customer would crush profitability.

Consolidation will likely continue in the foundry space (UMC?) as it is clear these entities are not viable in smaller sizes and are unable to withstand any sort of economic variation. Also you will see more companies go fabless which should help the foundry model some.

I believe Chartered's tech is fairly similar to GF/AMD as they are all part of the IBM fab club - AMD was already outsourcing small CPU volumes to Chartered at some point (can't remember when), so from a technology perspective this makes sense - I have no idea about the corporate cultures, but that is usually overestimated as an impact anyway.

I'm still curious to see what impact thi smight have on the license and am surprised no mainstream techsites have even raise d this as a question/concern.

SPARKS said...

G, you always say what I'm trying to say better than I do! Even more so, I failed to mention the economies of scale and the IBM fan club.

SPARKS

Anonymous said...

People who don't understrand cultural must have never worked or lived in different enviroments. To underestimate how this can impact productivity, decision making and plan effectivness is to not understand human nature and how we work and communicate.

As to economies of scale. I fail to see how fabs that run completely different technology leverage economies of scale. They may share the same tools so you get leverage in buying them, fixing them etc. But the processes are so different. Today the world is swimming in capacity at foundry levels. Demand for leading edge foundry is small and confined to a small set of players who need, can afford and have the volume to afford the design, validation, and tooling costs. From the prospective of taking another competing company away it is good, from the point that Global Fuckup and Charter have economies of scale is laughable. Unless the converge to run the same tools on the same process there won't be economies of scale.

Economies of scale is running thousands of wafers a day in multiple factories to get yield learning and have huge factory that makes it efficient. Two small factories running two different money losing process managed by people from very different culturals won't work

Looks good to stupid arabs but its fucking stupid

Anonymous said...

I know gross he don't have what it takes, trust me I know

InTheKnow said...

I read this link and just had to chuckle.

In case you didn't pick it up yourself, the funny part was

"Today's consumer cares about what they can do with their PC, not what's inside," Nigel Dessau, chief marketing officer for AMD, said in a statement. "They want a rich HD (high-definition) and entertainment experience on their PC, delivered by the combined technology of AMD CPUs and GPUs, without having to understand what gigahertz and gigabytes mean." (emphasis added).

So customers don't care what is in the box, but they do want AMD CPUs and GPUs.

Gotta love it when marketing guys say one thing and then contradict themselves in the very next sentence.

Anonymous said...

As to economies of scale. I fail to see how fabs that run completely different technology leverage economies of scale.

Ummmm... they both run a process based off IBM's research.

ummm...Chartered has done the low end CPU biz for AMD (do a google search)

ummm... Chartered gives them 'trailing' edge access as GF is focused on 40/32/28nm bulk Si (in adddition to the SOI processes) and is a bridge to that as you, astutely for a change, point out that foundries are not so much based on leading edge tech.

GF was developing a 40nm/32/28nm bulk Si process (as most of the semiconductor world is still on bare Si)... people (esp idiots baking in the heat of AZ) don't seem to understand that SOI is not that different from Si... yes you have floating body to consider - but the basic manufacturing is very similar, of course you'd have to work in a non-manufacturing site and have an understanding of process technology, to comprehend this.

Foundries are about scale - Chartered helps increase the scale of GF. Will this make them successful? Hard to say (foundry business is tough). But this is a positive move for them - it is one less competitor and it delivers instant access to a customer base. Yes there will be assimilation issues, but fabs are not so dependent on human labor and corporate culture as you might think...

People who don't understrand cultural must have never worked or lived in different enviroments

The acquisition is an acquisition of customers and an acquisition of fab assets - while people are important and integration takes time.... the instant assimilation of customer (as opposed to having to win business away from Chartered) is probably worth and assimilation/cultural issues.... the key for any model base on large fixed cost (read: semiconductors) is volume and utilization.

Simple question - what is different (from a fab pespective) about an SOI and Si process? And considering GF had to develop a bare Si process anyway (as most of the world is on this), how is Chartered a bad thing?

Anonymous said...

I agree with the above poster.

Buying Chartered (theoretically) means that GF no longer needs to develop a bulk process from scratch. They can sync and use Chartered's process. I hope.

Actualy *doing* the sync may be a bit difficult (which set of engineers wants change?), but can probably be done with the proper management.

However, from a leading edge logic perspective, I wonder if this will take GF's "eyes off the ball", so to speak.

Yes, you can run two parallel processes in different locations -- ie two separate teams -- but you have to wonder how the org chart is going to look.

It's clear that GF needed to do this AND needs to integrate quickly to get their fabs full and start making money. Having seen firsthand what can happen with a "sustaining operation" and a "devleopment operation" in the same building, what starts happening is that the sustaining (ie profitable NOW) work tends to start eating resources away from the development work.

Again, it's possible to overcome these with the proper culture and force of personality from the top... but it may be a large concern from an AMD perspective in the medium term.

Time will tell.

I still think it was a coup for GF.

SPARKS said...

"Ummmm... they both run a process based off IBM's research.

ummm...Chartered has done the low end CPU biz for AMD (do a google search)

ummm... Chartered gives them 'trailing' edge access as GF is focused on 40/32/28nm bulk Si (in adddition to the SOI processes) and is a bridge to that as you, astutely for a change, point out that foundries are not so much based on leading edge tech."

You see! That's what I'm talking about! I may not know what a "floating body" is, (besides the ones dredged up every so often in the East River), but my inexplicable interest in following this stuff over the years instinctively informed me of the significance and dynamics of these assoiciations. (I just needed some real brains to spell it all out!)

Without a doubt, these are the main reasons behind the 3.9B purchase. There WILL be more to follow folks. They're committed, they're focused, and they've got plenty of moola. Fail or not, this is something to watch.

Well done. (As usual)

SPARKS

Anonymous said...

GF never needed to theoretically develop a bulk process IBM has bulk and SOI process I highly doubt Charter had a superior one. All they get is the customers and a money losing business.

This is bankrupt thinking that merging multiple losing money business generates economies of scale that make 1+1=3. That is a myth perpetuated by bankers and finance wizards who know nothing about running a business. Look at the DRAM consolidations in Europe, Japan and Taiwan. If economies of scale can really help you'd think merging companies that run 30K/month and more would really generate big scales. Sorry when you got a bad business bigger means bigger losses.

I predicted years ago that AMD would fall and they have to nothing but a maker of crappy slow CPUs. I told everyone when they were sitting high and Hector thought he was so bold to challenge Paul to some benchmark comparison. Now Hector is gone his company is in ruins and his pride crown jewels are owned by fucking arabs and he has no technology just IBM technology.

Sorry Charter was a bit player merging them with GF is a bigger global fuckup. Complementary losing business one in low end CPUS the other on low end foundry, a match made in heaven and funded by stupid arabs.

Yawn, times were more interesting trying to convince you that AMD was going to lose when they were ahead. Its so sad how fall they've fallen and fall further and further.

I got to think who is going to make their GPUs now. TSMC is really really going to be motivated to give them good service.

Anonymous said...

I'm lost how people think so small. Because somehow Charter was the backup slow / low CPU manufaturing option for AMD at somepoint is advantage how? Today AMD's leading edge 45nm is getting killed by Intels trailing edge 45nm. INTEL is getting ready to unlease 32nm soon. How is charter a help today. Charter fabs are running older slower logic and can't evne fill the fab. TSMC is also struggling to fill their fab with 65nm and 90nm. What the hell is the fascination with Charters inferior silicon. The onlything Charter brings to the table is the appearance of a full kid of building blocks for customers. But merging with GF is of little complementary value as these collateral don't exist for 45nm and if they did it would compete in the leading edge fab. If there were customers who would they be. Who needs or can afford leading edge capacity foundry, GPU and CPU that is it. So again 1+1=2 money losing business.. sorry not going to fly.

Anonymous said...

I'm lost how people think so small. Because somehow Charter was the backup slow / low CPU manufaturing option for AMD at somepoint is advantage how? Today AMD's leading edge 45nm is getting killed by Intels trailing edge 45nm. INTEL is getting ready to unlease 32nm soon. How is charter a help today. Charter fabs are running older slower logic and can't evne fill the fab. TSMC is also struggling to fill their fab with 65nm and 90nm. What the hell is the fascination with Charters inferior silicon. The onlything Charter brings to the table is the appearance of a full kid of building blocks for customers. But merging with GF is of little complementary value as these collateral don't exist for 45nm and if they did it would compete in the leading edge fab. If there were customers who would they be. Who needs or can afford leading edge capacity foundry, GPU and CPU that is it. So again 1+1=2 money losing business.. sorry not going to fly.

It's obvious that a rabid fanboy like you can't grasp simple things like these. let me educate you:

The reason why GF acquired Charteres is because a) they have a well-known customer base, b) what you call "old tools" will be a benefit for GF in the sense that no one in their right mind will use a High-performance 45/32/22nm SOI High-K process to build wireless chips, DSPs or embedded devices when you can achieve the same milestone with cheaper 90nm and 65nm peocesses. C) With this move, Fab one and the upcoming Fab 2 will be committed to AMD and whoever that demands high-performance processes like High-K and SOI while leave the low performance and cheap process to Chartered. Chipsets don't need the latest cutting-edge process. If you don't believe me, ask Intel. ;)

If you only remove that intel blind-fold from your face, maybe you'll understand the reason why GF did this.

Anonymous said...

I'm lost how people think so small. Because somehow Charter was the backup slow / low CPU manufaturing option for AMD at somepoint is advantage how? Today AMD's leading edge 45nm is getting killed by Intels trailing edge 45nm. INTEL is getting ready to unlease 32nm soon. How is charter a help today.

Are you really this clueless? This has nothing to do with AMD... this is about expanding the foundry business and customer base. Most of Chartered's work is .13um (130nm)and worse and I think the majority of their fabs are still 200mm.

This is not about AMD and making CPU's this is rounding out fab capabilities and foundry work.... Here's a crazy thought - some people who are making low volume parts may not want to go through the time and expense of moving a design onto a leading edge tech (mask cost, eng cost, iteration costs, etc)... as the above poster pointed out.

It's a bit mind boggling how short sighted some people are... and to use the memory business as an example of why consolidation doesn't work... here's a hint economies of scale wasn't the issue it was the price war and drops that crushed this... Intel lost money when they were doing memory too (and I'm not talking pre-processor days, I'm talking the NOR work which they spun off to IM flash)...

It's one thing to be a fan, it's another to be an idiot. Consolidation is not going to hurt them - you can argue it won't help (and initially it probably won't make a difference) but I don't see how this is a bad long term move. You will have a bunch of older fabs to support legacy parts and now Chartered will have a more advanced option (and probably cheaper) for those customers who want to move up in tech nodes (whether it's a high volume part that makes sense to move onto a lower cost process or a new part where you can just choose to start on the new tech). There's a lot of value to being a "one stop shop" for customers who may not want to deal with multiple foundries for multiple parts.

Anonymous said...

GF=one dedicated chip making for AMD that has lots money for its whole life and will continue to lose money for the future. Not enough market share to generate enough revenue to ever break even for the amount of R&D and capital required.

charter= second tier money losing foundry that has always lived on subsidies.

Put the two togather = two complimentary and non overlapping money losing operations where one doesn't help the other or vice versa.

Two different money losing operations now under one stupid ownership changes nothing.

No economies of scale, no technology sharing, no customer sharing, nothing in common but their long money losing history and same sugar daddy.

How does this change their competitive position for AMD, not at all. how does it help GF, not at all. Sure they can claim to have charter's foundry doesn't change much.

InTheKnow said...

Too many people here are focusing on process capability. In my mind this is first and foremost about customers.

My "foundry" experience is in the PCB field, so let me go through how this works there.

The PCB shop will post a list of process capabilities. These are the design parameters that they are able to achieve. And then they will sell their services based on their design capabilities.

For example, they will give minimum line widths and spacings (analogous to some degree to process node), minimum via size and aspect ratios, ability to do blind vias, or exotic inner layer materials, etc.

The sales force, armed with this list of capabilities, then goes out and solicits customers. If you don't have the capabilities the customer needs, then you don't make a sale. You also need to be willing to make low end parts, even if this loses you money, in order to get the more lucrative high end business. Many customers tied their high end business to contracts to make their low end products since they knew it wasn't profitable to the board shops to make their low end products. As long as the net gain was positive the board shops considered it to be worthwhile.

This last item is what the Chartered acquisition brings to the table. It give GF the option to address the low end of their client's needs and opens a lot of doors that were currently closed. It also gives them an instant customer base. This acquisition gives them a foot in the door. Without it, there is a lot of business that they would never have even been able to bid for.

Chuckula said...

OK, I posted this on Scientia's blog in response to the fact that he just bought a new Phenom 965... and didn't wait for the upcoming price cuts. Posted here to defeat censorship:

OK so we all know you are an AMD fanboy immune to all facts about anything Intel makes unless you can find a way to spin it in AMD's favor BUT:

Why didn't you just wait 2 weeks for the inevitable $50 price drop that we all know AMD is going to make to the 965's to undercut the new i5 750?

Notice how I'm not asking you to even consider buying an Intel machine since no matter how many facts come out you'll find a way to cast the i5 as being a greater threat to humanity than Al Quaeda... but why the hell are you not even trying to get a good deal on an AMD box?

P. S. --> Because I know you don't like free speech, this post is being replicated on another blog so people will still see it. Censorship. Always. Fails.

SPARKS said...

Chuckula, arguing with Dementia is like pissing up a rope. It ain't gonna go anywhere, and you'll only soil yourself. The guy is so biased, it's very difficult to believe it's all not contrived in one way or another. I could be wrong, but his tenacity for defending AMD is only surpassed by the number of times he's been wrong over the years.

The only joy I ever got is when GURU pounded him into the pavement, and that was a long time ago.

Why bother?

SPARKS

SPARKS said...

Lex, Lex, LEX! Your swinging after the bell!

Come on, the GF move on Chartered was prudent, if not, imperative. I agree with you 100%. AMD's business model is busted up like Humpty Dumpty. They can't compete with our beloved INTC, true. INTC has been sand bagging chips since 2007. Where in hell do you get 8600's merrily buzzing along @ 4.4 Gig all day and night? Christ, enthusiasts are pissed when don't get 30% overclocks these days!

However, ya gotta think outside the box on this one. There's a lot more business in integrated circuits than CPU's, man! I mean really, do I have to go through the entire Mouser Catalogue to show you where GF is headed?

http://www.mouser.com/Integrated-Circuits/_/N-6j73k/

And that's not everything! Do I have to list all the manufacturers? Mouser lists over a hundred on that same page. And that ain't everybody!

I don't know how many do their own foundry work. I don't know if I care. What I do know is these companies need foundries to produce even the lowest form of process, say an NE555 timer or bullshit generic logic chips. There are tens of THOUSANDS logic IC floating around out there in Never-Never Land, forget the analog stuff.

Sure, some of 'em may be obsolete, but all those companies want to stay in business and they need FOUNDRIES.

Hell, open up any device these days, discrete device's are in the minority!

http://www.itechhd.com/

Those are intelligent sound amplifiers!


Look, AMD is one part of GF, so it sucks. Maybe they have a shot at some past glory. Mostly, I think INTC will be there to kick their teeth down their throats.

But I'll wager GF is looking at a VERY big picture here (aside from AMD), and they are hell bent on being in that picture. You said it, they've got money to burn! Thats whats scary! AMD was only a start (In their minds it was a Coup D'etat and it put them on the map, big time) They want to service the GLOBE with Global Foundries, not just with CPU's.

Get it?

SPARKS

Anonymous said...

Why didn't you just wait 2 weeks for the inevitable $50 price drop that we all know AMD is going to make to the 965's to undercut the new i5 750?


I feel sorry if the only tech site you read for reviews is anandtech. Intel's i5-750 barely competes with amd's 965BE. Indeed, intel's competitor is the 955BE, which is priced at the same level as the i5-750.

Get your facts straight and do some more through investigation instead of posting such nonsense here.

SPARKS said...

"Get your facts straight and do some more through investigation instead of posting such nonsense here"

Because it won't get deleted, Sci, even if he is right!

We'll wait the two weeks, and then we'll see! It will be recorded here forever.

There is a respect for everyone's opinion, Sci, even yours.


SPARKS

Tonus said...

"ntel's i5-750 barely competes with amd's 965BE."

That may be reason enough for a price drop, isn't it? Putting aside the realization that one of Intel's mid-range "budget" level chips is competing with AMD's top-of-the-line CPU, the question is whether or not AMD will need to drop pricing in order to keep the 965 attractive enough to OEMs to keep them from replacing it with the 750.

SPARKS said...

Uh---is there any info out there about the shake up at INTC, and PLEASE don't tell me Big Paulie is moving on to greener pastures.

The boys in the back room ain't calling Centrino "Latrino" anymore.

We've got an 80.6 percent market share, the best in 4 years. AMD schleps it's 11.4 percent. I'll bet i5 puts them in the single digits by years end.

Now he retires!?!?! WTF!!!

SPARKS

Anonymous said...

Here is one to noodle, could INTEL be following AMD in planning to spin off their manufacturing?

If you read the press release

These two quotes jump out:
" Paul Otellini, Intel's chief executive, will devote a higher quotient of his time to corporate strategy and driving the company's growth initiatives."

"Second, Intel's global manufacturing organization, the Technology and Manufacturing Group (TMG), will now report to Andy Bryant, Intel's chief administrative officer and also an executive vice president."

Seems like they are moving the manufacturing guys under their former finance guy. Sounds like manufacturing is going to be less important in the new Intel, no? So this free's up Paul time to focus on other more important things then manufacturing. Couldn't this be the first step to having a finance guy run manufacturing as almost seperate entity, next step spin it off, no?

Anonymous said...

Couldn't this be the first step to having a finance guy run manufacturing as almost seperate entity, next step spin it off, no?

They can always use GlobalFoundires if that's the plan in the long term. ;)

pointer said...

or acquire GF :)

A Nonny Moose said...

Dunno if you guys have seen this thread on a Gulftown ES sample, oc'd to 6+ GHz :).

http://www.xtremesystems.org/forums/showthread.php?t=234592

At a lowly 5.9GHz, the 32-nm 6-core i9 scored 51,177 on 3DMark's CPU score... One poster thinks the i7-975 only scored about 10,000 points at 5.7GHz.

Things are looking more bleak for AMD's "chances" (and I guess zero counts as a chance) of recapturing the perf crown anytime soon. Somebody go wake up that dozing Bull! :)

Tonus said...

Heh, the guy who posted the images is an AMD fan. :) He got the images from this forum thread at another site.

To be honest, I'm less impressed with it hitting 6.3GHz under extreme cooling (which required more than 1.9v) than I am with what they did with air-cooling: 4662MHz at 1.47v. Are you kidding me?!?!? I want one. o.O

A Nonny Moose said...

^ Yeah Tonus, I saw that later after I posted the first time - really impressive, & shows Intel's 32nm is gonna be another good node, if not their best one yet. 6 cores/12 threads at 4.7GHz on air! That's gotta have the AMDfanbozos crapping their pants with envy, if not worry for their green deity! :)

I was going to dump the i920 & go with the i860 and one of those AMD or nVidia 57xx or GT300 cards, since the low latency looked ideal for a single-card setup. But now I'm back to the i920 and a 1366 socket.

Anonymous said...

Pointer said:or acquire GF :)

Let me fix that quote for you:

Or be acquired by ATIC!!
:D

Anonymous said...

Why AMD is in trouble and GF is going to be nothing but a money sink forever!

The newest reviews are in on the AMD X4. For the price they look good, damm good for 99 bucks. The problems comes to the small detail of how late AMD is to 45nm and how big the die is link. The die is huge at 169mm, not uncompetitive compare to the INTEL offering. The big difference is AMD is a year+ late and for the last year INTEL has been ramping yields, got 4 factories going, and also used their early lead to sell much more expensive chips on the 45nm. Thus, INTEL's lead allowed them to invest the billions, recoup the billions by selling high margin products. Then AMD comes late yelling me too me too, with a decent product. Problem is it sells for a give away price of 99 bucks. How can AMD hope to ever recoup their R&D investment. To catch up they need to spend more, but their current cash flow is too broken to spend more to catch up as their products don't generate the necessary cash flow. What a sorry ass viscious cycle money losing business model. Now GF with their stupid pimp arabs has figured that two money losing business togather make for money making, yeah right. They got just 2x the number of money losing business that add up to nothing more then foundry customers and CPU customer all under one holding company losing money togather. But heah, at least GF can now say it got real customers, too bad they lose money with all of them. It don't matter if its is 5 dollar custom foundry, 100 dollar GPU, or 99.99 AMD, still the same outcome is a loss

Anonymous said...

Why AMD is in trouble and GF is going to be nothing but a money sink blah blah blah...

Time will tell us who will be the money-losing company in the long term. ;)

Also, your fanboy rant really entertained me. Maybe that's why I come and read this crappy blog, to get some comedy relief after a stressful day at job. :)

InTheKnow said...

The die is huge at 169mm

So let me get this straight. AMD makes a chip on a die that is 169mm^2. They sell it for $99.

Intel makes a chip on a die that is 25mm^2 and they sell it for $30.

So Intel gets ~$180 dollars for the equivalent die area, that AMD is getting $99 for. But all the Intel detractors claim Intel is losing money on this small chip.

Someone needs to take a closer look at who is losing money, and it isn't me. Because with equivalent defect density, the smaller chip will give better yields. The smaller die is also going to lose fewer die per wafer due to edge effects and other sources of parametric yield loss.

Sounds like the AMD supporters better start backtracking on Intel's ability to make money on Atom. Either that or admit that this chip is, at best, a break even deal for AMD.

Yeah, like that will happen in my lifetime.

Tonus said...

"Time will tell us who will be the money-losing company in the long term. ;)"

I guess the last forty years aren't enough of an indicator.

SPARKS said...

"Time will tell us who will be the money-losing company in the long term. ;)"

Does the little 'winkie-winkie' at the end of your sentence mean to suggest that you think you've made a good point or a shrewd observation?

Let's take a look.

"money-losing company"- From 4Q 2006 to 2Q 2009 AMD has lost over 7 BILLION DOLLARS. They haven't turned a profit since. They are still in debt to the shareholders for 3 BILLION payable by 2012 and 2013. Those bonds are literally JUNK status, perhaps lower.

"Long term"- I really don't know what the definition of "Long Term" is in your little world is. 4Q 2006 was nearly 3 years ago, 2012 is 3 years from now. It reminds me of a song, financially:

'Clowns to the left, Jokers to the right' and there they are, stuck in the middle with you.


"Time will tell"- How much time???

Unless your traveling around the solar system at nearly the speed of light, in your imaginary space ship, 7 years IS LONG TERM, PAL!

(Read the above post, from a guy who knows what the hell he's talking about.)


"winkie-winkie"


SPARKS

SPARKS said...

Yo Moose! That Gulftown is looking REAL TASTY!!! 2Q 2010, no?

From what I reading, it may be an XE chip exclusively, at first anyway, X58 only.

QX9770 is getting long in the tooth.

Sounds like it's time for overtime!

HOO YA

SPARKS

SPARKS said...

http://www.pcgameshardware.com/aid,695432
/Core-i9-Intels-6-core-desktop-CPU-pictured/News/

SPARKS

Anonymous said...

It is funny how the release of the new AMD chips and the fact they have 45nm all them cum lappers at AMD are popping a boner. Them analysts and dumb a$$ arabs also seem to be getting a boner too.

But as some people point out only look back in history to see the track records of the company. When one looks forward the mountain gets steeper at 32nm and even steeper at 22nm. That means more money, lots of it. But beyond money you have to have the experience, the skills. I let past track record tell you who has the skills and resources. Who do you think attracts the best process engineers. Lets be honest who wants to live in Phoenix, Albuquerque, Hillsboro, or Lexlip. Or do you want to work for a bit company with a long track record of losing money and set up house and family in upstate NY or Dresden, duh...

I love the enlightening obeservation from InTheKnow on revenue / sq mm of silicon area. Guess what AMD makes far less on 45nm then INTEL does selling them netbooks. And INTEL already got a year of even higher revenue from selling quadcores. Now they are moving to 32nm. let IDF begin and here about Westmere and all the other goodies coming.

AMD is finished, was finished even in 2002. Not even a question worht debating, they had a broken business plan then, and still do now.

Anonymous said...

AMD fanbois at AMDzone can close their eyes and stick their head up their butts, the have been doing that so long that the shit is starting to smell sweet. Do the have a clue, rooting for the little guy is just stupid. Buy the best platform = INTEL, buy the best company = INTEL. But the underdog that always loses money = AMD

Anonymous said...

...Buy the best platform = INTEL, buy the best company = INTEL

LOL :D

Hope you enjoy playing games with your crappy GMA graphics.

And to add salt to the wound, enjoy reading this:

Pat Gelsinger and the Larrabee Fiasco

I hope I didn't pinch a nerve with that one. LOL :)

Having_Fun said...

...And to add salt to the wound, enjoy reading this:

Pat Gelsinger and the Larrabee Fiasco

I hope I didn't pinch a nerve with that one. LOL :)


That was priceless. This is pure comedy relief.

I'll get my popcorn and see what these fanboys have to say about that. Good find.

InTheKnow said...


I'll get my popcorn and see what these fanboys have to say about that.


Theo Valich. 'Nuff said.

Having_Fun said...

InTheKnow Wrote:Theo Valich. 'Nuff said.

HA HA. LOL :)

I have my popcorn and my Coke beside me. C'mon guys, bring the comedy relief. :D

SPARKS said...

"I have my popcorn and my Coke beside me. C'mon guys, bring the comedy relief. :D"

Well, here's another transient flameboy twit with nothing to say about CPU's since AMD as been reduced to a has-been design company and got it ass handed to it during the last 3 years. So he shifts the focus from CPU's to graphics. Obviously, the AMD crowd, with one hand on their peckers and the other on the You Porn site, have no one to entertain themselves with, so they come here for a little fun

Look boys, stick with the hand jobs, flaming, and YouPorn. It's something your used to. Wrector Ruinz and what was AMD have been jerking you off years.

"Barcelona"
"Phenom"
"Phenom II"

---- oooh was it good for you?

SPARKS

Tonus said...

"Hope you enjoy playing games with your crappy GMA graphics."

Heh.

Person A: Intel's new lineup of CPUs looks really good. I notice that AMD is dropping prices even faster, and they're in real trouble now.

Person B: Yeah, but your integrated graphics really suck!!!

As Sparks said, you're really reaching. Then again, this one wasn't as good as the "long term" remark.

SPARKS said...

I've been holding this one back for a while, but now seems the right time.

Having received a new apprentice recently, I've come to discover he is an 'enthusiast.' He's the one who's been overclocking the water cooled 8600 to 4.4 GHz. Interestingly, he was a big fan of AMD. His friend, who has been a fan of AMD for quite some time, refuses to acknowledge that AMD's time has past. He buys AMD components hoping to overclock them to compete with Intel hardware. My apprentice tells me there's no talking to him even with the number of successive failures he's encountered.

While at coffee one morning, my boy comes to me, hands me his phone, and says, "Read this."

It was a text message that said:

"it don't overclock for shit. dude amd sucks."

I almost spit my coffee all over myself.

Naturally, trying to be the objective the old sage overclocking expert, I asked my boy to text him back and ask him if he overvolted the processor, the memory, etc.. He did.

Here's what came back:

"dude I tried everything amd sucks"

SPARKS

Tonus said...

This may spoil the taste of that popcorn and soda. It's a list of companies that are facing possible bankruptcy in the near future. At number eight...

"8. Advanced Micro Devices

When will AMD actually make money again? The question is becoming more important by the day since it carries over $5 billion in long-term debt.

After losing almost $3 billion from 2007 – 2008, analysts expect the company to lose more money in 2009 and 2010.

While the shares rallied from their February $2 low, they still appear stuck in a long-term down trend from $40 highs way back in 2006.
"

Anonymous said...

Dudes, I've been telling all of you boys and geeks that AMD was going to go BK a long time ago. All them AMD fanbois just don't understand the business at all, nor do them dumb fuck arabs. Not any dollars out of my pocket, but I do feel bad for all them engineers that end up in upstate NY with wives, kids and dogs to feed when they find out in a few years they are out of a job. Just like them poor engineers and their families in Virginia. AMD is going to go BK is a gurantee. Unless them liberals continue to rob from the hard working and give to the people who continue to piss it away

Anonymous said...

BTW AMD fanbois only a few % of the world give a fuck about gaming performance. Most people don't live in their momma's basement, they get out have a real job, got a wife and family and do real stuff versus care about overclock and gaming. For the rest of us INTEL does pretty good.

InTheKnow said...

Sparks, I think you will find this link about the Intel reorg interesting.

SPARKS said...

Thank You, ITK!!

Now THAT was good for ME!! Great video, too!

All said, I don't agree with Mr. Maloney regarding the future capabilities of portable devices. I may be a know nothing from Podunct, and perhaps he's downplaying the segment strategically, but interestingly, he does mention the "fertile" area that we spoke of so often in the past rather flippantly. Personally, I think it's far more important than he suggests, or willing to suggest. There are just too many people with their faces in these things to indicate otherwise.

As far as 'Big Paulie' is concerned, NO damn it! Keep him right where he is! I may fly out to the next shareholder meeting and piss on the carpet in protest if they HINT of anything else! So he's not a Grandstander like Nvidiots' 'One Hung Low' or that other BIG MOUTH at Misrosoft (sic). Big Deal! We need less of those clowns, not more.

Walk softly and carry 10 big FAB's, I say.

GOD, I love INTC, thanks again ITK.

SPARKS

(I think I'll walk to the LIRR via Times Square tomorrow and stand underneath that NASDAQ sign, just for fun.)

SPARKS said...

When does little ATOM and monster 55xx Xenons come together for supercomputing? Right here, right on your desk.

http://www.sgi.com/pdfs/4177.pdf

Wow.

SPARKS

Having_fun said...

Sparks Wrote:...GOD, I love INTC, thanks again ITK.


It's really sad how people like you root for convicted criminals. This show you how our American values went to the toilet. :/

There's an old saying which sattes the following: every pig awaits for its time. ;)

Anonymous said...

Tick Tock Tick Tock, just like clock work 22nm functional SRAM is here, 3rd generation HighK Metal Gate. Nobody else even has 1st generation and INTEL already have their 3rd process node with it.

Got to love that Sparks!

SPARKS said...

"This show you how our American values went to the toilet."

Quite right. You couldn't have said it better. Perhaps I should "root" for a non American company who made business blunders of monumental proportion, sold a majority of itself to Middle Eastern Oil, fouled up an entire generation of process, spent 5 billion on a graphics company they could ill afford, raped it's shareholders for billions, and in one fell swoop, single handily took on both Nvidia AND Intel?

Criminals? Where? Europe? Not here, not in this country. Drive off in your Japanese car, wish the worst on successful American companies like Microsoft and Intel, and sell out the USA in the process. You've already sold out Intel by default.

Buy a clue, wishing the best for Intel IS the best for the USA.

Not doing so is the MOST un-American I can think of. You should rethink your priorities on was is and is not American values. They are the winners, and they are the best in the world. You're not proud of this, as an American? You call yourself an American?

Maybe you cheered on the EU for fining the Intel "criminals" 1 billion Euros? (AMD will not collect a dime.) Ra,Ra European Union Ra, Ra. People like you make me sick to my stomach, with your twisted, sick, convoluted logic.

Anything less I call treason. I am damned proud of INTC and it's brilliant employees.

"This show you how our American values went to the toilet."

Well said, got a mirror? Take a GOOD look. Na, don't bother, you're too single minded and one dimensional to see anything this profound. Either that, or you're just a stupid kid.

SPARKS

InTheKnow said...

You're welcome, Sparks. It just took a few days for someone to step back from the media feeding frenzy and take an objective look at the shuffle. :)

Anonymous said...

http://www.eetimes.com/news/semi/showArticle.jhtml;jsessionid=KFFQUIFMMUFZ1QE1GHRSKH4ATMY32JVN?articleID=220100588

This is fairly impressive - 22nm SRAM test chips (not cells like IBM reported).... these are functional SRAM chips with 2.9 BILLLLLLLION (said in my best Mike Myers voice) transistors, as opposed to a functional 6 transistor cell IBM reported.

I do wonder if it is a planar transistor geometry or trigate... my bet would be on planar for the SRAM, but I would not rule out trigate for the final solution (50/50 chance?). I had previously though SOI/trigate for 22nm - I'm starting to wonder now though....

Tonus said...

HardOCP got a look at the Radeon HD 5870 and they liked what they saw... a lot. Looks like Asus will offer a version of the card with adjustable voltage!

Looks like I might be replacing these 4870s sometime soon...

Orthogonal said...

Anyone else excited about Lucid's Hydra chip? Looks like it's back in the news and it's not vaporware. It was making some noise last year with grandiose claims about near linear scaling with multi-gpu's. It seems we'll get a taste of it with the first Mobo embedded Hydra chip next month. If it works, I'm game, every time I upgrade to a new GPU the old one isn't worth selling so I've got a few laying around. Would be nice to put them to some good use.

http://anandtech.com/video
/showdoc.aspx?i=3646

InTheKnow said...

Sparks, unlike the last one, this article will make your blood boil.

Personally, I think Enderle is way off base claiming Intel is not hitting on all cylinders and that Otellini has no vision. I think Paul Otellini has done a much better job of painting a vision of the future for Intel than Craig Barrett ever did.

Having_fun said...

Quite right. You couldn't have said it better. Perhaps I should "root" for a non American company who made business blunders of monumental proportion, sold a majority of itself to Middle Eastern Oil, fouled up an entire generation of process, spent 5 billion on a graphics company they could ill afford, raped it's shareholders for billions, and in one fell swoop, single handily took on both Nvidia AND Intel?


LOL. So, according to your flawed fanboy logic, AMD is not american??

HAHA, LOL.

Then, I can surely say that Intel belongs to Israel, India and the chinese. Let's not forget that intel tech is being used in Iran also. :)

Not rooting for AMD is as anti-american as not rooting for intel. But let's get our facts straight here: INTEL IS a convicted criminal. It's so sad that a HUGE company like intel, even though having the resources, had to drop so low by threatening OEMs and retailers not to use AMD processors. Really lame if you tell me.

The good news is that AMD is working on the American civil lawsuit. This is the one (God spares life) where AMD will DEFINITELY get money from intel for compensations and we are not talking small money here. ;)

Intel lost in Korea, Japan and in the Europe. There's no reason for them to succeed in the States. Paul Otellini should be lucky not to be jailed if then.

Sooner rather than later, criminals will pay.

InTheKnow said...

Sparks, kindly ignore the troll. He isn't adding anything to the conversation, and frankly, isn't even entertaining.

I won't think any less of anyone here for just letting him babble to himself in the corner.

Having_Fun said...

InTheKnow WroteSparks, kindly ignore the troll. He isn't adding anything to the conversation, and frankly, isn't even entertaining.

I won't think any less of anyone here for just letting him babble to himself in the corner.


I know truth hurts, but you have to cope with it. ;)

SPARKS said...

"Sparks, kindly ignore the troll. He isn't adding anything to the conversation, and frankly, isn't even entertaining."

I had every intention of doing so.

BTW: I'm still digesting that last link you posted. More to follow.

SPARKS

SPARKS said...

"2.9 BILLLLLLLION (said in my best Mike Myers voice)"

Sorry, I beg to disagree. Carl Sagan said "BILLLLLLLION" best.

Hmmm, 32 before Christmas? They released QX9770 two years ago on the month of December, hmmm. Perhap's Santa will treat ole' Sparks to a nice new X58 platform, with all the trimmings, if you know what I mean.

SPARKS

InTheKnow said...

It looks like Global Foundry's schedule is slipping only 2 months after releasing it. Ironically, it is the introduction of their HK/MG SOI process that is slipping the most. It is slipping to the tune of about 6 months, or if you want to look at it another way, one quarter of the time it will take Intel to move from 32nm to 22nm.

I love the denial from GF, though.

GlobalFoundries denied its roadmap has slipped. ''Our roadmap for 32-nm SOI has not slipped,'' according to a spokesman for the company. ''Yes, the timeframe for introduction has been altered slightly compared to the roadmap we showed you in July, but that is not because of any issues with the technology. The roadmap has simply been adjusted to align with AMD's product needs.

It's not slipping, AMD just doesn't need it when GF was going to have it ready.

Like any good foundry, they are going to delay introducing the cutting edge process by 25% of its life cycle until everyone is ready for it, because all their customers are demanding that they move to this process in lock step.

Are these guys a foundry, or are they just AMD's manufacturing arm under a different name? Actions speak louder than words.

InTheKnow said...

Hmmm, 32 before Christmas? They released QX9770 two years ago on the month of December, hmmm.

Intel showed 32nm SRAM at IDF two years ago. Now they are showing 22nm SRAM at IDF two years later. Despite claims to the contrary, Intel seems to be sticking to their two year cadence pretty well. I certainly don't see a 6 month slip from Intel.

Love them, or hate them, only a lunatic would deny that Intel is a highly tuned manufacturing machine. If Chrysler and GM could execute like this, the Japanese would be buying American automobiles.

Anonymous said...

Are these guys a foundry, or are they just AMD's manufacturing arm under a different name

I assume this is a rhetorical question? :)

The SOI process right now is ONLY being used by AMD, so I'm not sure what "aligning customers" has to do with anything - the one other customer is using bare Si processes and I'd venture to guess most new customers will as well.

That said I don't think it is a slip I think the parsing of words and nuanced schedules are now just being figured out... The previous schedules were based on a bare Si, non-high K process, now when you talk about a high K, SOI performance based process the schedule is different. I think I've been posting for a while here that it's about a 6 month lag to the bare SI process and that's the key comparison date to Intel manufacturing, not the blow the smoke schedule at the press for the bare Si, non-high K process the press had been misunderstanding.

And now? Suddenly there's a six month slip! In fairness GF is right, but they deserve this crap because they were intentionally misleading the press about the 32nm schedules and made no effort to provide a clear timeline and the various different processes and were more than happy to let the press misconstrue it. This is no doubt from the IBM PR platbook (see SILK, 22nm SRAM cell announcement, air gap, high K for 45nm... do I need to keep on going?)

Now I have to go rest after patting myself on my back so hard! ;)

Tonus said...

So in essence, what Global Foundries is saying is "our schedule didn't slip, we were just lying to you earlier." Heh.

Anonymous said...

EU releases damming Intel antitrust emails

Quote: The EU might not be one of the most exciting institutions around, but from time to time they do tend to dabble in some juicy, gutter tabloid stuff.

After Intel launched a massive PR campaign to publicize its appeal against the European Commission's record €1.06 fine, Eurocrats responded by releasing some spicy details of the case, including internal Intel emails.

HP was one of the companies to rat on Intel. It confirmed that Intel required HP to by at least 95 percent of its business desktop platforms from Intel. In a 2002 email, an HP exec wrote: "PLEASE DO NOT... communicate to the regions, your team members or AMD that we are constrained to 5 percent AMD by pursuing the Intel agreement." To make matters worse, an internal HP presentation from 2002 clearly shows that the Athlon desktop CPU superior to Intel's products at the time. HP said that it “had a unique architecture” and was “more efficient on many tasks”.

In a Dell email to Intel, a Dell exec said: "AMD is a great threat to our business. Intel is increasingly uncompetitive to AMD which results in Dell being uncompetitive to [Dell competitors]. We have slower, hotter products that cost more across the board in the enterprise with no hope of closing the performance gap for 1-2 years." However, in February 2004 a Dell executive warned his colleagues that buying AMD processors could have dire consequences for Dell, and that Intel executive were, “prepared for [all-out war]” if Dell “joins the AMD exodus”.

Acer also received payments from Intel to postpone the launch of an AMD-based notebook in 2003. An Intel executive commented the matter: “Good news just came from [an Acer Senior Executive] that Acer [has] decide[d] to drop AMD K8 [a notebook product] throughout 2003 around the world.”

Intel also convinced Lenovo to postpone the launch of its AMD-based notebooks in 2006.

Clearly, Intel was very, very naughty child a few years back, but ultimately, it paid a huge price for its uncompetitive practices. However, it obviously managed to keep AMD from getting a significant number of design wins and more market share through some of the biggest PC makers on the planet. Back then, AMD truly did have superior products, especially in the server and desktop market, and the true cost of Intel's campaign to AMD's bottom line in the long term will probably never be clear.


More evidence of the criminal acts of Intel. Craig Barret and paul otellini deserve to be jailed for this but I'm no one to judge them, I'll leave that task to God. It's a shame. AMD probably would be at 30-40% market share and Nehalem would be competing with Bulldozer instead of K10.5.

Khorgano said...

It makes me sad to see such garbage thrown around by the EU and the undereducated masses. It's unfortunate that most folks don't even know what a market is, let alone how they even work. That Intel is by definition NOT a monopoly, nor ever has been and unlikely to ever be one even if AMD and VIA close up shop. The kinds of business practices Intel is accused of as being Anti-competitive are ironically the exact opposite and by definition competitive.

Tonus said...

"The kinds of business practices Intel is accused of as being Anti-competitive are ironically the exact opposite and by definition competitive."

I think that there's a pretty clear difference between saying "we'll give you an X percent discount for every 1000 Intel CPUs you buy" and "we'll give you an X percent discount if you stop buying AMD CPUs." The accusations against Intel, as against MS in the past, are for the latter, for attempting to tie pricing incentives to limits on what the competition sells. That is indeed anti-competitive.

SPARKS said...

"Now I have to go rest after patting myself on my back so hard! ;)"

Why not? You've earned it. Any speculations on INTC's Tri-Gate at 22?

SPARKS

Khorgano said...

Tonus, I understand the distinction you're trying to make, but in reality there is no difference in the scenario's you've made. Whether it's offering volume discounts or preference discounts it's all the same. The fact remains that it's not Intel that's being "anti-competitive", it is AMD that is being uncompetitive.

Let me explain; The reason AMD has struggled isn't because Intel was "playing dirty", but because AMD was unable to keep competitive pricing. It has everything to do with AMD's business model and cost/capital structure. Take the HP scenario for example: If Intel tries to broker a deal by saying we'll give you x% discount if you keep 95% of your volume with us, then HP can easily calculate the cost difference of accepting that deal. They can then go to AMD and ask for them to match or exceed what Intel offer's. If AMD had a better cost/capital structure than Intel with higher margins, then even as a smaller competitor, they would be in the driver seat for pricing. Then HP could simply give Intel the finger and do whatever was best for them. AMD could then drive costs to the point that is unsustainably low for Intel such that they suffer losses and AMD reaps the profits. This has everything to do with AMD's ineffeciencies and nothing to do with "sales tactics" by either side.

This happens all the time in every other market, large players lose market share or go out of business completely to smaller more nimble companies all the time because of their ineffeciencies and difficulties in reacting to market signals.

If AMD can reorganize and improve their business model and cost/capital structure, they will thrive and challenge Intel.

InTheKnow said...

The SOI process right now is ONLY being used by AMD, so I'm not sure what "aligning customers" has to do with anything

I'm sure this is what GF is thinking as well. And that is precisely my point. Waiting for customers to be ready isn't the right mindset if you are a foundry.

A foundry has to roll the process capability out the door and publish that design capability as soon as possible. It is sensible to hold off on ramping capacity until significant demand is expected, but you want to get the design capability on your spec sheet as quickly as possible.

The line of thought here is that even if your customers don't use your cutting edge tech, it makes the sale of your less challenging process easier. The sales force can say: Look, we can do all this really challenging stuff. If we can do that, you know we can make what you want easily and you won't have to worry about low yields delaying delivery.

With the wording of this announcement, GF just doesn't seem to be looking at this the way a PCB shop (and presumably a semi foundry) would.

SPARKS said...

ITK, Speaking of predictions, and Mr. Maloney's downplay of the "fertile" area, it seems that your prediction of the significance of low power handheld devices, and INTC's interest in the market, are rendered academic.

Mr. Maloney is a pretty shrewd speaker, despite the charismatic showmanship.

SoC, we're looking at 7 to 9 watts. Not bad. I'm quite certain the boy's at Apple are keeping an eye on this one.

http://hothardware.com/News/Intel-launches-new-TVoriented-SoC-Predicts-TV-Revolution/

SPARKS

SPARKS said...

Hmmm, Duo, Quad, ......, Octo, so I guess if we had to fill in the gap, it would be a SEXA-CORE? I don't think the marketing guys would be too happy about that, frig 'em. They don't get sexier than this. She's well built, with a big pair of 2-16X bust line, uh, buss lanes.

Core i9 1000 series, perhaps?

Anyway, I know INTC, and I'm glad I waited. In fact, INTC did me a favor. They made the venerable QX9770 such a powerful desktop processor, I didn't need to upgrade to Nehalem until the time was right. Albeit expensive at the time, the overall mileage was TERRIFIC. They knew it, I knew it, and now you know it.

The integrated PCI-E controller on Gulftown was the deal maker. PCI Express buss latency will be a mere memory, pun not intended. I suspect that was the reason why QX9770 did so well in certain graphics applications, comparable in some cases to i7-920, if not better. Well, this should solve that issue

In any event, it's time to exercise some options on the New Egg account. In short, I'm very enthusiastic about this new machine, and they threw in two extra cores just for shits and giggles (yeah, I know Istanbul, bla bla bla, server part bla bla bla). I really don't care. Those two extra cores sit there and play Pinochle. However, from what I'm reading, Win7 is heavily threaded so it may just use all six (12?, whoa!) of the little jewels. Plus, INTC is buying up software companies like they're in kind of goddamned fire sale. Read: economic downturn = software companies on the cheap. Interesting, INTC has paid over a billion to make it easier for the software folks to do their jobs. Remember 'hyc' and his beef with programmers? He was pretty much on the money.

Win7 OS release that's optimized for multithreads, with INTC's simultaneous release of new processor tech? Well, it feels like old times to this aging enthusiast.

Keep the ball rolling, boys. Compel me to spend my money. That's what your in business for. Well done.

Lets see, ASUS Maximus Extreme, 6 GB Corsair DDR3 1866,..............hmmm.

http://www.pcgameshardware.com/aid,695432/
Core-i9-Intels-6-core-desktop-CPU-pictured/News/
&menu=browser&image_id=1172563&article_id=
695432&page=1&show=n

SPARKS

Unknown said...

Sparks, have you got a link to show Gulftown having an integrated PCI-E controller? I thought it was only LGA1156 processors that could have that; Gulftown is on LGA1366.

SPARKS said...

LEM, this is the link I posted. It looks like an official Intel document/release. That said, I, too, heard that the PCI-E Controller would only be available on LGA socket 1156. Frankly, I thought one of the boys on the inside here would straighten this/me out. They are, however, very tight lipped and get very sensitive about anything they feel may compromise the company beyond an official company release/discloser, and for good reason.

Here's the link to the "Official" document. It clearly says, "2X16 or 4X8 discrete graphics with X58. They could be wrong, however, or perhaps the thing is bogus.

http://www.pcgameshardware.com/aid,695432/Core-i9-Intels-6-core-desktop-CPU-pictured/News/

SPARKS

Unknown said...

Sparks, thanks. I did look at the link you posted first, but that says nothing about a PCI-E controller on Gulftown itself. The PCI-E controller on X58 boards is in the discrete northbridge chip, with the CPU connected to it via QuickPath (going by the diagram on your link).

On LGA1156 there's no discrete northbridge chip, the PCI-E controller is on the CPU die, with the southbridge connected via DMI (see http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3570&p=2 for a diagram). So the CPU is wired directly to the PCI-E slots.

Maybe Gulftown, if it has a PCI-E controller, will be used on LGA1156 somehow? It doesn't appear it could be used on LGA1366 since to access the PCI-E slots, the CPU must go through X58.

Geeze I really went out on a limb here, I generally don't pay this much attention to Intel stuff. Forgive me if I've got it totally wrong :P

SPARKS said...

Geeze I really went out on a limb here, I generally don't pay this much attention to Intel stuff. Forgive me if I've got it totally wrong :P

Cheese LEM, it ain't that important, right or wrong. AMD or INTC, just as long as you like this stuff as much as some of the REAL high tech birds that roost here. You're pretty much given a pass on Motherboards and CPU architecture here. When it comes to process tech however, these guys are as sharp as they come, and I mean REAL pros. I'm talking FUTURE cutting edge stuff.

All said, I think they enjoy guys like us who bring the fruits of their work to the table in this forum on a more practical basis. Be advised, learning about electron tunneling from overvolting CPU's doesn't make for a happy day. Nor does the degradation of microscopic traces, and this is not to mention leakage and other esoteric issues from going past recommended design limitations. Hell, I once backed off on my multiplier for week! Silly me. I'll show 'em.

I would like to find out about that integrated PCI-E controller. We'll find out soon enough, now that it's on the table. Someone here will straighten us out during the week, they'll probably work for a big CPU company, if you know what I mean. And they won't hold it against us if we're both misinformed, trust me.

SPARKS

Orthogonal said...

"I would like to find out about that integrated PCI-E controller. We'll find out soon enough, now that it's on the table. Someone here will straighten us out during the week, they'll probably work for a big CPU company, if you know what I mean. And they won't hold it against us if we're both misinformed, trust me."

Hmmm, I wonder who this is directed too, but I'll take a stab at it ;). I think Lem is on the right track here, the X58 Tylersburg chipset already has the PCIe controller embedded. It isn't possible to integrate the PCIe controller onto the CPU in that platforms current configuration. Gulftown is not slated for socket 1156 as far as I know, I think it's 1366 exclusive so I don't think this is possible.

SPARKS said...

Thanks, Ortho, I'm still for the 6 core unit. But this is better

It seem the boys at INTC have their ears to the railroad tracks these days. They woke up and smelled the coffee. Yes, Virginia they design a new cooler for what (I'm guessing is i9), and it is a BEAUTY!

Not only are they FINALLY embraced the lunatic fringe enthusiast community, they have also taken a page from our book. Those miserable extruded aluminum heatsinks were a waste of environmental resources, at best.

Then there was the mounting abomination/tragedy called the "push pins" mounting system. Do us all a favor, find the guy who designed that thing and beat the snot out of him, please. At least pee in his coffee! Core i9 XE gets a serious cooler with eight heat pipes, and get this, SCREWS!!!!!

SCREWS!!!

One more time---- SCREWS!

Four of them!!!!

Why not, again---- SCREWS!!!

Out with the old guard with some fresh, open minded, new blood! And to the person who pushed this new heatsink though the pipe, give that individual some stock options!

http://www.nordichardware.com/news,9971.html


It's about TIME!!! Now I can go to my grave with sheer delight that INTC finally supplied a good cooler for my $1000+ chips. Hell, it only took TEN YEARS! and untold hundreds of billions to wait!

I swear to GOD, I'm going to buy an extra one, as a collectors piece, and display it in my curio case!

SPARKS

Having_Fun said...

Seems like Intel is in "Turbo Desperate Mode" :)

Settlement looks likely in AMD-Intel civil case

Quote: Let's quickly go back to the beginnings of this case. AMD sued Intel in a Delaware district court in June 2005 in a very long filing that alleged all manners of misbehavior by the chip giant.

Since then, the case has developed legs greater than a million millipedes, with filing upon filing constituting a veritable K2 of documents and no doubt filling lawyers' pockets with endless dollars.

A number of other things has happened too - the most recent being a European Commission filing a week ago and one, moreover, that Intel's CEO Paul Otellini swiftly rubbished in a Q&A at its Developer Forum last week.

Many of the documents are redacted, that is to say edited, so we'll never know the complexities of the case, alas.

But talking to both parties off the record, it appears that it's in both AMD's and Intel's interests that the case never comes to trial. Say the trial begins in front of a jury in Spring next year and the jury finds in Intel's favor. AMD will appeal immediately. Say the trial begins in front of a jury in Spring next year and the jury finds in AMD's flavor. Intel will appeal immediately.

Intel stands to be fined something like $9 billion - that's about twice as much as two fabs cost to build, or nine times the cost of moving from one process technology to another.

Sources close to both companies told us last week that there was every indication that the case will never come to trial and there is likely to be a settlement. If and when that happens, all we'll ever see is a press release saying both companies have come to an agreement, with financial terms not disclosed.

There is, however, a complicating factor. After AMD filed the original civil antitrust suit, 80 other class actions started - these were consolidated some time ago. While it's definitely in AMD and Intel's interests to come to a settlement and put the past behind them, they cannot dispose of the class actions so simply.

Unless the ways and means committees at the chip companies come up with some cunning plan, that means both may be forced to continue shelling shedloads of dollars into the pockets of lawyers already grown fat on the surfeit of documents.

Is it in the interests of consumers - that is to say people - that this case be heard? Yes. Will this case be heard despite the best efforts of Intel and AMD and because of the class action suits? On that, the jury


This clearly says that Intel is admitting their wrongdoings and they want to "settle" things out before they lose those 9 billions.

OUCH!!!

Having_Fun said...

More news on the topic:
European Union Exposes Intel

Quote: The EU produced internal email excerpts of leading PC makers, such as Dell (NasdaqGS: DELL - News), Hewlett Packard Company (NYSE: HPQ - News), NEC Corporation of Japan and Lenovo of China, as well as Intel’s own employees.

The email excerpts displayed the PC makers’ concern regarding the usage of chips made by competitors such as Advanced Micro Devices (NYSE: AMD - News). They also expressed the favor Intel was likely to bestow on those players that exclusively used Intel products.

Specific allegations include special rebates to PC-makers on the condition that they do little or no business with AMD, making special payments to PC-makers to delay or cancel AMD-based products and offering server microprocessors below cost to strategic customers to prevent customer losses.

The EU stood by its decision to impose the fine, as it believes that the evidence is sufficient to prove Intel’s guilt. It reprimanded Intel’s practices, which are commonly perceived to be anti-competitive and anti-innovation, resulting in reduction of consumer choices.

The negative publicity notwithstanding, Intel continues to maintain its leading position in the microprocessor market, with an 80.6% share of global microprocessor revenue in the second quarter of 2009 (iSuppli). This was a 1.5% sequential increase and 1.4% increase from the comparable quarter of the prior year.

However, AMD’s market share in the second quarter declined 1.4% sequentially and 0.4% from the prior-year period, mainly on account of lower prices.


I agree with what Tonus said regarding the whole issue: ...there's a pretty clear difference between saying "we'll give you an X percent discount for every 1000 Intel CPUs you buy" and "we'll give you an X percent discount if you stop buying AMD CPUs."

Khorgano said...

"I agree with what Tonus said regarding the whole issue: ...there's a pretty clear difference between saying "we'll give you an X percent discount for every 1000 Intel CPUs you buy" and "we'll give you an X percent discount if you stop buying AMD CPUs.""

Like I said, there is virtually NO difference in either scenario. There is almost no difference in aggregate demand, economically speaking, with either business tactic. They are both legitimate business model's regardless of company size or marketshare.

Look at it this way. If AMD was the more efficient company with a superior cost model, then regardless of what Intel does, AMD can position themselves in the market accordingly and the so-called "anti-competitive" problem with Intel will take care of itself. This is not an Intel competition problem, it is an AMD competition problem and the ball is in their court to change the tides in their favor.

A Nonny Moose said...

According to the analysts this quarter which ends tomorrow is not gonna be a good one for AMD, although perhaps better than last one: http://finance.yahoo.com/q/ae?s=AMD

Looks like the average street expectation is 42 cents per share loss, or about $260M down the drain once again. Time for those Abu Dhabba Doo guys to ride in on their camels and bail AMD out again :).

SPARKS said...

TONUS, good buddy, may I suggest you read the below excerpt before you put any credibility in those EC released "E-mails," and the person or persons who sent them.

Below is the link for the entire paper.

"However, one important OEM, Dell, which the Decision says was coerced by fear of Intel “punishment” to buy exclusively from Intel, has confirmed publicly that it always considered itself entirely free to choose to buy from AMD, without fear of reprisal or punishment. The record before the Commission contains sworn testimony of Dell executives that contradicts this essential premise of the Commission.s case. The Decision nevertheless disregarded this evidence and instead relied on the speculation of a single lower level employee, who was not a decision maker and not even at Dell for much of the relevant period."


"The Commission made clear errors of factual assessment in making its findings concerning the Intel transactions at issue in the case. Over and over, it chose to accept less credible, and less numerous evidence – typically unauthenticated emails – while ignoring more credible, and more numerous evidence – such as written declarations (see, e.g., ¶ 440), statements under oath (see, e.g., ¶ 302), and even statements made by third parties under formal Commission procedures (see, e.g., ¶ 573). It ignored, again and again, highly probative evidence, and intentionally failed to gather other readily available evidence, including evidence in AMD v. Intel, United States District Court for the District of Delaware, No. 05-441, where substantial evidence regarding AMD.s allegations has been adduced."


Hmmm, he wasn't even employed by Dell during that period.

"It's the difference between trail law and paper law"- A Few Good Men

http://www.intel.com/pressroom/legal/docs/
Intel_Response_to_Redacted_Decision.pdf


SPARKS

Having_Fun said...

Sparks Wrote:Below is the link for the entire paper.

"However, one important OEM, Dell, which the Decision says was coerced by fear of Intel “punishment” to buy exclusively from Intel, has confirmed publicly that it always considered itself entirely free to choose to buy from AMD, without fear of reprisal or punishment. The record before the Commission contains sworn testimony of Dell executives that contradicts this essential premise of the Commission.s case. The Decision nevertheless disregarded this evidence and instead relied on the speculation of a single lower level employee, who was not a decision maker and not even at Dell for much of the relevant period."


"The Commission made clear errors of factual assessment in making its findings concerning the Intel transactions at issue in the case. Over and over, it chose to accept less credible, and less numerous evidence – typically unauthenticated emails – while ignoring more credible, and more numerous evidence – such as written declarations (see, e.g., ¶ 440), statements under oath (see, e.g., ¶ 302), and even statements made by third parties under formal Commission procedures (see, e.g., ¶ 573). It ignored, again and again, highly probative evidence, and intentionally failed to gather other readily available evidence, including evidence in AMD v. Intel, United States District Court for the District of Delaware, No. 05-441, where substantial evidence regarding AMD.s allegations has been adduced."


Hmmm, he wasn't even employed by Dell during that period.

"It's the difference between trail law and paper law"- A Few Good Men

http://www.intel.com/pressroom/legal/docs/
Intel_Response_to_Redacted_Decision.pdf


If that's the best Intel can show to the DOJ, I feel pity for them. It would be wise for intel to save their "dirty earned" money instead of that arrogant investing frenzy they're in now.

...And talking about the kettle calling the pot black:

Taiwan Semiconductor Manufacturing Company has scheduled a ramp up of monthly capacity at its 300mm (12-inch) wafer fab (Fab 14) to 6,000 wafers by the end of 2009, and to about 35,000 wafers in 2010, Digitimes reports.

The foundry is now purchasing backend packaging and testing equipment and plans to allocate capacity of 5,000 - 6,000 wafers for production of Intel Atom-based processors on a 40nm process.

Earlier this year, Intel and TSMC reached an agreement that they would collaborate on system-on-chip (SoC) based on the Atom processor. Under the agreement, Intel would port its Atom processor CPU cores to the TSMC technology platform including processes, IP, libraries, and design flows. This partnership will reportedly help Intel lower its cost structure to better compete in the mobile space, especially against ARM-based processors.

TSMC is also said to be reserving some capacity for production of Larrabee graphics chips in the future. However, Intel claimed the Larrabee silicon demoed at IDF did not come from TSMC.


Source is here:
en.expreview.com

SPARKS said...

ITK, you were right, he is in the corner babbling to himself.

D.O.J.? That rebuttal was sent to the EU As for the TSMC spew, I haven't a clue.

His penchant for irrelevancy demonstrates his stupidity as well as his ignorance, despicable.

SPARKS

Anonymous said...

People who think AMD and Intel will settle for any signficant money are just stupid. I can see a settlement where INTEl admits no wrong and says it will abide by being some pussy commpetitor. Beliavable as AMD is a castrated company these days. No balls just like all their supporters.

They are behind by two generations in technology, one generation in scaling. They sale entry level yugo like products. WTF would any reasonable OEM want their junk except to fill out the cheap end of the stack.

Another reason INTEL can't settle as some dumbfucks think is that if they admit anything then they got problems with almost every goverment on earth.

What INTEL did was simply agressive competition. Coke does it, Starbucks does it, McDonalds does it. The only reason people are going after them is its sounds like they are doing the consumer good.

Look into history was the attempt to break IBM and ATT a good thing or a bad thing. If you are for Arabs root for AMD, if you are for innovation and good business sense to be rewarded you are rooting for INTEL. AMD has and always will be nothing but a parasite.
Ruiz is a nothing but an afirmative action CEO.

Khorgano said...

And it's crackpots like this Anon joker that make it difficult to have intelligent economic discussions. With guys like this, it's no wonder the EU is pushing charges. At least the EU has an incentive to go after Intel on the order of $1B+ freebie pay days. It's unclear what incentives this numbskull gets, but guessing by his bigotry and tactless prose, just the thought of AMD's losses give him arousal in some warped unbridled sense of patriotism for his country.

Having_Fun said...

Khorgano Wrote: And it's crackpots like this Anon joker that make it difficult to have intelligent economic discussions. With guys like this, it's no wonder the EU is pushing charges. At least the EU has an incentive to go after Intel on the order of $1B+ freebie pay days. It's unclear what incentives this numbskull gets, but guessing by his bigotry and tactless prose, just the thought of AMD's losses give him arousal in some warped unbridled sense of patriotism for his country.

Khorgano, yo and Tonus seem to be the ONLY intelligent members in this crappy forum.

To me, it looks like you're living in denial and I know that deep inside you, you would like to believe in what you wrote awhile ago but the truth is far from that. Intel did wrong and they're going to pay for it, sooner or later. You CAN'T distort reality in Intel's favor and then later on believe that everything is nice and dandy. Pinch yourself, THIS IS reality!

What wrong with that?
Do you want a monopoly which will charge you $500 for a crappy Atom, or do you want competition which you and I know is good for the whole computing world?

If it wasn't for Intel, AMD would be far more competitive AND profitable than what they are now. Just have some common sense and think, meditate about it for a while before rebutting my post.

And since you talk about patriotism, yes, I am patriotic. God Bless America!!

InTheKnow said...

As for the TSMC spew, I haven't a clue.

Considering that he is using a source that is still pushing the "cost structure" thing, I think says it all. I shot that theory full of holes a while back.

Intel can afford to make Atom just fine. As Otellini reiterated just a few days ago at IDF, the TSMC deal is about expanding the customer base and has nothing to do with cost or capacity.

As to 5-6K wafer starts (I'm assuming that is a month since the quote doesn't say), that is about a weeks worth of output from just one of Intel's 45nm fabs. So the quantity is just spit in the ocean.

Anonymous said...

People here simply don't think very clearly.

If you look back at how the computer volumes/profits have grown INTEL and AMD are for larger and in the case of INTEL far more profitable now than 15 years ago.

Lets review, 15 years ago we paid 1000 for a high end desktop processor and close to 2.5K for a computer. Today 1.5K dollars ( something like an inflation adjusted 500 bucks ) gets us a pretty damm good computer except for the game fanatics. INTEl and AMD got here by selling chips for less and making it up in volume. People who think an Atom for 500 bucks would have been a hit are stupid. It would have been a dud as people would have continued buying Celeron.

Atom is a great example of how a Monopoly ISN"T bad. Look today ATOM has NO competition from AMD. Today netbooks fly off the shelf at less than 400 bucks. No AMD here just big bad dominating INTEL making a boatload of money with their monopoly. What is the number one reason people are buying Atom powered netbooks? PRICE PRICE PRICE. Sorry what did I miss? Do you think INTEL should earn a profit for the risk it took. Last I check they spent 7 billion of their OWN money to invest in 45nm capacity betting on Atom and other products.

If you look back when AMD was selling top end CPUs for more than INTEL because they kicked Intel's butt in the high end, did they take the risk and invest in more production? NOPE

Thus if INTEL was shackled than where would you be today?

No Atom no 45nm capacity from INTEL and we'd still have 999 laptops as we'd be compacity constrained. Yes AMD might be profitable you think their track record in execution on technology and product makes you think Barcelona would have come out better? That they'd have 3 fabs instead of 1? No we'd all be screwed with higher prices and less performance.

AMD fanbois are just like rooting for the underdog, AMD fans just don't understand the business.

Khorgano said...

To me, it looks like you're living in denial and I know that deep inside you, you would like to believe in what you wrote awhile ago but the truth is far from that. Intel did wrong and they're going to pay for it, sooner or later. You CAN'T distort reality in Intel's favor and then later on believe that everything is nice and dandy. Pinch yourself, THIS IS reality!

What wrong with that?
Do you want a monopoly which will charge you $500 for a crappy Atom, or do you want competition which you and I know is good for the whole computing world?


I appreciate your candor and moderate tone, but you're missing a few pieces of the puzzle to truly understand what is and is not acceptable business practices. Full disclosure, I believe in the Libertarian philosophy and free market capitalism and my personal beliefs align more closely with voluntaryism.

Intel is not a monopoly, never has been, and likely never will be. If AMD and Via were to close up shop, they still wouldn't be a Monopoly. They would simply be a single source supplier as selected by the market at that point in time. A Monopoly can only be granted by the State which places artificial barriers to entry.(i.e. Your electric and water utility companies). If you want to go after Intel's IP and open up the x86 license, that's another discussion altogether.

Intel playing hardball with OEM's is perfectly fine. No government should force Intel to agree to certain sale terms for OEM's, and no government should force OEM's to accept Intel's terms. Only voluntary mutual exchange is permissable.

Look at it this way. Legitimate/Acceptable business practices regardless of company size would be: Sales, Discounts, Rebates etc...

Illegitimate/Unacceptable business practices: Fraud, Physical Coercion/Violence, Use of Lobbies and Congress to shape policy and regulations that limit competition and create artificial barriers to entry.

Surely you can see the difference. As far as I have seen, Intel is not guilty of the latter and if they are, you had better bring compelling evidence.

SPARKS said...

"Considering that he is using a source that is still pushing the "cost structure" thing, I think says it all. I shot that theory full of holes a while back."

ITK, Yes you did, quite well.


"Surely you can see the difference. As far as I have seen, Intel is not guilty of the latter and if they are, you had better bring compelling evidence."

KORGANO, That piece was excellent, coherent, and well thought out. KUDOS. However, he doesn't have the capacity to do either.

SPARKS

SPARKS

SPARKS said...

"Other partners were a bit more cautious about the new program. One prominent AMD partner on the West Coast confessed that he had not even heard about the Fusion Partner Program a week before its launch. And another Midwestern partner of both AMD and Intel said the legacy of AMD's deals with Tier 1 partners in recent years "still stings."

"The Dell (NSDQ:Dell) thing is still on a lot of people's minds," said the source, who asked not to be named. AMD's late 2006 deal to supply Dell with desktop processors was seen by many in AMD's whitebox channel as a slap in the face."

Remember when poor beleagered AMD thought they had it all? One bitten twice shy. Oh no wait, let me think, how can we twist this so we can blame INTC.

http://www.crn.com/white-box/220300557;
jsessionid=FFHVYNL2H2Y1BQE1GHOSKH4ATMY32JVN

SPARKS

SPARKS said...

There's an article at X-Bit Labs entitled:

"Globalfoundries on Track for 50% Natural Yield by Year End with 32nm Process Technology."

Ok, what is a "Natural Yield"? Is it like a "Operating Profit" in the process world?

SPARKS

SPARKS said...

Well, that didn't work.

"It says to us: "let's put handcuffs and shackles on you and duct tape your mouth"."

http://channel.hexus.net/content/
item.php?item=20509

SPARKS

InTheKnow said...

Sparks, I have no idea what a "natural" yield is. Since semiconductor devices don't spontaneously assemble themselves the term seems out of place to me.

But you missed the really juicy parts of the article.

The 28nm technology offers the smallest SRAM cell size (0.120µm2) currently reported in the foundry industry, and an advantage in die size relative to 28nm “Gate Last” approaches. In addition, the Globalfoundries’ “Gate First” approach to HKMG simplifies 28nm design implementation and IP re-use for customers using conventional poly/SiON-based technology at the 45/40nm and 32nm nodes due to similar process flows and design rules.

and

IBM and its research partners first introduced the “Gate First” HKMG innovation in 2007 as the basis for a long-sought improvement to the transistor to deal with power leakage that emerged at the 45nm node.

First, I'd like to know where I can get one of these 28nm "Gate Last" devices. Since Intel isn't doing any half-nodes that I'm aware of, I have no idea what they are comparing to unless the IBM consortium evaluated a gate last device as an option.

Then there is the bit about the process flows being similar. I'm here to tell you the circuit designers that are going to purchase foundry designs couldn't care less about the process flow. The toughest customer we ever built PCBs for was EMC. They documented our process flows and validated all of our process control systems. But they never once tried to convince us that some other process flow might be better. All they really wanted to know was "can you hit our specs and can you assure us that you can do it reproducibly."

Now the bit about design rules does have some value, particularly if you want to do a dumb shrink from 32nm to 28nm. But if your design rules are too restrictive or don't fit my needs, I'll just take my design somewhere else. So a foundry's design rules need to be as broad as possible. I just don't think they are bringing anything different to the table here that what I would expect from any other foundry.

And that last bit is my favorite. They "introduced" Hk/Mg in 2007. Tomorrow starts the 4th quarter of 2009. So just where can I get a gate first Hk/Mg device 2 years after it's introduction? I can't you say? Surely you jest!

Oh, and lets not forget that gate leakage is something that "emerged" at the 45nm node. Didn't gate's leak prior to the development of the 45nm node. Did the IBM consortium discover gate leakage too?

In short, all I see here is a huge steaming pile of marketing manure.

«Oldest ‹Older   601 – 800 of 966   Newer› Newest»