Petz -

You are on record as saying that Centrino isn't pulling it's weight, whatever that means. You made no qualification and that statement isn't very specific so who's to say how far you are off? I'd like to comment but I know Banias volumes so I can't argue with you one way or the other....

Blauboad - A couple things suggest good Barton yields/splits.

Management has said as much in the conference calls.

Oh please!


The chips are extremely overclockable to the extent that all grades below 3000+ appeared to have been downbinned for demand reasons. The fastest chips are also good overclockers, which suggests that there is alot of headroom in the process.

This is not something that can be measured. There is no data to back up this claim. It would seem that if there was so much headroom AMD would release a higher bin unless they don't want to make A64 look bad as you suggest here:

AMD is about to release the AMD64, and I expect that they will want it to have a clear performance advantage over the fastest Bartons. That gives them a reason not to push Barton as far as it will go, since in September it will become a budget chip and its premiums evaporate. It would be a difficult marketing position to have your A64 3600+ priced several times higher from a concurrently selling Barton 3600+ that, based on the QS rating, should perform identically well. So, they may lose a few high-end Barton sales in the meantime, but AMD is betting the company on Hammer already anyway. What's a few more chips in the pot? Like I've said, if Barton does get released in a higher grade/s, that indicates very good things for A64.

Again this makes sense if the premise is correct but where is the independent data to show that Barton is highly overclockable?

No reported shortages or acquisition difficulties for Barton.

Depressed demand for AMD products?

[me] I have more confidence in my bias than yours: [you] I couldn't agree more :)

As it should be :)

Blauboad - Intel is also restrained by demand and supply, albeit to a lesser extent given their dominance and resourced. If they weren't, the Celeron wouldn't exist. Best Buy would be all .65 dual-core Itanium clusters.

Agreed.

It seems to me that AMD's production is going quite well, certainly on the bulk process. IMO, Barton's problem is a demand and not a supply problem. But this is an educated guess.

Then present a case and we can debate whether or not it is plausable. You haven't done that.

In reality, we have no clue what yields are on Barton or P-M and no easy way to guage demand.

I should tell you that I know exactly what yields are on P-M and all Intel products. That's why I can never comment on Intel's yields.

Interpretation on this matter tends just to reveal the interpreter's bias.

Yes, but I have more confidence in my bias than yours <G>

blauboad - I never used it as a technical term, anymore than "churning out Opterons" would mean that AMD is fabricating them from milkfat in wooden buckets.

Fair enough I supposed but nobody ever claimed or implied that some Opterons were "churned out" while others were produced by some other means.

NAS - Now at the distributor/retail level you could test for speed assuming the manufacturer downbinned parts for marketing reasons. At that point someone could cherry pick. Of course that only matters to the retailer/end user.

Yes, they could do that and I'm sure than many do. That could be called "cherry picking". All parts will perform in excess of their frequency rating if voltage and temperature are well controlled within their ratings. The point is that some people have the idea that Intel or AMD does "cherry picking" of some parts, while at the same time there must be some that aren't "cherry picked" and for the life of me I have no idea what those non "cherry picked" parts would be.

Edgar - I like some (combjelly,sgolds,yourbankruptcy) on this board were trading AMD when it shot to $80 a share (presplit) and also we where there on the slide back to 3. All in a couple years.

Don't feel too bad. I bought a ton of INTC at $66.

Blauboad - Whether or not Centrino is selling well probably depends on what you mean by "well." If Intel keeps the price premium on it, then that probably means that sell-through is very good. (Though when AMD manages to keep a premium on its high-speed chips, that can only mean yields are bad :)

It's all about balancing production capability with market forces, right? To assume that Intel and AMD price the same way is to assume their products are subject to the same capability/forces. They aren't IMHO. AMD has a few enthusiasts who will pay a premium price. At the same time AMD has real difficulty producing premium products so they need to control demand. Intel may not have the same restrictions.

Blauboad -

Then that's probably why you don't understand that terminology :)

I guess I didn't make my point very well. If you think some parts are hand-picked, you must be assuming the others aren't. How can that be? You pick cherrys by looking for the good ones. You can see the good ones and you can see the bad ones. You inspect them with your eyes. With processors you have to test them and the fast ones get the same test as the slow ones (well potentially). So if the fast ones are cherry picked, so are the slow ones. In fact the slow ones get more testing than the fast ones. So the term cherry-picking has no special meaning when everything is treated the same way.

If you're referring to the fab process itself then there is a strong possibility that AMD is taking some non standard steps to enhance binsplits at the expense of yield. When forming transistors there is a target size for the channel and controlling it defines, in part, the limits of a process generation. In simple terms, you can aim for Xnm with a standard deviation of Ynm, X>Y. If you miss on the high side you're OK but the part will be slower, good yield-slow parts. If you're on the low side you may have exceeded the limits of the other process aspects and the transistors won't shut off. So the lower the target the faster the parts but the more parts that don't work because you've gone too far. But this is not on a die by die basis and once again the term "cherry picking" has no meaning because you can't cherry-pick a wafer. So can you explain to me what you think "cherry-picking" is and how is a part treated that isn't cherry-picked?

Blauboad - This suggests decent binsplits, as I don't think AMD will release anything with less performance than the fastest Bartons for reasons I've posted before. If a 256k cache chip can manage that, it must be clocking well. And if they have that many at launch, it would be difficult to say they were handpicked. We'll see.

Can you explain something to me? I've been testing parts for over 20 years and people who have no test experience seem to understand "handpicked" verses "non handpicked" but the concept eludes me completely. What do you mean by "handpicked" and what would be "non handpicked"?

wbmw - Well, I assume you mean Q2

Of course. How silly of me...

Semi -

I would expect Intel to say something about Banias volumes in their earnings announcement. And I expect Petz to rewrite history too.

Perhaps Petz might venture a guess about Banias volume during Q1?

Jerry R

They do not consider AMD a for profit company. They consider it as the instrument through which their Intel hatred is expressed. AMD's financial statements represent the necessary battle losses in its war against big bad Intel. It doesn't matter to them how much money they lose. As long as they are fighting Intel, the AMDroids are happy.

A conclusion many of us have come to.

Keith -

The CPUs that go into what you call lowend servers sell for up to 3700 US$ per unit. Sounds like a good market to be in.

Yes it does. Where are the products and where are the profits?

BTW, why do you never put a subject into the "subject line" - because that´s what it´s for, as the name indicates.

Just a matter of style. I think it's polite to address someone by name.

Haddock -

To hear Haertel tell it that's not how it went at all

Not addressing this at you, but how ironic it is to hear AMD fans criticize the P4 as a bad design when it's smaller, much easier to manufacture and outperforms AMD's best. We would have to believe that to the AMD faithful, a good design is one that makes for a great slide presentation even if it's actual performance is not up to promise and is nearly impossible to produce. To AMD's credit they've proven that if a company has no regard for their shareholders and is willing to spend themselves into oblivion, they can present a design to match Intel's best lowend servers.

CJ -

Who knows? They very well might. But it is not particularly common to issue a single datasheet for multiple products without breaking out the different specs for the differrent products. Unless you want to claim that the 240, 242 and 244 have the same TDP...

As higher frequencies are unreachable without process improvements, perhaps they don't know yet?

sgolds -

Face it: The primary reason for the P4 queue design was to come up with a high MHz alternative to Athlon because they thought the public didn't know the difference. From the start of the P4 design (which was rather large and unwieldy), how many good performance ideas were thrown out just so they could get a high MHz chip to the market?

Face it, you can't use your premise to prove your point. You have to prove your premise. You haven't presented anything except to keep repeating your premise.

The only reason for the P4 is that Itanium is not ready for the general desktop market, and Intel needed something to counter Athlon. (Yes, I know the P4 was in design longer than that, I'm talking about the commitment to actually manufacture it. P4 was a backup design for the desktop at a time when Intel wanted to - wisely - hedge their bets against Itanium delays.)

And your evidence for this is....?

You know if you'd just say "IMHO" you would keep some credibility. As it is now you're long on rhetoric but short on facts.

sgolds

While I think the P4 concept is a cynical attempt to ripoff consumers with high clock rates

I thought you said you spent 20 years as an engineer. Doesn't sound like it here. You imply that Intel designed the uA of the P4 to run at higher clockrates for the single purpose of ripping off the consumer. Frankly that's a dumb statement. The P3 uA was out of gas on the highend and the fact that P4 even kicks Opteron's butt in single processor applications makes your statement look terribly foolish. P4 will scale to 10GHz over it's lifetime and with each process generation the performance advantages become more obvious and your statement becomes more foolish.

CJ -

Been ther, done that. I note that the 85W is for the Opteron family. True, currently it goes to 1.8GHz max, but 2.4GHz is planned for this year. Since they only give the one power figure, it is reasonable to infer that this power figure is to give the designers guidelines for all of those chips.

I'm sure you're right. After all, no one is allowed to issue a new datasheet along with a new product.

Tenchu -

what if Intel "merely" hits the high end of guidance, as Niles is predicting? People will most likely sell on the news.

Maybe. I quite trying to predict Intel stock movement a long time ago.

Intel to Hit High End of Guidance, Researcher Says


Online staff -- Electronic News, 7/11/2003


There was good news for Intel Corp. today from the Lehman Brothers.

The market researcher gave the MPU giant's Q2 revenue guidance its vote of confidence and looked positively at Q3.

"We expect Q2 revs near the upper end of guidance of $6.6 to $6.8 billion, with EPS of $0.13 to $0.14," Dan Niles, an analyst with the firm, said in a report. "We expect Q3 revs guidance of $6.9 to $7.5 billion (up 7 percent at the midpoint q/q) with a 1 percent increase in GMs q/q to 51 percent, while operating expenses increase very slightly q/q. As a result, we expect to only modestly increase our EPS estimates."

Lehman in June stated that it still had concerns for Intel, mostly based on the idea that the continuing build of inventory in Q2 around SARS concerns may end as fears subsided, cutting into Intel's MPU business from major customers like Dell Computer Corp.

Intel, too, was less than bullish. This year's Q2 mid-quarter call was the first in which the company did not make any changes versus prior guidance, saying its Intel Architecture business is trending toward the high-end of the normal seasonal patterns as stated in April with weakness in its flash/communications segment.

"We believe Intel expects normal seasonality driven by many trends including better corporate profits, increased Centrino adoption, good server business, a rebound in demand in Asia post SARS, and continued market share gains against AMD," Niles said. "Also flash revs, which were down 29 percent q/q in Q1 and were flattish in Q2, should start to grow solidly again in Q3 as cell phone demand improves post SARS."

Lehman expects no capex changes when Intel announces its Q2 results on Tuesday.

SZ -

So, when the corporation has profits and or significant positive cash flow, they take some of the excess money in the bank that is not needed to pay off humungous long term debt and interest payments on loans that were made to pay off other loans or chuaffeurs or body guards or private corporate jets for the Chairman of the Board - and they send a check to each shareholder.

What a remarkably irresponsible use of capitol. That money could be used to attack Intel. Why give it to shareholders who don't deserve it anyway?

MauiDude -

As long as you are the registered owner on the day that MS specifies as the dividend effective day you would be entitled to the dividend. Many people will be playing this arbitrage and it will be factored into the price.

wbmw -

http://news.com.com/2100-1006_3-1023436.html?tag=cd_mh

From the end of 1986 through the first quarter of this year, AMD's cumulative net earnings, including sell-offs and acquisition expenses, come to about $221 million, about the same amount of money Intel makes every three weeks. AMD is expected to lose money in the second quarter as well.

Q2 will put them at just about zero over the last 17 years. Jerry has walked away with far more money than AMD made.

To paraphrase that guy from Contact again:

"I would like to thank those from whom I have taken so much"

wbmw -

Look at the open interest on those Oct $9.00 calls. There's only 2 contracts open.

I sold some Aug $7.50 calls today for $0.40

morrowinder -

Not only is this initially third hand information and pure speculation

Sounds like Dan is raising his standards...

morrowinder

You failed to mention that the Xeon 4-way beats the Opteron 4-way.

blauboad

But they don't really have any other choice, do they?

Wasn't all this obvious when the decision was made to develop hammer in the first place? Add to that the extreme uncertainty around SOI and you have one of the most irresponsible decisions of the modern tech era. Once again the shareholders foot the bill for Jerry's vendetta.

Winsock -

I needed a good laugh so I went over to the SI Mod thread and read NiceGuy's posts.. The Sun could go supernova and he'd be convinced it was a great sign for AMD... What a character.

YB -

Obviously production means "experimental".

Development means experimental. Production means production.

DGanon819 -

I haven't been able to find anything on AMD paying IBM $40 million to fix a broken process.

AMD's 2002 Annual Report Form 10K, page 19, 3rd paragraph contains this sentence:

"In addition, research and development expenses in 2002 included a $42 million charge for amounts paid to a third party in exchange for product development services provided in connection with a discrete PC processor research and development project in the fourth quarter of 2002".

CJ -

So, anyone hear any recent rumors about Dell adopting AMD processors?

There's no reason why we can't start one!

Intel gets more key Alpha alums


By Michael Kanellos and Stephen Shankland
Staff Writer, CNET News.com
July 2, 2003, 4:34 PM PT


Pete Bannon, one of the key architects behind the touted Alpha processor, and a number of other Alpha engineers are joining Intel to work on future versions of Itanium.
Bannon and approximately 50 Alpha engineers will move from Hewlett-Packard to Intel this month, the company announced this week. The transition is the latest stage in a technological wagon train that began in 2001 as the result of a massive development agreement between the Santa Clara, Calif.-based chipmaker and Compaq Computer, which at the time controlled Alpha.

Around 300 former Alpha engineers already work at Intel. By the time the agreement is concluded, more than 450 are expected to move to Intel, company spokeswoman Barbara Grimes said. Most of the Alpha alumni are working on a version of the Itanium that will follow Montecito, a version of Itanium with two separate processors in a single piece of silicon coming in 2005.



Although the Alpha chip has never sold as well as other server chips, such as the UltraSparc family from Sun Microsystems, analysts and engineers have praised its performance. Digital Equipment first released the chip in the early 1990s. Compaq acquired Alpha in 1998 when it bought Digital, and HP became the owner when it bought Compaq.

The chip, and the Alpha development team, directly and indirectly influenced other major processors. The original Athlon chip from Advanced Micro Devices used a bus initially created for the Alpha. The new Opteron chip features a high-speed chip-to-chip interconnect called HyperTransport and an integrated memory controller--technologies similar to those touted earlier in Alpha. Dirk Meyer, AMD's top processor executive, worked on the Alpha.

Similarly, Hyperthreading, a technology in the Pentium 4 that lets a chip do multiple tasks simultaneously, was inspired in part by research performed at Digital first.

Brannon on Tuesday joined Intel as an "Intel fellow," the job title given to the company's deep thinkers. Like their counterparts at IBM, fellows at Intel sketch out strategic technological plans. While an employee at Compaq, Bannon was known for making disparaging comments about Itanium at the Microprocessor Forum, one of the big, annual events for semiconductor makers.

Bannon "will be leading a system interface architecture team for a future Itanium product," Grimes said. That processor will be released after Montecito.

Kevin Krewell, a senior editor at The Microprocessor Report, said it won't be surprising if Brannon and his team work on technologies similar to Hypertransport or integrating a memory controller. In June, Mike Fister, general manager of the Intel enterprise platform group, said the company is considering integrating a memory controller into its server chips and improving the chip interconnects.

The move for the Alpha team likely won't be difficult. Intel owns the building in Hudson, Mass., where they work. "It is a migration from one end of the building to another," Krewell said.

Joemoney -

Well, I not going to question your opinion.

What he said was not an opinion, it was a fact. Go to the SPEC website and look at the scores for yourself.

http://www.spec.org/osg/cpu2000/results/cpu2000.html

There you will find IBM's fastest scores and Intel's fastest scores. Judge for yourself which one is faster.

BTW, look under DELL for the current P4 3.2GHz scores.

B2L

I see your point and I guess it makes some sense. I would add that in this case, Intel is losing an 8 inch fab which may be of significance. All those 12inch fabs to make processors and where will the chipsets come from?

CJ -

Sorry but you're not convincing me that IBM needs AMD's help in process technology or design. AMD has shown a consistent inability to match their designs with the capability of their process. The apparent poor yields they have shown over the last few years are best explained, IMHO, by over aggressive targeting of transistor channels in the attempt to gain more frequency. The addition of SOI and 2 pipeline stages has resulted in a slowing down of Opteron relative to it's higher clocking predecessor Athlon. These are not the signs of a level of process expertise that IBM needs to help them improve PPC. If they want poor yields, long delays and slower products then AMD has paved the way.

wbmw -

CJ has been one of the more level headed posters here. I'd hate to see him lose it.

CJ -

And this isn't trivial. Despite the good point that Haddock made about IBM having excess fab capacity, I still don't see how IBM can deliver a 90nm 3GHz PPC970 without AMD's help.

Please, tell me you're not serious...

Sgolds -

Let's revisit in three months, it will be interesting to see updated Itanium and Opteron scores.

P4 could lead in 3 months...

Sgolds -

those SPEC scores - why is Itanium so mediocre?


Because HP's scores haven't been posted at SPEC yet.


http://www.intel.com/pressroom/archive/releases/20030630comp.htm

Itanium® 2 processor result of 1322 on SPECint_base2000* measured on HP Server rx2600 using Itanium® 2 processor 6M at 1.5GHz, HP-UX operating system and submitted to SPEC. SPECint* is a trademark of SPEC*.

Itanium® 2 processor result of 2119 on SPECfp_base2000* measured on HP Server rx2600 using Itanium® 2 processor 6M at 1.5GHz, RedHat Linux AS2.1 operating system and submitted to SPEC. SPECfp* is a trademark of SPEC*.

K -

A simple lease of a fully equipped fab is a possibility as well. Or in a more sophisticated sale and leaseback - structure. Or one of many many other possibilities. Which ones will be feasible and which one will be the structure of choice can hardly be judged from today's viewpoint. However, I agree in the core of your posting: IBM is confident in its development partner. Maybe even confident enough to leave the definitive agreement open until 2005.

I'm more skeptical myself. IBM must have a plan that considers the possibility of default. There is just too much chance of that fab being handed back to them in the future and IBM has to have a use for it that doesn't rely on AMD.