Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Keith -
The datasheet gives 85W typical and 101W max, however, as the 3,06 uses 1,525V instead of 1,500V (according to the datasheet), the max power is actually higher than 101W for the 3,06 Ghz, despite the prescribed adaptive voltage regulation.
The article is incorrect. While I haven't looked at the new datasheet since yesterday, I believe it is the same Vcc as the P4 which is not 1.525V but 1.425V. You are quoting the VTT number, not the VCC number. If you have a Intel reference to a Vcc greater then 1.425V then show us. Also max Icc is 65.4A so Max power is 93.195 Watts, not the 101W claimed.
[Edit: I just looked at the new Xeon datasheet and the Vcc has risen to 1.467V. Icc_max is now 69.1 giving the 101W posted, which was correct]
EP
Paul -
re: Elmer, wbmw, SemiconEng
It would be great if we could remember to keep things impersonal and unemotional with them.
Don't worry, I can take a few insults. When you come on an AMD board and dare to say the emperor has no cloths you get used to personal attacks. I just draw the line when it comes to falsifying my statements. I don't think I'm being unreasonable.
EP
Klaus -
As for "proven unmanufacturable", not yet
Please quote me correctly. I said "so far, unmanufacturable".
EP
Keith -
This will be the last post from me to you.
Good! Then I won't have to worry about you falsifying my statements again. I will however feel free to take your statements, edit them to suit my needs and them claim them as your statements. Fair enough?
EP
Keith -
the design has proven to be, so far, unmanufacturable
That's right, that was my statement. Here's what you attributed to me in quotes;
"has proven to be unmanufacturable"
You took my statment, cut out the operative words and claimed it was mine. So you clearly misquoted me. That's dishonest and you should be ashamed of yourself.
It seems to me you are totally lacking the ability to admit mistakes, and rather choose the route of subtle personal attacks like the one above to somehow save yourself.
You have demonstrated you are willing to edit someone's statements to suit your needs, claim they made the edited statement and then criticize them for what you have falsely presented.
If you are going to quote me fine, quote me but don't lie about it by editing my statements to change the meaning just to suit your agenda.
Spokeshave -
How is that not suspension of belief in its purest sense?
How many clues do people have to put right under your nose before you can figure that out?
then I submit that you just violated your NDA by clearly intimating that, based on your confidential information, Itanic sales were greater than the Bloomberg article stated.
Getting a little overly dramatic aren't you?
Spokeshave -
Pick a different FUD buzzword. This one is no good anymore.
This word has a definition in the industry. The fact that you don't know that is not a reflection on me.
Spokeshave -
Bull. You don't believe that for an instant - at least not in general terms. If you did, then you would have objected vehemently when wbmw claimed that he did not believe the Bloomberg data on Itanic, and that he believed that Itanic 2 was doing far better than anyone thought.
Once again your reasoning is faulty. I happen to agree with wbmw on that matter and for the same reason. He didn't disclose his reason and I won't disclose mine but I don't believe the Bloomberg data. I'm not trying to persuade you because I know that's hopeless. I'm just stating my position. I agree with wbmw on that one.
Dan3 -
As you once again demonstrate that you have no idea whatsoever what you are talking about, I am reminded of why I "ignored" you on SI. You have absolutely no respect for accuracy. You make things up ad lib to suit your needs. I'm just not going to waste my time correcting your nonsense.
BYe, you're back on ignore.
Keith -
Just because you don´t doesn´t mean we don´t.
I suspect that you wouldn't even know what features were turned off. As long as they were fully operational you wouldn't know the difference.
While it is nice to see you can make up definitions to somehow fit your argument, to say that a product that hasn´t even been launched yet "has proven to be unmanufacturable" is, especially with your definition, absurd.
Please show us where I made this statement you are attributing to me. The only reference I can find is a statement you made. Are you quoting yourself and criticizing me for your own statements? That's dishonest. Don't go that route because nobody wins.
Spokeshave -
How is it possible that you have such a deep understanding of the confidence and expectations of all who follow AMD? By what magic have you gleaned all of our innermost thoughts? Please share you great and powerful secrets with us mere mortals.
I can only go by the posts the AMD faithful make here. The suspension of disbelief is not healthy for an open mind.
I hereby decree that there will be no further discussion of anything that *might* happen.
Discussions of what might come is a good thing. What is not good is to accept, as a foregone conclusion, the performance of a company that has heretofore failed to perform. Failure is not a crime and it is no guarantee of future failure. AMD could perform in the future, it is not impossible. However, a reasonable person must allow their expectations to be biased by past performance. Failure to do so is intellectual dishonesty.
Dang. I can't believe that I typed that with a straight face.
I don't believe you did.
Spokeshave -
You clearly did not read the post. Klaus was talking about the approach. An approach is not a product, and is not manufacturable
You clearly missed the point.
Your logic is ludicrous. By the same line of reasoning, it would make no sense to look at the weather forcast for tomorrow. After all, if it is not raining today, why should we consider the possibility that it will rain tomorrow?
I should start charging for straightening you out. Of course we should listed to the weather forecast as long as they maintain a reasonable accuracy rate. You guys just listen to the same forecasts that say tomorrow will be clear, warm and sunny but it keeps raining day after day. They don't recalibrate their confidence in tomorrow's forecast based on poor past performance. They just keep believing the new forecasts with no recollection of the past. That is in my mind the sign of very poor judgement on your(collectively) part. There are a couple here who have noticed that the forecasts keep turning out wrong (CJ comes to mind but there are others) but most others have not changed your expectations one tiny bit despite the many delays and disappointments. That's dumb in my view and the sign of a faith based belief system.
Keith -
AMD already sold [up to] 500 1U 2P development machines which are fully functional, with CPU´s running at launch speed (that would make it a thousand Opterons).
AMD sold systems that have to be considered engineering samples and while they may have been fully operational, we don't know they were fully functional. And what frequency do those run at anyway?
and to say "Opteron is unmanufacturable" also doesn´t seem quite right
Do I need to provide a definition of manufacturable? It's something that can be produced at a profit, and it usually requires some repeatability and meeting of schedules. 1 die per wafer sold for $5000 could make the case that it is manufacturable but I don't think that's likely the case.
It´s only one more month, and then we´ll see how the story goes.
No doubt there'll be an announcement but what else remains to be seen.
EP
Paul -
It won't be long (later this year)before Opteron has surpassed Itanium's combined total volume in 2 generations.
Once again, this has proven much easier to say than to do. So far AMD hasn't shipped a single unit nor demonstrated the ability to do so, but you already have them shipping volume. No wonder we call AMD a faith based system. As for Itanium volume, that is unknown and wide open to interpretation. 64 processor systems are shipping today and larger ones will be shipping later this year.
kpf -
What convinced me from the very first moment of the idea is its evolutionary approach of transition.
In other words you believed the hype. As I've said many times on many boards, a product is the marraige between a design and a process and the design has proven to be, so far, unmanufacturable. If I may offer some advice, pay more attention to what AMD can actually produce and less attention to the hype.
EP
sgolds -
It is an evolutionary growth path for x86, a well proven technology. Early indications are that 64-bit optimized for Hammer gets a 30% boost over 32-bit code. (That is more than I would have predicted.)
After slipping almost 1.5 years it's hard to claim it's well proven technology. Why in the world did it take so long if it's so well proven? BTW, the "early indications" are that Opteron doesn't meet the design targets. Perhaps we should wait (and wait and wait) for some real data?
EP
sgolds -
Opteron will be at 3GHz on 90nm, next year. Should not be anywhere near 90W. That is one of the advantages of making a performance chip, vs. an inefficiently fast chip. We match or exceed Intel performance at much lower power usage, and lower heat dissipation.
At the risk of being redundant, you have yet to demonstrate that you can produce Opteron on any technology and your recent delivery to schedule performance leaves customers with little confidence that your promises of 90nm next year can be relied on. Deliver something, anything, in volume and then make claims about future products. Right now, these promises are 2 generations out.
EP
kpf -
You sure know a lot more about PowerPC than me. However, what I know is the architecture was done quite a while ago. Today, neither MOT, IBM nor anybody elso would make any thought of designing and manufactoring a leading-edge CPU based on definitions of Apple, you bet - unless Apple is capable and willing to carry the cost for it. Apple can only buy whatever they basically already have (which is good enough - Apple can still make USP-products with what they get).
PPC was designed by IBM, Moto and Apple in the early 90s. It was considered the newest and greatest design because it combined the genius of those 3 stellar industry leaders. x86 was condemmed as an aging piece of crap by comparison. Unfortunately the silicon didn't quite live up to the foils and the PPC is an embarrassment to those who praised it's glories. Perhaps we will see a replay with Hammer?
EP
Dew -
You pose some good questions. All I will offer is the cliché that the darkest time is just before dawn.
A truism no doubt but you don't know if it was the darkest yet until you see the dawn.
EP
Edgar -
re:Xeon at 3 Ghz, what Intel is not telling clients is that it's at around 90 watts. Put some of those in a server and your running 1 Kilowatt an hour.
It's more like 81.8 watts typical and 81.8 watts for 1 hour is 81.8 watts per hour per processor. An Athlon MP will save a few watts per hour and perform much worse as measured by SPEC and SQL benchmarks recently posted here.
EP
Klaus -
In contrary to SUN, Apple is not designing and producing CPUs
Apple played a big role in defining the PowerPC architecture.
EP
Intel® Xeon™ Processors Speed Past 3 GHz
SANTA CLARA, Calif., March 10, 2003 - Intel Corporation today introduced its fastest Intel® Xeon™ processors, exceeding 3 GHz, for dual processor servers and workstations. The company also said that more than 9,000 companies around the world are now selling components or systems based on the Intel Xeon processor family.
The Intel Xeon processor at 3.06 GHz features a 512 KB level two cache and a 533 MHz system bus and the Intel Xeon processor at 3 GHz includes a 512 KB level two cache and a 400 MHz system bus.
"Intel continues to raise the bar on system performance and manufacturing execution by delivering the first processors to exceed 3 GHz for dual processor servers and workstations to system vendors worldwide," said Richard Dracott, director of marketing for Intel's Enterprise Platforms Group. "In 2003, we will continue to enhance the Intel Xeon processor family with industry leading performance and scalability."
Systems based on Intel Xeon processors are used as general-purpose servers for Web hosting, data caching, search engines, security and streaming media, and as workstations for digital content creation, mechanical and electrical design, financial analysis, and 3D modeling.
The Intel Xeon processor at 3.06 GHz is drop-in compatible with existing systems designed with the Intel® E7501 or Intel® E7505 chipsets, which have shipped in volume since November 2002, while the Intel Xeon processor at 3 GHz is drop-in compatible with systems designed with the Intel® E7500 or Intel® 860 chipsets.
The Intel Xeon processor at 3.06 GHz and the Intel Xeon processor at 3 GHz are available worldwide. Intel's suggested list prices are $722 and $658 respectively in 1,000-unit quantities.
Intel, the world's largest chip maker, is also a leading manufacturer of computer, networking and communications products. Additional information about Intel is available at www.intel.com/pressroom.
Intel® Xeon™ Processors Speed Past 3 GHz
SANTA CLARA, Calif., March 10, 2003 - Intel Corporation today introduced its fastest Intel® Xeon™ processors, exceeding 3 GHz, for dual processor servers and workstations. The company also said that more than 9,000 companies around the world are now selling components or systems based on the Intel Xeon processor family.
The Intel Xeon processor at 3.06 GHz features a 512 KB level two cache and a 533 MHz system bus and the Intel Xeon processor at 3 GHz includes a 512 KB level two cache and a 400 MHz system bus.
"Intel continues to raise the bar on system performance and manufacturing execution by delivering the first processors to exceed 3 GHz for dual processor servers and workstations to system vendors worldwide," said Richard Dracott, director of marketing for Intel's Enterprise Platforms Group. "In 2003, we will continue to enhance the Intel Xeon processor family with industry leading performance and scalability."
Systems based on Intel Xeon processors are used as general-purpose servers for Web hosting, data caching, search engines, security and streaming media, and as workstations for digital content creation, mechanical and electrical design, financial analysis, and 3D modeling.
The Intel Xeon processor at 3.06 GHz is drop-in compatible with existing systems designed with the Intel® E7501 or Intel® E7505 chipsets, which have shipped in volume since November 2002, while the Intel Xeon processor at 3 GHz is drop-in compatible with systems designed with the Intel® E7500 or Intel® 860 chipsets.
The Intel Xeon processor at 3.06 GHz and the Intel Xeon processor at 3 GHz are available worldwide. Intel's suggested list prices are $722 and $658 respectively in 1,000-unit quantities.
Intel, the world's largest chip maker, is also a leading manufacturer of computer, networking and communications products. Additional information about Intel is available at www.intel.com/pressroom.
belgiangenius
Does anyone think we're going to find our way out of this $5 mess we're in?...Any comments?
Why not write some Covered Calls?
YB -
Actually I buy that 25,000 Itaniums will be sold this year. Last year it was 2,000 servers and if there is 4 cpu in average, then it was 8,000 Itaniums. So this year it will be 4,000 servers and 16,000 cpus. Maybe someone will assemble a 10,000 node cluster and it will be 26,000 Itaniums total. Fine.
If you are in any way basing this discussion on the article in the Inquirer you should at least be consistent and acknowledge that the article said 25,000 Servers. Some of those Servers will be 64-Processors and more. The 64-Processor systems are already released and shipping. Larger systems will appear throughout the year.
Klaus -
What would you understand from the (semiconductor language) term "begin producing during the second halt of 2003"?
This is the same company that planned to announce Hammer in 2001...
EP
Edgar -
I have been researching and trading AMD for going on 9 years now and Intel always leaks news spam to the media just before AMD is about to release a new "great" product. How do I know that? Because if you read a lot you notice a greater frequency of Intel related stories concerning,Intel is greater and Intel is better just before those releases.
While everyone is entitled to an opinion I think this is utter nonsense. You have no data, no links and no evidence. You only have your feelings.
By the way get ready because here comes Opteron.
Where have I heard that one before?
sgolds -
This reinforces my view that Barrett is the very worst thing that could have happened to Intel, he certainly is no Moore!
Did you happen to notice that there's a worldwide recession in technology? During this time Barrett has kept Intel profitable. I think far too much blame has been dumped on Dr Barrett for things beyond anyone's control.
CJ -
Nope, they won't. Primarily because it was so darned dumb.
Yes it was. Nobody argues that because there's no other way to spin it. Just plain stupid.
BTW, do you ever hear the name Albert Yu anymore?
EP
sgolds -
Reliability testing - to some extent, it is a useful fiction. Useful in that different products tested with similar methodology can be compared with each other. A fiction because lab conditions, despite attempts to accelerate wear, do not come close to a good model of the outside world
The only real metric is Customer perception, however accelerating reliability problems by means of burnin correlates very well with early life failure. Test methods are constantly improving and long term reliability also correlates well with fault coverage. Humidity and temp cycling experiments are required for package qualification. Board testing and quality is a field I don't know so I won't go there but I would think that a server manufacturer can specify temperature and humidity requirements and hopefully minimize the use of salt...
Nevertheless, the process is not perfect nor would anybody be willing to pay for an absolute guarantee if one existed.
EP
Economaniac -
Reliability testing, Comb probably can speak better to this but my understanding is that semi co's guarantee that no more than a certain percentage will fail over a fixed period of time subject to certain conditions- ie less than .5% failure over 20 years at -20-120C. For consumer products the manufacturer can then calculate the total probability of failure over the warrantied product life.
A commonly used term is DPM, or Defects Per Million. It is broken down into 2 basic parts, quality and reliability. Quality means how many dead soldiers do you have at time zero after shipping to the customer. Reliability means how many die prematurly during the useable life of the product. Cosmetic defects fall into the Quality category. I would expect that AMD is not far behind Intel in these metrics although I have no data. For Desktop products a level of <500 DPM (Total of Quality and Reliability failures) is probably what Intel and AMD are shipping however Servers may require an even lower defecr rate. I have seen DPM requirements drop into the low single digits for some applications where failures are extremely problematic. How these levels are achieved is a science unto itself and where my expertize lies.
EP
Paul
How long will it take them to do this, and how is it done? Is it actually physically/electronically done to a large number of chips, or is there some sort or thoeretical extrapolation done?
Yes to both. Experiments are performed on a sample of units large enough to product data that is statistically significant. No one can absolutely guarantee any individual device (except to the extent that they can promise to give you your money back if it fails). What they can do is demonstrate to a customer that a sample base of x units showed no more that y failures under the conditions of the experiment. That gives a confidence level that the customer can rely on and predict.
EP
sgolds -
how many processors do you think are in each Itanium server shipped?
I really don't know and I don't want to speculate. I just wanted to introduce the fact that the number could b higher than 25K.
EP
Edgarcayce -
Intel to sell 25000 Itaniums this year,LOL.
I have no idea if the article is true or not, but it quotes 25,000 servers. Seeing as some of those will be 64 processors or more, you may want to stop laughing so hard...
EP
FNF -
ephud Intel did have problem with .13
Remember the 1.13GHZ P3 which had to be recalled.
That was not a .13u product and it was not an example of a process problem. It was an example of pushing a process too far.
Also Intel had problems getting volume out initially which is well documented.
Well Documented? You've got to be kidding.
wbmw -
So on this set of industry standard benchmarks The Athlon MP "2600+" scores below a 2.2GHz Xeon in Integer and below a 1.5GHz Xeon in FP. Combining this with the poor showing in SQL tests posted recently and it's clear that AMD isn't going to make much headway without a competitive Hammer.
Jack -
) to/too/two
An example- I'm going to the store too, we'll get two.
2) they're/there/theirs
An example- They're over there and it might be theirs.
How about your/you're?
Chris -
You certainly commented greatly on it at the time when it was referenced on the RB board. Of course, it's hard to call your attention to it as your subject line rarely includes a subject, just a name. (Is this standard practice for FUDsters, so they can't later be called to task on their gloomy predictions?)
I think you are making this up. Show me a reference to my commenting on this article, or is this just standard FUD from a champion like you?
sgolds -
delays at Intel? Other than a single rumor from theRegister, what evidence is there of an Intel delay? Don't take that to the bank yet!
I didn't even see the rumor.
Semi -
I believe I read quite awhile back that intel had functional transistors on 0.09u on 300mm, so it seems to me that 0.13u can no longer be called intel's "cutting edge" process. It may not be a candidate for 300mm at all.
Huh? I certainly hope Intel has functional transistors on 90nm on 300mm wafers. I wouldn't want to think Prescott was done with discrete components. And you think 130nm might not be a candidate for 300mm? What are those 300mm fabs doing right now?
EP