7.03.2007

AMD's Performance Promises - Theo's Source

Theo posted an article in the Inquirer that detailed AMD's promise of a superior Barcelona performance at 2.3GHz. I thought this was old news and was the primary reason why I never bothered posting it. But before you accuse Theo of pulling out numbers from his excretory opening, let me be the first to say that he isn't. You could probably accuse him of recycling news because first of all his numbers comes from a predated AMD slide and secondly, they include a slide title that says "Performance Projections".



If you wish to know what the number means, you can go ahead and compare them with Intel's benchmarks here: Intel® Xeon® Processor. On paper, Barcelona holds up very well against Intel's Clovertown exceeding it in most cases. But the big story now is when will we see Barcelona with speeds of 2.3Ghz and above.
I seriously caution against getting too excited about the NDA slides. Barcelona has gone through so many iterations that projections made before them can easily go either way.
Of course, Thank You's go to the anonymous gutter rats for the tip.

!-- Update: DailyTech has started joining in the fray and AMD's "simulated benchmark" mess has snowballed. It looks like
Kristopher Kubicki agrees with George Ou. Another bummer for Scientia since his long list of Intel pumpers is about to get longer. Come to think of it, Phil Hughes should be included in the list since he appears to be contradicting Scientia. "We are working to remove that stuff from our website now. It isn't an accurate reflection of the highest performance [Intel processors]," Hughes acknowledged.

The key point isn't whether the slides were done in April before AMD's 2Ghz announcement, but that they were distributed last week to journalist. After losing the performance crown and large sums of money, it appears like losing face is something AMD is willing to depart from.

25 comments:

Anonymous said...

Theo's numbers for the specrate are wrong. That is a 2.6GHz Barcelona that we've already seen.

For the others, I'm not sure how you came away with the impression that Barcelona "holds up very well", "exceeding it in most cases".

For TPC-C, let's say an average of 60%. K8 scores 140K. 1.6x that is 224K. 224K is less than Clovey's 240K.

No comment on the obviously retired benches.

K8 scores 97K on specjava test. 1.62x that is 157K for K10. 157K is less than Clovey's 225K.

K8 scores 15K in specweb. 2x that is 30K. Finally, beating Clovey by 50%! Too bad K10 won't clock 2.3GHz until Christmas.

K8 is 1.1K in SAP-SD. 1.7x that is 1870. Matches Clovertown, would lose if I didn't round.

Anonymous said...

BTW, the situation turns from dire in 2P to desirous in 4P compared to Tulsa. Too bad Tigerton is going to roar.

Anonymous said...

Always great to see some "detailed" data on chips that have absolutely no launch schedule, yet a complete lack of details on chips that will be launched (1.9&2.0GHz's).

Bait and switch anyone? Look we project can beat Intel just like we said; we can't sell you any of these right now but you probably want these more "energy efficient" chips anyway... benchmarks on those? estimates? Well that's not important right now...did I mention they are ENERGY EFFICIENT!

What you really need to do is upgrade to the energy efficient models for 6-9months and then UPGRADE again to out super-duper model later...Did I mention they are drop-ins? Meaning you only need to drop some money (several times) to get the performance we promised...

Anonymous said...

BTW anyway want to ask how much of the "energy efficiency" of the K10 design is actual design related?

If you look at the drop in Vt's moving from 90nm to 65nm, couple that with a serious decrease in clockspeed, and how much of the power benefits is due to design again? Outside of independent clocking of cores is it really more power efficient?

Anyone yonder why AMD didn't scale Opteron to 65nm?
A) 65nm process is not as good as they are claiming it to be?
B) You would get significant active power reductions (through lower Vt, Vcore), which may make the benefit of moving to K10 a bit less impressive?
C) AMD has problems with getting out product shrinks quickly? (see Athlon)
D) They seriously overestimate how quickly they could launch K10?

For a company that promotes energy efficiency and perfomance per watt it seems odd that they didn't move their flagship product for the last 3 years to a lower power process...which would only enhance this "lead" / "philosophy"

Anonymous said...

This is all preposterous ‘Power Point’ paper positioning. Again, we’re looking for clues at the scene of the crime. We have been, like so many others, bandying this thing back and forth for OVER 2 quarters, HELLO! The fact of the mater is the damned thing ain’t even out the door, yet. AMD is not beating Clovertown or Penryn, for that mater. The only thing their doing is beating their meat.

More importantly, with ALL the previously discussed partners and hardware manufacturers’ respective shortfalls, there is no product. That means NO foreseeable revenue in 2007, none, squat, nada, zip! How the hell they can stay afloat even if they meet their entire Power Point pissing pump by December is beyond me. THEY NEED MOOLA, ------------- BIGTIME! This thing ain’t gonna deliver it. WORD!

SPARKS

Anonymous said...

LOL - the only saving grace is that this all happened on or after iPod Friday and right before a major US Holiday.

Why hasn't AMD scheduled their Q2 Earnings call yet? What are they waiting for?

Someone should show up at the special shareholder's meeting coming up and ask Hector WTF?

Maybe Paul Otellini should buy 1 share of AMD and go ask WTF?

Roborat, Ph.D said...

anonymous said: "Bait and switch anyone? Look we project can beat Intel just like we said; we can't sell you any of these right now but you probably want these more "energy efficient" chips anyway."

i'm thinking Sun and Cray must be absolutely gutted with AMD' execution.

Ahmar Abbasi said...

http://blogs.zdnet.com/Ou/?p=567

Check this out roborat great info on AMD's projected numbers on their website of a 2.6GHz barcelona which does not exist....

Anonymous said...

Roborat, how does "Barcelona holds up very well against Intel's Clovertown exceeding it in most cases"?

Roborat, Ph.D said...

anonymous asked: Roborat, how does "Barcelona holds up very well against Intel's Clovertown exceeding it in most cases"?

they don't! I made a mistake. Please disregard that statement as I was unfortunately reading AMD slides that were "blatantly misleading". :(

Anonymous said...

Poor thing, you fell into the evil power of the powerpoint! Your next post should say "I got spun by AMD. Really, true power of K10;useless in int_rate, useless in database, uselss in java, useless in ERP, AMD desperate that they have to bench on old retired benches, amd only good in fp_rate and SPECweb2005_Ecommerce" Klovertown kills.

Anonymous said...

Doc,

“i'm thinking Sun and Cray must be absolutely gutted with AMD' execution.”

After reading this, I am compelled to respond, after a bit of thought. The ramifications, as you well know, go seriously beyond SUN and CRAY. Motherboard makers, system builders, and any company who spent hundreds of millions investing in wild-eyed speculation and PP launches must be absolutely furious with AMD.

Secondly, the entire IT industry was waiting patiently for a ‘drop in’ upgrade to their entire infrastructure which is now being rendered obsolete day by day. They are still waiting as Intel trounces and ‘tic tocks’ AMD to the pavement. Corporate big shots who INVESTED millions on AMD platforms on AMD’s ‘one hit wonder’ will not take the heat for second a second tier product, again. They will blame some poor bastard lower in the food chain for going with the “Scrappy Little Company”, based on the poor slob’s best recommendations. READ: Not so cheap in the long term, eh? Heads will roll when entire systems are gutted. Sure AMD turned heads last year, but somewhere, someone’s going to ask “How the hell did we put our eggs in this basket?”, and “We should have stayed with INTEL.” If I were that poor slob, or the corporate Bigwig, you could bet your ass I wouldn’t do that again.

Third, AMD’s ascension to the serious player forefront has cut an unprecedented swath of destruction in one year. Everyone, from the channel, to the enthusiast’s, to the manufactures, and HPC, have felt the pain as AMD burned every goddamned bridge they crossed. Would you trust these guy’s again? Not with my frigg’en money you don’t. Unless, of course, we can get BARGAIN BASEMENT PRICES on working, available product! That’s the rub, isn’t it? AMD can’t make any money. It always has been.

Smart players like DELL, HP, Toshiba, and Sun saw the hand writing on the wall. They hedged their bets. Apple, smartest of all, didn’t even consider AMD in the line up. Look how well they’re doing! You can’t smoke and mirror Steve Jobs. He’s been on this roller coaster much too long.

The bottom line here is performance in an industry is one thing, consistent performance AND credibility is quite another. This ridiculous AMD side show ascension has revealed every flaw we all knew existed, poor manufacturing and poor execution. I believe, at this juncture, this HUGE eight ball will be impossible for AMD to get around, now, and in the foreseeable future, despite all the hype.

Finally, there are the investors. God help them.

SPARKS

Anonymous said...

More fuel to fan the fire

http://www.pcdoctor-guide.com/wordpress/?p=4394

Anonymous said...

Scientia: I'm just amazed that someone could actually argue that it is fair to compare a quadcore against a dualcore after complaining that AMD's graphs were outdated. That is insane.

Is it just me or is he stupid?

Anonymous said...

Scientia: I'm just amazed that someone could actually argue that it is fair to compare a quadcore against a dualcore after complaining that AMD's graphs were outdated. That is insane.

First of all, this isn't Scientia's blog so who are you addressing yourself to?

Secondly, Intel is shipping quad core today and AMD is shipping vapor.

So comparing the best of AMD (dual) vs the best of Intel's (quad) is fair and valid - and the price points at which the Intel quads are going to be selling soon will support it even more.

I think Scientia and Sharikou are busy getting Paxil refills or have been frantically trying to get a hold of the stockbroker who sold them on AMD stock and has now disappeared to the Bahamas :)

Anonymous said...

Welcome to 2007: "Simulated Benchmarks"

http://www.dailytech.com/article.aspx?newsid=7927

Like fine wine, this story gets better with time ;)

Anonymous said...

What's funny is the AMD fanboys actually read right past the point of the article (AMD is using deceptive bench'market'ing which is ironic as AMD cried foul about this very thing) and saying that Ou, et. al are conveniently ignoring the specfp charts because they show more of a lead...

They just don't get it, the whole point of the article/blog was not another pissing contest on who's got the best chip. It was an article about AMD ESTIMATING performance of a product that will not be part of the actual initial launch and comparing it to a lower tier of the competitor products.

It is rather clear to anyone with half a brain what AMD is trying to do and how hypocritical it is with respect to their previous whining about this very thing.

Hey look the 2.6GHz chip is superior (*we think/estimate*), so buy our much lower clocked part which we will conveniently not include in the comparison. Oh and ignore the higher bin part from our competitor and please also ignore the fact that the competitor's data point is no longer up to date - it's rather difficult and time consuming to update a single data point on an excel graph, so we'll let the readers do that for themselves. After all the #'s change all the time anyway and it's not like their PUBLIC domain, how can we be held responsible for keeping a comparison up to date?

BAIT and SWITCH - look at the 2.6GHz but buy the only chip that is available the 2.0GHz. The 2.6 is better so why wouldn't the 2.0 be better also?

This is done all of the time - let's put an item on sale, but make sure the stocking level is really low / non-existant (Circuit City, Best Buy anyone?) - and we'll get people in the store and since we are out of that product we'll try to sell them another product...

Anonymous said...

It is rather clear to anyone with half a brain what AMD is trying to do and how hypocritical it is with respect to their previous whining about this very thing.

Indeed - and the crybabies will come out of the woodwork to declare how AMD would not be in this position were it not for the "Intel monopolist" blah blah blah.

In the meantime, hundreds of loyal AMD employees await the inevitable restructuring that is to come due in no part due to the overly ambitious and meglomaniac leadership that has run this company into the ground.

Anonymous said...

Cleaning up my last post a bit.

It is rather clear to anyone with half a brain what AMD is trying to do and how hypocritical it is with respect to their previous whining about this very thing.

Indeed - and the crybabies will come out of the woodwork to declare how AMD would not be in this position were it not for the "Intel monopolist" blah blah blah.

In the meantime, hundreds of loyal AMD employees await the inevitable restructuring that is to come due to the overly ambitious and meglomaniac leadership that has run this company into the ground.

Anonymous said...

"AMD promises honest Barcelona benchmarks"
http://www.vnunet.com/vnunet/news/2193610/amd-promises-honest-barcelona

'The AMD web page doesn't list a publication date, but AMD spokesperson Phil Hughes said the information was posted early this year and that it was accurate at the time of posting.

"We are working to remove that stuff from our website now. It isn't an accurate reflection of the highest performance [Intel processors]," Hughes acknowledged to vnunet.com.'

The funny part? They "are working on removing that stuff from our website" How long does it take to pull the info? Do they ot even have a summer intern who is capable of updating a web page? (I assume the rest of AMD is furiously trying to squeeze clockspeed out of K10)

My guess is that it will take just long enough to make up another benchmark?!?!?

"AMD's Hughes promised that the company would post performance benchmark data for its 2.0GHz chips "in the coming weeks", at which time it would also use the most recent benchmarks on Intel's chips."

Apparently AMD does not have the technical expertise to take the info down yet as of 9:38pm PDT, 7/5/07:

http://multicore.amd.com/us-en/AMD-Multi-Core/Products/Barcelona/Performance.aspx

but hey they're "working on it"....

Anonymous said...

Off topic, dumb question... of course there are no such things as dumb questions, just dumb people who ask those question so feel free to rag on me!

Scientia keeps claiming Intel will have no answer for DTX and/or D or DTX will be a significant advantage for AMD.... Is the DTX form factor proprietary to AMD? Is there anything to prevent MOBO manufacturers from making DTX boards with a 775 socket in order to use Intel chips if the market starts accepting/flourishing with DTX?

If not is there anything innate to the AMD chips that makes only those chips capable of working on a DTX form factor (my gut feeling on this is that if the socket was modified for an Intel chip there would be no other major differences, no?) Sure there is a time to market advantage but is there anything other than that?

anonymous said...

Is the DTX form factor proprietary to AMD? Is there anything to prevent MOBO manufacturers from making DTX boards with a 775 socket in order to use Intel chips if the market starts accepting/flourishing with DTX?


I posted exactly that on Scientia's site a few threads back (1 or 2). My argument is that a mobo form factor is not defensible technology. If it will influence sales or improve margins, mobo manufacturers will make Intel versions simply because it will have a bigger impact on their financials. Not rocket science...

Anonymous said...

It appears that Scientia is changing his tone. Touching writetup in the beginning about his wife.

On The Quality Of Things I've Said

http://scientiasblog.blogspot.com/2007/07/on-quality-of-things-ive-said.html

Is Sharikou next?

Anonymous said...

So is his sympathy towards his wife now empathized towards AMD?

Anonymous said...

So is his sympathy towards his wife now empathized towards AMD?

Absolutely not