From uberpulse:
AMD explains why 16Core Barcelona got fragged by an 8C Clovertown:
"The objective of our demo was to show performance scaling from our current dual-core processors to our upcoming quad-core processors within the same thermal envelope and drive home the point through a real-world demonstration that customers could expect to see 2x the performance without an increase in power consumption. "
Sounds like disaster control to me. Assuming that AMD is telling the truth and did intend to show performance-per-watt scaling, then I don't really see the value of the demo. The fact still remains that Intel's V8 showed better performance using less silicon. It is great that K10 scales well against K8 but if Barcelona comes out as a non-performer, who cares? Nobody is interested in comparisons to an already beaten product.
The question AMD deperately needs to answer is why is it that the best they can come up with is a crippled Barcelona while Intel's Penryn looks stable and ready to go.
AMD also said: "One thing to note, the system we showed, while it was a 4P, it was running only 6GB of memory".
I'm sorry, this is really a lame excuse. Why cripple the benchmark? I never heard of anyone running a benchmark and not optimise the setup to get the full potential of the system. Or maybe AMD can't afford the extra memory due to cost cutting.
[thanks to the anonymous poster for the link:]
5.23.2007
AMD Explains the Fragging
Subscribe to:
Post Comments (Atom)
11 comments:
Go go go go go Intel.
While the most likely explanation is that AMD is in spin control as they rev the Si to address some issue (critical speed path? critical bug? power issues at higher clocks? 65nm process issue? etc), I don't think folks should dismiss the fact that the architecture could be good, but just not stable enough right now to demo.
If you look at the R600 debacle, and the fact that it appears to be getting more stable/better performance - maybe AMD doesn't want a repeat. It could be just a schedule issue with K10 and AMD trying to do spin control by releasing areas that they know show something (albeit they are doing a poor job about this) or it could be 65nm process issues (remember despite AMD's claims of things being on track, all AMD high end parts curiously remain on 90nm).
The other thing AMD could be doing is trying to lower expectations so that if they exceed the lower bar than people will look at it as a success. This is classic spinning at it's best say things are really bad, exceed them and then say "hey things are not so bad". (R600 was a classic ATTEMPT at this as the delay was due to desire to had launch, as opposed to the now obvious driver maturity issues)
Unfortunately AMD start trumpeting K10 so early with the 40% claims that I think it is too late for AMD to lower expectations and the press will see K10 as a failure/weak product unless it outperforms the top Intel chip by 40% - the press won't care about perf/watt, clock for clock, IPC, 2 to 4 core scaling... or nuances of specfp vs specint.
What is funny though is that the AMD fans have clearly lowered expectations (most saying if it's at least as good as Intel it's a success), as they know if they set the bar really low they will feel better if AMD exceeds it (rather than measure AMD against their own claims...).
AMD is in damage control mode and right now their marketing machines are in full SPIN cycle
They've pissed off the MOBO makers
They've set low expectations with the OEMs with respect to volumes
They've got Intel and NVDA angry and are out for blood
They didn't deliver anything substantial at the Microprocessor Forum (sorry, Griffin wasn't it)
The channel will get screwed again and will become a dumping ground for lower end of foodchain parts (K8 vintage)
Their employees are being kept in the dark. HR is looking at trimming benefits to essentials including reducing profit sharing contribution %, SPP plan participation limit and plan choices/coverage.
Who else is left to piss off?
How did this site manage to beat 16 cores with just 4GB?
AMD also said: "One thing to note, the system we showed, while it was a 4P, it was running only 6GB of memory".
Anonymous said...
How did this site manage to beat 16 cores with just 4GB?
it will be really funny when Intel come out and say:" actually we are only having 4GB in out test system ..." hahahahah
some one should post the link provided by anonymous to the same site!! :)
apparently that particular benchmark is not affected by system memory. After 4 gigs it doesnt matter how much more you put in the result will not make a drastic difference so AMD is pulling excuses out of their asses just like their fanbois.
heat
"apparently that particular benchmark is not affected by system memory."
Bingo!
Actually even 512M is more than enough for PovRay AND the OS, assuming it is not Vista.
Let AMD explain this one to the Federal Trade Commission
Whoops, ATI Radeon HD 2900 XT Lacks UVD
http://www.dailytech.com/Whoops+ATI+Radeon+HD+2900+XT+Lacks+UVD/article7447c.htm
I think it's time for another article on Dementia's, Uh, I mean Scientia's excellent analysis. His latest post on Q1'07, outside of straying a bit while talking about compilers is a bit "off":
He continues to spout this crap about AMD being only 6 months behind on 45nm? THIS IS NOT WHAT AMD STATED! Is he referring to the earth calendar? Based on this statement I would expect 45nm product ON THE MARKET (not shipping, sampling, PR bs, etc..) in early Q2? He has no facts, no links, no actual info to back this up. The only thing AMD has stated is they are set for manufacturing in "mid 2008" which could mean ANYTHING - sampling, shipping, wafers started in the fab, or actual product in the market. He also ignores the fact that AMD's 45nm process will be an early rev without much performance improvement (save power), so he is comparing Intel's finished process to the beginning of AMD's... when AMD is meeting 45nm roadmap specs and has product on the market based on this then you can accurately compare the time between the 2 companies.
"The simplest way to handle the question of bankruptcy is to look at AMD's history and assume that AMD can survive whatever it has already survived."
Huh? Simplest way if you are an idiot and have no financial background. He looks at ONE financial metric in isolation and concludes AMD is no worse off than previoulsy. To take this to the extreme if someone survives a lightning strike the next time they get hit you should assume (without looking at other peripheral circumstances) that they would survive again?
"For example, I've seen comment after comment about AMD's R600 delay yet no mention of Intel's 965 chipset being two months late"
Ahh.. attack the press... Was R600 only 2 months late, funny I thought it was more than this. Also I don't see comparing a new graphics card generation to a chipset revision? But heh, I'm biased too..(everyone is except for Dementia)
"However, other demonstrations suggest that AMD could hit 2.97Ghz in Q4 while Intel could pull 3.33Ghz Penryn into Q4."
This is absolutley hilarious - the AMD "demonstration" was an article by Fudzilla (and it was no demo, just some rumor that the new stepping could do this). Of course Intel DEMO'D the 3.33Ghz Penryn a while back so this would obviously seem to be the less likely of the two assumptions to be true.... no bias here...
"So, AMD could potentially improve its volume and revenue position substantially by end of 2008"
Wow talk about lowering expectations, just one quarter ago Dementia was saying how AMD was MUCH BETTER POSITIONED than Intel going into 2007! Now he thinks, maybe, potentially AMD could improve substantially by end of 2008!
BTW - here's some excerpts from his Q4 "analysis":
"AMD's margins should be up in Q1 and their channel distribution should be fine." (OOPS!)
"Likewise FAB 36 is converting to 65nm and this too reduces cost. So, by continuing with the same strategy AMD could expect to have better profits in Q1 07 although they may not quite match the Q3 06 numbers"
"Presumably, AMD could introduce Barcelona in Q2 with a higher price tag" (Q2? I guess if "introduce" means you are not actually selling any?)
"However, overall I think it is most likely that AMD will continue to nibble at Intel's share each quarter"
"AMD however will have new offerings for server, desktop, and mobile" (Mobile? Last I heard was mid-2008 for mobile, unless he is considering a 65nm product a new offering? Desktop will also be pretty minimal in 2007)
"Intel does appear to have gained back 0.8% of revenue share. However, with the boost in volume in Q4 this is probably only a temporary change."
Does anyone know when Scientia start guzzling the AMD kool-aid? I mean he always was an AMD fan, but until recently he seemed a little more objective...
It will sucketh at AMD's teat until the teat runs out of juice.
Ahh... more Dementia (Scientia) logic and math comparing Intel and AMD capacity (the first part refers to 2005):
"Since a 300mm FAB is 2X the capacity of a 200mm FAB this put Intel at 6X AMD's capacity.
Two 45nm x 2 = 4
Four 65nm + 4 = 8.
8 / 2 = 4
This would suggest that Intel will have a smaller production advantage than they had in 2005. How is that an improvment?"
Anyone else see the lunacy of this statement / "analysis" (and I use this term VERY loosely)? IT IGNORES WAFER STARTS COMPLETELY and ASSUMES (INCORRECTLY) ALL FABS ARE THE SAME SIZE!
To put this n perspective one of Intel's HVM fabs (like NM) does over 6500wafer starts per week. AMD F36 is ramping (at least right now) to ~20K wafer starts per month.
If we take the low end of 6500 WSPW this is a little over 28000 WSPM (before someone tries to hammer me on the math I'm assuming ~4.3wks/month. Of course this is the low end for Intel's large fabs.
He also ignores the other 2 45nm fabs. He assumes only 2 based on current work underway yet ignores the fact that F30 will hardly be a full ramped 65nm fab by end of the year (per AMD analyst foils on their own website) This guy is getting crazier and crazier, I hope K10 comes out and does reasonably well so he doesn't completely lose it trying to defend AMD at all costs...
Of course there's also yield which noone really knows who's better so it will have to be assumed to be a wash in this example for lack of any real data.
Post a Comment