Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Ali Chen who wrote (227028)	2/28/2007 5:32:17 PM
From: Biomaven	Read Replies (2) \| Respond to of 275872

for SPECint2006: spec.org - 18.5 base spec.org - 11.9 base -> 55% For FP, the best at the date results are: spec.org - 17.1 for Xeon, and spec.org - 12.7 for AMD -> 35% Looks to me like you compared 2 cores to 1 core in the SEPCint2006, and 4 cores to 1 core in the FP. Peter

To: Ali Chen who wrote (227028)	2/28/2007 8:32:14 PM
From: pgerassi	Read Replies (1) \| Respond to of 275872

Dear Ali: Base is useless for architecture comparisons. It is more a compiler test on a given architecture. Besides as I stated Intel compilers automatically optimize for Intel CPU doing all of the peak switches in base. So normal is the correct comparison. 13.3 for Opteron. And using the same third party type compiler on the same OS with the same clock speed keeping things equal of course, we use the ones I came up with comparing apples to apples and not apples to cactii. Woodcrest has to use slower FBDIMMs like Opteron uses the slower RDDR. Try using a A64 FX-74 with PC2-8500 versus Extreme 6800. They have at least the same type of memory. spec.org SPECint_2006: 16.3 spec.org SPECint_2006: 13.3 -->84% Surprise! With a non Intel compiler base no longer equals peak. Also given that base is spelled out and peak is just plain missing from the name on official SPEC pages, peak is the valid comparison. For FP if you want the best Opteron scores: spec.org SPECfp_2006: 15.8 spec.org SPECfp_2006: 14.7 -->93% Not using Studio: spec.org SPECfp_2006: 13.2 -->84% Same as above SPECint_2006. With Studio, many cores can run subthreads that are added together to produce a faster single thread performance. And with Barcelona's 4 cores, it will improve even more to the point where memory performance is the bottleneck. 4 Opteron cores beat silly, 8 faster C2D cores in SPECfp_rate2000. But a 7.5% increase in clock gained less than 2% in score. Showing that memory speed was the bottleneck and not core clock. And Intel apologists don't realize that most server code is not compiled with Intel compilers but with third party ones so it makes these Intel compiler scores of yours just irrelevant. 20% more IPC and AMD matches Intel using the same third party compiler. 40% more and Intel is in the K8Hs dust. The 70% more in OLTP and C2D is just blown away. Just the faster memory alone might allow Opteron to beat Woodcrest in SPECfp_2006. I did notice that there are no Cloverton scores for SPECxxx_rate2006. It must do so poorly no one wants to release them. Likely they are slower than Woodcrest's. And in SPECfp_rate2006, there are no multisocket C2D scores. Is Intel ashamed of what they reveal? No wonder Intel supporters hate SPECxxx_rate2006. THat must be terrible for you that nno one even wants to step up to the plate with C2D. Well, anyone has a freedom to dream and stretch reality beyond any limit. Comforting yourself with irrelevant excuses is not productive. Pete