SI
SI
discoversearch

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)
AMD 207.67+2.2%3:59 PM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: dougSF30 who wrote (239949)9/7/2007 1:15:36 PM
From: PetzRead Replies (2) of 275872
 
Intel's dirty little secret -- poor frequency scaling

re: AMD folks have publicly said that cHT is a problem w.r.t. higher frequencies?

Yea, maybe at 4 GHz. Doesn't seem to be a problem at 3.2 GHz, does it? Ever heard of 8224SE?

are you ignoring the fact that he's comparing 4-socket systems? Tigerton vs. 4S Barc?

I can't get the Chinese site to translate, but if I were you I wouldn't take much comfort from "clock-for-clock" comparisons at 2 GHz. Here's the scaling percentages for Clovertown from 2 GHz to 3 GHz:

SPECint 67%
SPECfp 56% from 2 GHz to 2.67 GHz -- no results for 3 GHz!
SPECint_rate 52%
SPECfp_rate 36%
note: all above based on Intel submission using Supermicro motherboard. 2-socket numbers, 4 socket undoubtedly worse.

Yes folks, it's true. Even using SPECint_rate, which the AMD bashers here claim is a good predictor of server performance, you get a 26% performance increase by increasing the clock 50%! Floating point? Forget about it! Even single thread, single core scaling is abysmal. Anyone who buys Intel's two highest speed grades, which also require more power, is stupid.

So Intel "just as fast" at 2 GHz may translate into Intel 20% behind at 3 GHz not many days from now.

Petz
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext