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 221.42-0.1%Dec 10 3:59 PM EST
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: eracer who wrote (212713)10/5/2006 7:14:02 PM
From: pgerassiRead Replies (1) of 275872
 
Eracer:

Since neither FX-74 or Kentsfield is out, the only comparisons would be by samples and apparent NDA breakers. And they may be cherry picked to boot. There is a way of getting FX-72 4x4 numbers by using two Opteron 2220s on a MB with only the higher power using registered ECC memory versus the unbuffered non ECC memory that would be used in 4x4s. Somehow I don't think you want to use 2 2.66GHz Woodcrests as a standin as many here know that uses more power than the 2 Opteron 2220s both fully decked out with memory.

As for HPC loads, many of them have working sets that don't fit into the Woocrest's caches and thus force it to use main memory and there no one disputes AMD's leadership when that happens. Heck in SPECfp_rate2006, Woodcrest loses to Opteron and that holds 12 different applications.

And as for Coprocessing, as many supercomputer makers agree, Clearspeed plus Opteron is far better than Woodcrest or Itanium. And more are turning to DCA than FSBs.

And FSB is a bottleneck. Just look at the AMD-64 white paper with the SPEC graph with 3 different FSB speeds. Each speed reaches a plateau where increases in clock speed doesn't boost the score and that level rises with faster FSBs. Yes cache can mitigate it to a point, but it eventually runs out of gas. Why do you think Intel is investigating a way to bring multiple FSBs to a single socket? Yes, its mostly a stop gap until CSI, but thats at least 3 years away.

And due to the increased loads, the FSB speed may need to drop to 200MHz from 266MHz for desktops. So instead of a 2.66GHz Kentsfield@266 QDR, you get a 2.6GHz Kentsfield@200 QDR with a good loss of performance. Hey its happened before when most expected with Woodcrest going at 333MHz QDR (1333), that Conroe would too. It dropped to 266MHz QDR (1066) instead. That might help lower TDP at a cost of lower performance.

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

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2025 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News