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.
Politics : Formerly About Advanced Micro Devices

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Paul Engel who wrote (94698)2/23/2000 12:52:00 AM
From: Dan3  Read Replies (1) of 1575761
 
Re: Why don't you ask why the Athwipey ISN'T THREE TIMES as fast as a Coppermine in FPU calculations...

For solving LP optimization problems it often is: when we received our first Athlon 550s late last fall, we ran one of the models we're working on it and on a Pentium III/550 (not an E, obviously) as a test, and the same problem that ran 19 seconds on the Pentium ran 4 seconds on the Athlon. I've not brought it up before, because the comparison was not carefully controlled, and the PIII was running 98 while the Athlon was running NT (both IDE drives, both 128 meg PC100 SDRAM)

The problem fits in main memory, so it wasn't an issue of virtual memory efficiency on the NT box, no disk swapping is observed when this model is run.

My guess is that part of the code that fit in Athlon's L1 didn't fit in Pentium's L1 - the difference shouldn't have been that large. But, for whatever reason, it was.

There is more to computing than games and word processing, and the code for much of that work comes out of Microsoft or Borland compilers, and isn't carefully optimized for any hardware. That's the code used in much of the business and scientific world, and that's the code that the Athlon runs well on - SIMD2 (or 3D-Now) won't help on most of the applications where performance really matters.

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