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.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Elmer who wrote (98541)	3/15/2000 8:33:00 AM
From: DRBES	Respond to of 1570657

re: "This is incorrect. The GHz PIII is available however in very limited quantities." Indeed...and so are 2 gig Athlons. DARBES

To: Elmer who wrote (98541)	3/15/2000 11:38:00 AM
From: Petz	Respond to of 1570657

RE:<Athlon FP is only slightly faster> The Athlon FPU can do a double precision multiply-add in one clock cycle where the P3 needs two clock cycles. Thats a little more than "slightly" in my book. The problem is that unless the data is coming from and going to the L1 cache the memory bandwidth can't keep up. On chip L2, or, better yet, on chip L2 with 256 bit interface as JC's has reported, will give up to 2-8x the bandwidth to the L2. (probably 4x) If one or more of the data arrays are huge and residing in main memory, PC133 will help by 33% and DDR 2100 will double throughput. Petz

To: Elmer who wrote (98541)	5/31/2000 12:58:00 PM
From: Ali Chen	Respond to of 1570657

<I believe this is incorrect on 2 counts. First Athlon FP is only slightly faster and second, Intel's compilers only use the prefetch instructions which Athlon uses also.> I am confident that your believes are incorrect on both counts. For the FPU performance there is a pure FPU test from SPEC suite: spec.org It is special in the sense that it does not depend on main memory and exercises only FPU and caches, so any concern about better or not memory is set aside. For precisely identical conditions, apples-to-apples, the results are: K7, (IFC v.2.4): spec.org 145.fpppp = 137s, base ratio = 70.1 P-III-Katmai: (IFC v2.4) spec.org 145.fpppp = 400s, base ratio = 24.0 It is apparent that the K7 is about 3 times faster. Even if the SSE/prefetch is engaged, K7 is still more than 2 times faster as compared to equivalently clocked Coppermine: P-III-SSE (compiler v4.5) spec.org 145.fpppp = 289s, base ratio = 33 <Intel's compilers are publicly available and AMD can do an apples to apples benchmark comparison> This is highly misleading. When Intel compiler 4.5 detects a processor other than P-III, all prefetch optimizations gets by-passed to a 486-like code. Therefore Athlon cannot use prefetches in Intel's 4.5 code. Got it? When you will stop doing this shameless plugs about "publicly available" Intel's compilers and AMD Athlons?