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) -- Ignore unavailable to you. Want to Upgrade?


To: porn_start878 who wrote (63002)11/8/2001 2:47:54 PM
From: wanna_bmwRespond to of 275872
 
Max, Re: "Unlikely IMO. The 64-bit mode isn't really performance oriented (only a couple more registers) and the performance predictions seem to follow model numbers, which are based on 32-bit benchmarks."

At MPF, Fred Webber mentioned that the largest gain in performance comes from literally smaller code once recompiled. In x86, there are a lot of instructions that generate addresses, for example, that would be unnecessary in a 64-bit instruction set. Instructions that move data between registers will also be reduced due to more registers being available. In the end, less code means that it gets done faster, and I think Fred Webber estimated that some applications could get a 10-15% increase using x86-64.

wanna_bmw



To: porn_start878 who wrote (63002)11/8/2001 2:57:13 PM
From: jcholewaRespond to of 275872
 
> Unlikely IMO. The 64-bit mode isn't really performance
> oriented (only a couple more registers) and the
> performance predictions seem to follow model numbers,
> which are based on 32-bit benchmarks.

You are likely correct here; however, the 8reg SSE support is likely counted in the boost, since it will largely negate Intel's "instruction advantage". The 16reg SSE support is likely not counted (imho), and we'll have to see how that plays out.



To: porn_start878 who wrote (63002)11/8/2001 2:59:27 PM
From: pgerassiRead Replies (1) | Respond to of 275872
 
Dear Max:

X86-64 has both twice the number of GPU registers at twice the size each. It also had 8 additional SSE registers. Thunderbird has 8 GPUs at 32 bits and 8 at 80 bits for a total of 112 bytes of registers. X64-64 has 16 GPUs at 64 bits, 8 at 80 bits and 16 at 128 bits for a total of 464 bytes of registers over 4 times that of Tbird. Hardly a couple more registers (16 << 40)!

Pete

PS: As M2000 implies that AXP will be at 2000+ which is confirmed by the latest BIOSes at 1.66GHz before the end of this year! Probably mid December. Also, Sledgehammer will have more L2 than Clawhammer (>= 1MB)!