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 -- Ignore unavailable to you. Want to Upgrade?


To: Elmer who wrote (74555)10/7/1999 5:43:00 PM
From: Tenchusatsu  Read Replies (2) | Respond to of 1573062
 
Elmer, <So you are saying the SpecFP95 scores improved from
~16 for a 600MHz PIII to ~25 for 600MHz CuMine without the
use of SIMD instructions? That's a 56% improvement. Is that
correct????>

The foil said there was a 20% increase from a Pentium III
600B to 600EB. I assume that meant the 600B was also
benefitting from cache-streaming instructions and the
820/RDRAM platform.

By the way, one correction here. The scores are
SPECint_base95 and SPECfp_base95. So to extrapolate numbers
here:

SPECint_base95 SPECfp_base95
-------------- -------------
PIII 600, no SSE 24.6 14.9
PIII 600B, w/ SSE 25.9 20.8
PIII 600EB, w/ SSE 29 25

(The PIII 600 numbers were taken from the Dell Precision
Workstation 410. I don't know what the highest numbers
should be for that processor.)

Tenchusatsu



To: Elmer who wrote (74555)10/8/1999 12:35:00 AM
From: Haim Barad  Respond to of 1573062
 
Remember the extra fill buffers on CuMine: 50% more than Katmai. Aparently prefetch was used.

Now if the opcode for prefetch is the same for AMD (I think it is), then that means that AMD could also benefit from the new code. I'm sure we'll hear something right away about this.

Haim