mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2004-01-17, 23:49   #1
GSV3MiaC
 
Jan 2004
Shropshire, UK

24 Posts
Default Athlon 64 Benchmark(s)

Has anyone actually got any data for Prime95 running on an Athlon64 (either an XP3200+ or the FX51) under WinXP?

Yes, I looked at the 'perpetual benchmark' thread, but didn't see anything definitive.
GSV3MiaC is offline   Reply With Quote
Old 2004-01-18, 11:03   #2
kaeptn_kork
 
Sep 2002

23·3 Posts
Default

they're still kind of rare

didn't find another bench than that I managed to get once of an overclocked a64, perhaps it's still a bit useful
RAM was running 236 MHz at 2-3-4-7

Quote:
AMD Athlon(tm) 64 Processor 3200+
CPU speed: 2362.53 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 1024 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.7, RdtscTiming=1
Best time for 384K FFT length: 20.614 ms.
Best time for 448K FFT length: 24.733 ms.
Best time for 512K FFT length: 27.828 ms.
Best time for 640K FFT length: 35.469 ms.
Best time for 768K FFT length: 43.169 ms.
Best time for 896K FFT length: 52.332 ms.
Best time for 1024K FFT length: 58.733 ms.
Best time for 1280K FFT length: 78.609 ms.
Best time for 1536K FFT length: 95.939 ms.
Best time for 1792K FFT length: 115.331 ms.
Best time for 2048K FFT length: 130.165 ms.

Last fiddled with by kaeptn_kork on 2004-01-18 at 11:04
kaeptn_kork is offline   Reply With Quote
Old 2004-01-18, 12:29   #3
GSV3MiaC
 
Jan 2004
Shropshire, UK

100002 Posts
Default

Thanks, that's the one I had already seen .. and if you back-out the 18% over-clock, the times are only about 5-10% better than I get with my XP2600+ (333Mhz memory, 2.088Ghz cpu speed), which doesn't seem very good (probably not as good as a regular XP3200+ Athlon (no '64')).

I gather that Athlon64 SSE2 is 'sort of not working very well', but by that benchmark it isn't working at all (i.e. no better than using the regular x87 FPU instructions).
GSV3MiaC is offline   Reply With Quote
Old 2004-01-18, 12:54   #4
Dresdenboy
 
Dresdenboy's Avatar
 
Apr 2003
Berlin, Germany

5518 Posts
Default

On this thread pages you'll find more benches - for example with and w/o using SSE2. There are many in this thread, starting on page 4 IIRC.

Quote:
Originally posted by GSV3MiaC
I gather that Athlon64 SSE2 is 'sort of not working very well', but by that benchmark it isn't working at all (i.e. no better than using the regular x87 FPU instructions).
Not really.

There are other reasons (but surely not that obvious):
  1. The K8 in currently available variants (Opteron, A64 etc.) has some SSE/SSE2 memory bandwidth bottleneck while loading SSE2 registers. Usually the K8 can load 2 64bit values (as much as one SSE register) per cycle when using the integer or MMX units and probably with x87 code too. But with SSE/SSE2 that reduces to half of that. There are workarounds - but these have to be implemented.
  2. The max throughput of the K8 FPU (1mul/1add per cycle with double precision) can be reached by x87/scalar SSE2 and vector SSE2 code. P4 can do that only with vector SSE2 code. So it looks like P4 vector code improves a lot and on K8 not at all (or only a little due to positive side effects of SSE2). But in fact both have the same max. throughput - but at different clock speeds. And because of obstacles like (1) the code is still faster per clock on P4 than on A64.

So currently (without new GIMPS client software) the A64 is not so useful for Prime95 crunching farms. Although this year could be interesting in view of AMDs 90nm CPUs with 512kB cache and dual channel - maybe they have a higher Prime95 throughput per Watt than Northwood or Prescott.
Dresdenboy is offline   Reply With Quote
Old 2004-01-26, 10:43   #5
DonDan
 
Jan 2004

716 Posts
Default

I ran the bench on my 3200 with 2*512 RAM (2-3-3-8 400DDR).
There u go:
AMD Athlon(tm) 64 Processor 3200+
CPU speed: 2009.81 MHz
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 1024 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 version 23.8, RdtscTiming=1
Best time for 384K FFT length: 24.581 ms.
Best time for 448K FFT length: 29.233 ms.
Best time for 512K FFT length: 33.991 ms.
Best time for 640K FFT length: 42.029 ms.
Best time for 768K FFT length: 52.164 ms.
Best time for 896K FFT length: 63.580 ms.
Best time for 1024K FFT length: 70.780 ms.
Best time for 1280K FFT length: 95.559 ms.
Best time for 1536K FFT length: 117.257 ms.
Best time for 1792K FFT length: 141.958 ms.
Best time for 2048K FFT length: 159.141 ms.
DonDan is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Athlon II 620 x4 henryzz Hardware 5 2009-10-28 03:25
amd athlon 64 bug? Unregistered Hardware 17 2004-06-01 22:57
Athlon 64 and SSE2 ThomRuley Hardware 17 2004-05-14 19:26
Athlon 64 ThomRuley Hardware 7 2003-09-28 13:31
Athlon vs. P4 ThomRuley Lone Mersenne Hunters 1 2003-07-11 23:15

All times are UTC. The time now is 16:56.

Sat May 15 16:56:46 UTC 2021 up 37 days, 11:37, 0 users, load averages: 2.00, 2.01, 1.93

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.