Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is SPEC a bad test organisation according to Hyatt?

Author: Vincent Diepeveen

Date: 19:53:39 03/01/04

Go up one level in this thread


On March 01, 2004 at 18:57:03, Frank E. Oldham wrote:

>On March 01, 2004 at 18:05:55, Vincent Diepeveen wrote:
>
>>
>>So all we know it's a slow CPU, that's all. How slow is not clear simply.
>>
>
>Slow? It's not quite up to opteron levels, but with the poor gcc compiler (e.g.,
>won't pass bitboard in a register) I get:
>
>crafty19.10 w/futility bench
>
>1 x 2GHz G5
>Total nodes: 89942714
>Raw nodes per second: 1249204
>Total elapsed time: 72
>SMP time-to-ply measurement: 8.888889
>
>
>2 x 2GHz G5
>Total nodes: 83479388
>Raw nodes per second: 2455276
>Total elapsed time: 34
>SMP time-to-ply measurement: 18.823529
>
>Frank

The testsets they benchmarked publicly the G5, they used GCC for the x86
hardware. Then they claimed victory. Later those tests were compiled by decent
compilers and eated that G5 alive.

Anyhow, you won't find the G5.

Bob already got 2.2+ Mln nps at his outdated dual Xeon 2.8Ghz by the way.

I get 1.2MLN nps at my old K7 MP2600 already single cpu.

There is 3.2Ghz Xeons and of course new generation Nocona and Opterons are there
now. Try a dual opteron with pathscale compiler and crafty and you'll see.

Power4 at 1.7Ghz is nowhere impressive:

http://www.spec.org/osg/cpu2000/results/res2004q1/cpu2000-20040126-02725.html

This despite using unified caches.

Your G5 is such a cpu with less cache that's all.

What is your RAM and RAM timings?




This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.