Author: Howard Exner
Date: 04:53:34 12/05/97
Go up one level in this thread
On December 04, 1997 at 18:16:54, Michael G. Kramer wrote:
>Those interested in chess software lack a reliable benchmark to compare
>hardware setups.
>
>Suggest that for the moment one engine which counts positions (or times
>ply depth search) and one position be used to test all hardware. It may
>be neccessary to establish a Dos standard and a Windows standard. (maybe
>even a w95 standard0
>
>As motherboard, win accel boards, specific memory types all profoundly
>effect actual system speed, two k-6 200 systems with different mboards
>etc could be remarkably different in performance.
>
>using a stopwatch or perhaps a specific engine utility or ply depth and
>studying the time to reach that ply depth or the positions achieved to
>reach that time mark any hardware could be compared in a fair and
>objective way.
>
>Has anyone a specific suggestion as to an engine and a position and a
>utility which has wide ease of use which can fulfill this task? I would
>like to make some comparisons myself.
>
>I think it would be good for us to think about a standard which we all
>can accept as useful for the moment. I recommend use of an existing
>chess engine as this will test specifically the type of computation
>which interests us. A generic speed test may be somewhat deceptive but
>then again I wonder. Is there a benchmark existing which is highly
>valuable already in this matter?
>
>It would be handy if one could read ("The Magma 8 generates 400,003
>positions on the "Standard Chess Benchmark" )This could eventually
>become a familiar and useful hardware test.
Ed Schroder has done his homework on this one. Here are some
selected results from his homepage:
Rebel8Mark
R8 Time R8 Cmark P90=1
Dx-66 2353 40 0.47
P90 1050 90 1.00
P166 530 178 1.98
P200 480 197 2.19
P6-200 341 277 3.08
P200MMX 319 296 3.29
K6-200 260 363 4.04
K6-233 219 432 4.79
The data from column one is the sum of three tests given on
Ed's Homepage. The next two columns I've added here to assist in
comparing speeds.
Each program I've discoverd is machine dependent but each programmer
has probably done similar tests on hardware. If the data were
translated into a column where for example a P-90 machine is given a
chessMark of 90(then compared for the other machines as in column two)
one could have a table with a quick view of which machines enhance
which programs.
This has implications on the call by many for equal platform tests.
There may not exist an equal platform as some programs may benefit
more from a 200MMX than others.
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.