Author: Robert Hyatt
Date: 19:55:31 11/23/03
Go up one level in this thread
On November 23, 2003 at 22:48:29, Zach Wegner wrote: >On November 23, 2003 at 21:04:43, Robert Hyatt wrote: > >>On November 23, 2003 at 20:57:01, Sean Mintz wrote: >> >>>I did a test of the latest crafty (19.5) using an executable I made with Intel C >>>5.0 (/Ox /G6 /Gr /Ob2 /Qipo optimizations). My machine is a dual 1.2 ghz athlon >>>MP running at 1.35 ghz with 150mhz fsb. >>> >>>My crafty.rc file has this: >>> >>>ansi off >>>mode normal >>>adaptive 1.4M 6M 192M 1M 48M >>>smpmt 2 >>>book on >>>tbpath f:\chess\tablebases\ >>>cache 8M >>>learn 7 >>>ponder on >>>swindle on >>>flag on >>>log off >>>exit >>> >>>--- >>>Crafty v19.5 (2 cpus) >>> >>>White(1): mt 0 >>>parallel threads disabled. >>>White(1): bench >>>Running benchmark. . . >>>...... >>>Total nodes: 100409437 >>>Raw nodes per second: 772380 >>>Total elapsed time: 130 >>>SMP time-to-ply measurement: 4.923077 >>>White(1): quit >>>--- >>> >>>--- >>>Crafty v19.5 (2 cpus) >>> >>>White(1): bench >>>Running benchmark. . . >>>...... >>>Total nodes: 98693325 >>>Raw nodes per second: 1542083 >>>Total elapsed time: 64 >>>SMP time-to-ply measurement: 10.000000 >>>White(1): quit >>>--- >>> >>>1542083 / 772380 = 1.9965 x faster >>> >>> >>>Sean Mintz >> >> >>Told you it was better. :) > >this looks rather suspect. the multi-threaded bench is searching less nodes than >the single-threaded. are there any explanations for that? or are you just that >good? That happens all the time. That's not the number I look at here. I look _only_ at the raw NPS numbers... that compares CPU speed without any SMP oddities...
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.