Author: Ernst A. Heinz
Date: 04:59:43 05/20/98
Go up one level in this thread
On May 20, 1998 at 07:38:39, Albert Silver wrote: >On May 20, 1998 at 06:42:55, Ernst A. Heinz wrote: > >>Hi all, >> >>The "DarkThought" team is happy to announce the long-awaited update of >>our WWW pages at URL = <http://wwwipd.ira.uka.de/Tichy/DarkThought>. >> >>The updated pages report about our latest results with standard test >>suites (BS-2830 = 2720, BT-2630 = 2521, LCT-II = 2610) and in Nunn test >>games against strong PC programs (60 games at tournament time controls). >> >>Please enjoy our pages and have fun replaying the games! >> >>=Ernst= > >Hi Ernst, > >First of all, congratulations on your impressive results. I haven't had >time to look at the games yet, but had a few questions regarding the >Nunn position matches. I am not a technocrat, but couldn't help >wondering about the hardware of Dark Thought's opponents. According to >your site, the Alpha used scored a SPECint of 18.4 whereas the K6/233 a >9.6 (a near two-fold difference in score though I don't know if this >means a two-fold difference in speed), yet Dark Thought's opponents only >benefitted from a 50% increase in time. Finally, and Moritz is bound to >point this out, Fritz 5 would have greatly benefitted from an increase >in RAM as well as an Intel processor. The RAM would have significantly >helped it particularly, and this can hardly be considered asking for too >much as Dark Thought itself was using 170 MB. > > Albert Silver Give me an AMD/KryoTech K6-3D 400MHz or an Intel Pentium-II 400MHz machine with 256MB of RAM and I will happily test all the other programs on faster hardware, too. :-) The additional 50% time bonus seemed to be a good compromise to [citation from our WWW page] "partially compensate" for the hardware difference. If the slower hardware gets too much of a bonus, the faster tends to benefit from this even more because of the time gained while pondering on correctly predicted moves of the opponent. BTW, if we were to play a serious match against PC programs today "DarkThought" would surely run on the new DEC Alpha-21264 EV6 CPU. This CPU gives "DarkThought" another 200%-300% performance increase at 600MHz as compared to the current DEC Alpha-21164a EV56 CPU. Thence, the hardware we tested on was also not the fastet commercially available for "DarkThought" right now. Anyway, we never claimed to run a uniform platform or equal hardware match. Neither did we claim that "DarkThought" plays better chess than all the other strong programs out there. It just plays well enough to stand its own against reasonable opposition. =Ernst=
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.