Author: Uri Blass
Date: 07:09:35 12/07/05
Go up one level in this thread
On December 07, 2005 at 10:00:02, Russell wrote: >On December 07, 2005 at 09:19:09, Uri Blass wrote: > >>On December 07, 2005 at 08:10:37, Kirill Kryukov wrote: >> >>>Hi all, >>> >>>I'm happy to announce the update of CEGT 40/40 lists and statistics. We have now >>>more than 540 games of Rybka 1.0 Beta 32-bit, and also 104 games of 64-bit >>>version. 32-bit Rybka is now prowdly leading our main rating list. >> >>only if you decide that the main rating list is the list that include only >>programs with more than 300 games. >> >>I think that the list that include all programs has more information so I >>consider it as the main rating list. >> >>Leading does not mean best so I see no problem with claiming that 64-bit is >>leading(in this case we can say that 64-bit is also better than 32-bit based on >>the knowledge that the only difference between 64-bit and 32-bit is speed) >> >> Rating of >>>32-bit version is not that incredible as it was expected, but still good enough >>>to secure the 1-st place, for now. 64-bit version is probably stronger. >> >>You can be practically sure that it is stronger based on the knowledge that the >>only difference is speed and the only question is how much stronger. >> >>I wonder what is the speed advantage of 64-bit relative to 32 bit >> >>The readme give the following details >> >>32-bit Rybka 1.0 Beta: 104 knps >>64-bit Rybka 1.0 Beta: 166 knps >> >>The main question is if 32-bit Rybka is not tested also on machines that are not >>64 bit machines because in this case the question is if 32 bit Rybka earn the >>same speed from 64 bit machines like other chess programs. >> >>It may be productive if we get information about the relative speed of programs >>on different machines as first step >> >>second step may be information about the number of games of every program in >>different machines. >> >>Uri > >The 64-Bit may be better than 32 in fast (Blitz games). In long game, I doubt it >would be significance Diff in Elo. The difference may be smaller but still significant. Nevertheless, One adv of Reb is that it speed >(NPS) very low compare to other engines therefore 64M Hash is like Huge for this >engine. I think that in the CEGT games the difference is going to be bigger relative to SSDF mainly because of the small opening book that the CEGT use. I also think that speed is significant here and I am not going to be surprised if we discover that Rybka 64 bit is 50 elo better than 32 bit. I do not think that nodes per second are important here and I do not think that Rybka earns more elo from speed improvement relative to other programs. Uri
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.