Author: Robert Hyatt
Date: 17:34:27 07/28/99
Go up one level in this thread
On July 28, 1999 at 05:23:53, blass uri wrote: > >On July 27, 1999 at 22:56:03, Robert Hyatt wrote: > >>On July 27, 1999 at 02:41:30, blass uri wrote: >> >>> >>>On July 26, 1999 at 19:23:01, John Wentworth wrote: >>> >>>>In my opinion, Cray Blitz running >>>>on a Cray Supercomputer would still blow away any of the current programs >>>>running on any PC today. >>> >>>I want to see games to see if Cray blitz of today is really better than parallel >>>crafty of today. >>> >>>I am interested to see results of games at different time control. >>> >>>I guess that cray blitz has good chances at blitz but is going to lose at slower >>>time control. >>> >>>Uri >> >> >>It isn't going to lose. Crafty on a quad xeon is doing about 700K nodes >>per second. Cray Blitz is doing 7M nodes per second. That is 10X, which is >>very hard to overcome. Cray Blitz also has some evaluation things that I can >>do with vector hardware cheaply, but which I can't do in normal scalar machines >>because of the cost. >> >>But even if knowledge was equal, 10x is murder, speed-wise. > >I agree that 10x is a big advantage if the knowledge is equal but you stopped to >work on cray-blitz some years ago when you continue to work on crafty. > >This is the reason that I guessed that the knowledge of crafty may be better >than the knowledge of cray-blitz and it has chances at longer time controls. > >You also had a problem to get time to work on the cray and this may cause >cray-blitz more bugs then crafty(another reason that crafty has more chances). > >I know that cray-blitz did not do impressive results in the last tournaments >that it played. >I understood that it lost against wchess because of a bug but there may be more >bugs that you are not aware to them. > >Uri I would never argue that point. If crafty might have _any_ chance, that chance would be based on lots of games against human GM players, tens of thousands of such games in fact. So it would not (likely) have any detectable bugs when it might meet CB. But many CB versions were also very well tested, just not always on a Cray. IE in 1986 at the WCCC the last 3 games it played were done very well... I had crafty annotate those games and posted the results a while back, and CB didn't make any mistakes that Crafty could find... and in a couple of places CB found tactical shots that Crafty couldn't... But to put my point in perspective: A "solid" Cray Blitz would destroy a "Solid" crafty. A "buggy" cray blitz could win or lose depending on luck. I was _always_ referring to a "solid" version and not necessarily what played in the most recent ACM events... By the same token, several Crafty versions have had serious bugs as well... while several have not had any that I knew of...
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.