Author: Albert Bertilsson
Date: 21:10:01 12/17/03
Go up one level in this thread
On December 17, 2003 at 18:34:54, Keith Evans wrote: >On December 17, 2003 at 16:01:06, Slater Wold wrote: > >>On December 17, 2003 at 15:40:07, Mathieu Pagé wrote: >> >>>Just for fun i calculated the nodes/secondes for the last 24 hours >>> >>>and it is: 4450300925 nodes /seconds >>>or : 4450 millions nodes / seconds >>>or : 4,5 giga nodes / seconds >>> >>>This system is faster than my own chess engine by a factor of 1:4450 >>> >>>I think it's an all time record for a perft. >>> >>>impressive, isn't it ? >>> >>>Thanks Albert to privide us with this fun project (not usefull, but fun :) >>> >>>let's compute guys. >>> >>>Mathieu >> >>And Bob thinks his quad opteron is fast! Pftt! >> >>4.5B NPS. Imagine that chess engine... > >All Bob needs to do to boost this dramatically is to change how he counts nodes. > >My understanding is that the nps given for the distributed perft project counts >all of the nodes that would have been visited if hash tables were not used. But >then hash tables are used, so not all of the nodes are actually visited. Is this >understanding correct? > >-Keith You're right Keith, some considerable cheating is done by using hash tables to "boost" node counting. But even without it I still think that the project would reach VERY high node counts, perhaps almost 1BNPS (when testing at home I solve problems twice as fast with hash tables so perhaps even 2BNPS are counted by the computers). /Regards Albert
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.