Author: Robert Hyatt
Date: 21:40:54 04/15/03
Go up one level in this thread
On April 15, 2003 at 23:36:06, Rupesh wrote: >Dear All, > >I have noticed that there does not exist a universal optimum hash table size for >the optimum performance for every program. The size seems to depend on the >program that one is using. (context: number of 3+2 marches/ played on playchess >with a xeon 2.60 GHZ dual and 2 Gig DDR) > >For example, i see that the performance of shredder improves when i use 32/64 MB >in comparison to 256 MB. Whereas i see that if i raise the hash size to 512 MB/ >1024 MB for deep fritz, the performance improves significantly. Surprisingly, >raising the hash does not seem to lose on time ( although the speed of search >definitely slows down) as the program seems to play from hash!! > The answer is yes, the optimal size can vary from program to program, and from machine to machine... >Can somebody comment as to whethee such a thing as universal optimum hash size >is a subject of trial and error or is it really some scientific calculation >which is suggested in T-notes of chessbase. > >Any thoughts will be much appreciated. > >Amitesh
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.