Author: Moritz Berger
Date: 12:57:25 08/25/98
Go up one level in this thread
On August 25, 1998 at 11:48:50, blass uri wrote: >I want to know after how much time the hash tables of Junior are full >(Fritz5 gives me this information). > >I see that if I give Junior big hash tables it needs time to begin to think so >too big hash tables are counter productive and I want to know what >is the optimal hash tables per time. > >I found that at least in Junior4.6(I did not check it in Junior5) >if I give Junior 100Mbytes hash tables it may be slower than 20Mbytes >and not only because of the first time it needs to clear the hash tables. >It looks at less positions to go to the same depth but need more time to look at >them. > >This fact make me think that maybe too big hash tables for Junior are counter >productive also in long time control. >can it be counter productive because of the computer? > >Is it possible that in pentium200MMX small hash tables are better and in a newer >computer big hash tables are better? > >Uri Junior 4.6 took about 5 minutes to load with e.g. 128 MB hash tables under Windows NT. Junior 5 (32 bit) doesn't have this strange problem, so I guess your observation doesn't apply to J5. Clearing e.g. 100MB hash tables takes a fraction of a second on a fast machine. Junior does about 80%-90% as many NPS as Fritz 5 (rough estimate), so as a rule of thumb it ought to fill up hash tables almost as fast as Fritz does. Moritz
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.