Author: Mark Young
Date: 10:58:10 05/03/98
Go up one level in this thread
On May 03, 1998 at 12:45:06, blass uri wrote: > >On May 03, 1998 at 12:19:30, Mark Young wrote: > > >>I don't think the fritzmark has anything to do with KN/s. It has a set >>position that it does a timed test on to a fixed depth. So the fritzmark >>is just a timed test to a fix depth. This way it can take CPU speed and >>bus speed and hash size into account. > >sometimes if I want to arrive to a fixed depth >big hash tables cause fritz5 to be slower >in reaching the same brute force depth because when it uses big hash >tables it increases the selective depth >so if this is the fritzmark maybe it can be better with smaller hash >tables. > >Uri -------------------------------------- Yes this is true. In the short fritzmark test I got lower fritzmark with big hash settings. But on the long fritzmark test it is better to have bigger hash settings. I have also found that some hash settings are much better then others on the fritzmark test long. I found that on my system 49 megs is best on the test. If I set it to 48megs or 50 megs my fritzmark can drop as much as 10 to 15 percent. I found this strange behavior by setting my hash at 64k and going up 1 meg of hash per test to see what the speed advantage more hash table would give me 1 meg at a time. It repeated this behavoir over and over again in win95. When I tested it on win98 this behavior stopped and I got more of a steady upward trend as I increased the hash table size.
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.