Author: Bert
Date: 12:18:20 01/16/98
Go up one level in this thread
About performance, but not necessarily AMD vs. Intel: I recently went from a 100mhz 486 to a 300mhz PII with Fritz and a couple of other chess programs. I find that although search speed is, of course, improved dramatically, there appears to be a hash table limitation on performance. I cannot set a fixed time search of more than about twenty to thirty seconds per move in play mode without seriously degrading overall performance, because--and you can verify this with the formula that ChessBase supplies in their user manual--at 300mhz, I would have to have a hash table approaching my total 48 meg memory size in order to search as long as one minute per move, which means that execution stops in order to write memory-resident code out onto hard disk, in order to make room for the hash table. This causes the program to stop and wait for the disk writes to complete, which can take an unacceptable long time. This is supposed to occur only once, but may occur more often. Also, if you have to write out code that Windows 95 needs during the operation of your program, you lose time with rollout/roll-in. I find I am realistically limited to about twenty or twenty-five seconds per move maximum. Another factor that affects performance--and not just of chess programs--is that you are using the same old 66 mhz bus with all these superchips. Still, it is a whole lot better than a 486.
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.