Author: Vincent Diepeveen
Date: 05:34:34 11/07/02
Go up one level in this thread
On November 06, 2002 at 18:02:57, Severi Salminen wrote: >>>A good program must be both a quick searcher and a good evaluator. >> >>Tiger14 is not a fast searcher and I guess that the same is for tiger15. >>I do not think that a good program must be a quick searcher. > >"a quick searcher". Please define "quick". What was quick in 1980, is now slow. >But the next holds allways: the faster, the better. "Must not be quick" means >practically nothing. > >Severi quick is for example less than 1000 clocks a node like fritz 5.32/5.16 needed. quick nowadays is 2000 clocks a node. note that a cpu today can do more instructions a clock than it could do a few years ago. I don't need to mention that if fritz got like 900 clocks a node on a P5-133Mhz, that then needing around 1800 clocks in 2002 on a K7-1.6Ghz means simply that it is a lot slower now. How much of that is caused by extra knowledge and how much is caused by seemingly no longer incrementally generating moves, will probably remain a mystery.
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.