Author: Uri Blass
Date: 04:27:08 02/15/02
Go up one level in this thread
On February 15, 2002 at 04:18:48, Gian-Carlo Pascutto wrote: >On February 14, 2002 at 19:13:28, Uri Blass wrote: > >>>For your example, I don't see why it should. If it can play on the XT >>>in the required contraints, the _same program_ likely will gain very little >>>or no strength moving to an Athlon. > >What part of 'same program' didn't you understand? > >-- >GCP I do not understand. Do you mean that B is not the same program as A in my example? It is possible to define the following levels for program in go: level 1:choose a move by some fast algorithm level 2:try all the legal moves and play games against yourself at level 1 choose the legal move that gives the best result based on the game against yourself level n:try all the legal moves and play games against yourself at level n-1 choose the legal move that gives the best result based on the game against yourself. I expect level n to be better than level n-1 and it is possible also to have level between level n and level n-1(if you play in the game against yourself the first move at level n-1 and the rest of the moves at level n-2) The program can choose level based on the time control or the hardware so I expect better hardware to help. I do not say that it is the best idea to earn from hardware but even this idea can be productive. I did not get a reply for the question how much rating programs can get from better hardware. Even 1 elo per doubling the speed of the hardware is an answer but I cannot believe that nothing can be earned by doubling the hardware. If the difference from doubling the hardware is too small to find out then it is possible to find how much program can earn from being 1000 times faster and use the result to estimate the gain from being twice faster. Uri
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.