Author: Robert Hyatt
Date: 07:11:45 12/21/99
Go up one level in this thread
On December 21, 1999 at 00:32:58, Greg Lindahl wrote: >On December 20, 1999 at 21:37:43, Robert Hyatt wrote: > >>Crafty's eval is 50% of the total time. Nobody is over 90% of total time in >>eval. > >You are correct about Crafty. You are wrong about "nobody". No I'm not. But I expect you can cite your example case? Diep is one of the slowest programs I know of, and Vincent has claimed for a couple of years to use maybe 90% of the cpu for eval in the opening, but that it tapers off as the game goes on. > >> The problem that >>you may be overlooking is this: Suppose a very fast program (fritz) searches >>300K nodes per second with almost no eval. Your hardware won't help much. >>Suppose another program searches at 30K with 90% of the time spent in eval. >>You speed that program up to almost 300K. That isn't enough to make a huge >>difference, particularly when someone can compile on other platforms (like >>the 21264) and get a big boost there as well... > >First off, I almost exclusively use the 21264 already, which you remember >because you've been carefully reading what I write. Second off, you seem to be >assuming that searching N nodes is the same as searching N nodes. But you know >that's wrong -- quality makes a difference. > >-- g I didn't say that at all. But I did say that if you start down a road where the asymptote is 10x, it is a tough road. Because you _never_ reach that asymptote. And in reality, you might not get very close. However, I can run on a 16X cpu 21264 system and get much more than a factor of 10x. Which was my point. 10X would be nice, should it actually happen. I think reality is closer to 1/2 of that for the most eval-bound program there is. And you are going to discover that you aren't going to get much access to information from commercial programs in this regard. Which means that you either get to interact with all the non- commercial guys, or only one commercial author. The latter isn't going to be easy with little help. The former isn't going to see anywhere near that 10x performance. The realities are sometimes a pain..
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.