Author: Dann Corbit
Date: 11:49:56 06/19/02
Go up one level in this thread
On June 19, 2002 at 13:11:27, Peter Berger wrote: >On June 19, 2002 at 13:03:13, Robert Hyatt wrote: > >>b) ponder=on can cause problems. Most of the time both engines will compute >>100% of the time. And get 50% of the cpu. But suppose one engine is not so >>sophisticated and fails (on occasion) to have anything to "ponder"? It will >>sit idle, giving the other program 100% of the cpu and a 2:1 time advantage >>for that search. >> > >This is the _major_ problem with ponder=on matches on single CPU IMHO . There >are several very strong engines that will have no ponder move quite regularly in >critical phases of the game (after fail-highs for example) - this looks like a >design decision. It looks like a rather principal problem for Ponder=On-matches >on a single CPU to me. > > >With time management and ponder=off - this looks like a lesser issue in >comparison IMHO ( for example for Crafty it was tested by Volker Pittlik and the >difference in strength wasn't measurable). This might be different for other >engines though. Some engines are far more effective at grabbing CPU cycles than others. For instance, I do not like to run EPD tests on my machine for either Bringer or GLC because they grab so much CPU, I cannot even move the mouse. The machine is completely locked up for the duration of the test. Other engines are far less harsh. Then, if you face these two engines against each other both with ponder on, the harsh engines will get 80% of the CPU or so and the less harsh engines will get less than 20%. Running ponder-on matches on a single CPU system is a very bad idea. Either get two machines or get a dual cpu system.
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.