Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The perfect cure:

Author: William Penn

Date: 06:12:38 03/14/04

Go up one level in this thread


On March 14, 2004 at 00:39:51, Mike S. wrote:

>**Never attempt to change the hash size while the engine is calculating.** :-))
>
>(That's why no other user ever has had that problem, nor will...)
>
>It's like trying to change a tyre at 50 mph. I guess the insurance won't pay but
>claim that was improper use...
>
>But seriously, you'd just have to set the hash size to a "save" size before
>starting longer tasks. I guess you wanted to reduce the size because you needed
>ram for other tasks running parallel (I see no other sense)? So it's just a
>matter of planning. The performance will not be significantly worse anyway, i.e.
>with 128 MB compared to 256 MB.
>
>Regards,
>M.Scheidl

I change hash sizes because it significantly affects the engine speed (kN/s). I
need to reduce hash size in endgames when tablebases are heavily accessed.
Otherwise my system becomes overloaded somehow, the hard drive starts churning
very heavily (pagefile?), and the engine speed slows down greatly.

Note that I only use this software in infinite analysis mode, and run each
analysis for several hours. So with such long analysis times, I need to use
maximum hash size for best performance, but tablebase access interferes with
that ideal.
WP



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.