Author: Brian Katz
Date: 15:21:32 01/14/02
Go up one level in this thread
On January 14, 2002 at 17:41:57, Jeroen van Dorp wrote: >The hash table calculation was a guide for use on PC's with older processors. > >One should make the hash table as large as possible, that means without Windows >*continuing* to swap to the HD. If that doesn't happen, there's noting wrong >with having big hash tables. > >In most cases and move big hash tables will only fill up to maybe 20%, but >sometimes the engine takes longer to search and then it can use bigger HT for >storage of more moves and extensions. > >Start the engine, e.g. with the command Alt-F2, listen if Windows starts >clearing memory (i.e. swapping to HD to store),wait till it silences, then stop >the engine and start it again. >If at this time windows remains quiet, there will be probably no swapping and >you have the biggest HT size available without the drawback of swapping, and >with enough spave for the engine itself cached in memory. > >J. Thanks James and Jeroen for your reply. First off let me apologize for my incorrect math. It should read: 2 x 350 =700 x 300 div by 40 = 7.25 div. by 1000 = 5.075 At 288 MB Hash Tables (which is what Fritz prompts me to use as a max.)I hear very little swapping. Start the engine for a few minutes and all the swapping stops. At 256 MB Hash I hear almost no swapping at all and there is no need to start the engine as above. I was under the impression however, that if the hash is too high, that the engine will take too much time too analyse. I will try to set the Hash at 288 mb on the internet with a 5/10 time control and see what happens. Interesting note, if I set the hash at 288 it reads 287 above the game score, so I don't know if it is really 288 or 287. Chessbase does suggest that you use only even numbers. Thanks again for your replies. Greatly appreciated.
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.