Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Newbie Questions: Winboard/Crafty on Windows PC

Author: Jeremiah Penery

Date: 13:31:45 09/15/99

Go up one level in this thread


On September 15, 1999 at 13:46:13, José de Jesús García Ruvalcaba wrote:

>On September 14, 1999 at 18:18:46, Jeremiah Penery wrote:
>
>>On September 14, 1999 at 09:39:58, Robert Hyatt wrote:
>>
>>>On September 14, 1999 at 09:26:28, James Flanagan wrote:
>>
>>>>How big should the cache be with 64Mb of RAM?
>>>>
>>>>Are there any other critical parameters that I need to set to get optimum
>>>>performance?
>>>>
>>>
>>>
>>>the best you can do is probably hash=48m, hashp=8m.  and if you are using
>>>compressed tablebases, that won't work.  Then you will have to try
>>>hash=24m, hashp=8m and cache=8m and see how that does...
>>
>>With Win98, I wouldn't suggest setting hash=48m, especially if you are running
>>ANYTHING else.  The OS just eats up way too much memory, and you'll get disk
>>paging, which will slow you down a lot.  With 64MB on my machine, I usually use
>>hash=12m and hashp=4m.  It's possible to set hash=24m...Sometimes it works ok
>>and sometimes not.  Either way, there will be VERY LITTLE difference in
>>strength/speed/etc.  Bob did an experiment with different hash sizes a while
>>back to see the performance impact...Perhaps I can dig it up a little later and
>>post it.  IIRC, there was no significant difference anywhere between a small
>>hash (4m, even?) and a very large one (384m).
>>
>>Jeremiah
>
>	NPS is not the way to compare crafty's performance with different hashtable
>sizes in the same machine. Actually, crafty scores a higher number of nodes with
>smaller hash tables, but does less useful work as many of them are repeated.
>	The search depth reached in the same amount of time is more significative. As
>long as you do not swap to disc, the bigger hashtables allow crafty to search
>deeper in the same amount of time (or to search to the same depth in less time).
>José.

I never mentioned NPS.  Bob's experiment compared the times to reach ply-N with
the different hash sizes.  I really need to look this up, however, as I am not
100% sure.

Jeremiah



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.