Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: CM6000>They say to use 16mb Hash Table! What table size for S. Death Gam

Author: Paulo Soares

Date: 15:39:02 04/05/99

Go up one level in this thread


On April 05, 1999 at 17:46:18, blass uri wrote:

>
>On April 05, 1999 at 09:34:57, Micheal Cummings wrote:
>
>>
>>On April 05, 1999 at 02:12:06, Paulo Soares wrote:
>>
>>>On April 05, 1999 at 00:48:11, Terry Ripple wrote:
>>>
>>>>  If 16mb is more than enough for 40/2hr, then how much Hash Table Size would be
>>>>enough for playing Sudden Death Games of 60/S.D.,30/S.D.,15/S.D. and 5/S.D.
>>>>
>>>>  Would it be better to use smaller Hash Tables for these Sudden Death Games and
>>>>if so, then what would be the best size to use?
>>>
>>>Terry,
>>>


Paulo Soares
>>>The equation below is used to calculate the minimum hashtables size,
>>>and it is in the manual of Fritz5.  I believe that it serves for the
>>>CM6000.
>>>
>>>	HT(Kb)=2.0*Pfreq(Mgz)*t(S)
>>>
>>>	HT= hashtable size in Kb.
>>>	Pfreq= clock speed of the processor in Mgz.
>>>	t= average time avaiable per move in seconds.
>>>
>>>Examples:
>>>1. Processor with 200Mgz
>>>   Control time=40/2- Average time 3min=180s
>>>
>>>	       	HT=2.0*200*180=72000Kb= 72Mb
>>>
>>>2. Processor with 200Mgz
>>>   Control time= Blitz(5min)- Average time=10s(esteemed)
>>>
>>>		HT=2.0*200*10=4000Kb= 4Mb
>>>
>>>3. Processor with 400 Mgz
>>>   Control time= 40/1- Average time=1.5min=90s
>>>
>>>		HT=2.0*400*90=90000Kb= 90Mb
>>>
>>>The program, in some moves, will spend a bigger time than the average
>>>time, then I think that is always good to use a bigger hastable than
>>>the calculated  for the equation above.
>>>
>>>Best regards,
>>>Paulo Soares, from Brazil.
>>
>>
>>Well Johann suggest that a hash table of no bigger than 16meg should be used for
>>CM6K. So I think you equation is off in the case of CM6K. I would not even
>>consider using  a 90 hash table even on my PII 450.
>
>I think that he said it only for pentium200MMX
>and i f this is the case maybe a similiar formula can be for chessmaster but you
>should multiply the optimal hash of Fritz by 16/72 so you get for
>Processor with 200Mgz
>
>Control time=40/2- Average time 3min=180s
>
>HT=2.0*200*180*16/72=16000K=16Mb
>
>Uri


Uri,

Why to multiply by 16/72? I don't understand.

Paulo Soares



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.