Author: Dan Homan
Date: 08:21:36 06/01/98
Go up one level in this thread
On June 01, 1998 at 10:37:12, Ernst A. Heinz wrote: >On June 01, 1998 at 10:20:48, Dan Homan wrote: > >>On June 01, 1998 at 09:21:05, Ernst A. Heinz wrote: >> >>>On May 31, 1998 at 10:17:51, Moritz Berger wrote: >>>> >>>>Fritz gets 23% on the 35 positions in the LCT II testsuite from a >>>>roughly 4-fold hash table increase from 12MB to 98MB (on P233MMX >>>>hardware). >>> >>>Good example for the validity of the generally accepted 7% rule per >>>doubling of transposition table size which I mentioned earlier ... >>> >>>Going from 12MB to 98MB is not a 4-fold but an *8-fold* increase >>>in size which is equivalent to 3x doubling. The 7% rule then predicts >>>a roughly 3x 7% = 21% better performance. Very close to your observed >>>23% for "Fritz5", isnt' it? >>> >>>=Ernst= >> >>Actually, the 7% per doubling leads to (1.07)^3 = 1.225 or a 22.5% >>increase >>which is even closer to the 23% observed :) >> >> - Dan > >Dan, > >I intentionally wrote "... predicts a *roughly* 3x 7% = 21% better >performance." because I assumed not everybody reading this thread >easily understands the correct (1.07)^D for D times doubling. > >=Ernst= I know, I was kidding... hence the :) Of course your approximation is completely correct - especially considering 7% is a rule of thumb... I just noticed that it worked out to 23% (with rounding) and thought it was cute that the rule of thumb predicted Moritz's observation even more precisely than you stated. - Dan
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.