Author: Bertil Eklund
Date: 09:50:46 01/29/99
Go up one level in this thread
On January 28, 1999 at 19:32:42, KarinsDad wrote: >On January 28, 1999 at 18:39:04, Bertil Eklund wrote: > >>On January 28, 1999 at 12:21:35, Enrique Irazoqui wrote: >> >>>On January 28, 1999 at 12:07:44, KarinsDad wrote: >>> >>>>I've noticed a rating increase from the Pentium 90 MHz to the P200 MMX (as >>>>expected). This implies that a doubling of the speed (plus bus improvements, >>>>minus hash table changes, etc.) results in approximately 70 points in increased >>>>rating as per the derived chart below (these being based on a x2.2). >>> >>>It is more like a 3x between a P200MMX and a P90, which means than doubling the >>>speed translates into some 50 points more. >>When we upgraded to P200MMX from P90 I run a test with about ten programs snd >>the average speed-increase was 2,4x. > >I was wondering if an increase in the hash table size (which seems to often be >the case on the P200MMX) would offset some of the speed improvements with the >bus and processor. So at 2.4x, we would have an average of about 65 rating >points (rough guess) per doubling (other things being equal). > >Are the nodes per second used to calculate the increase in speed or is some >other metric used? If nodes per second, do those totals include re-searches (I >would assume so)? > >KarinsDad I choosed a fixed depth and played a move out of book i.e. 1.h2-h4 and compared the times. But you are right the difference between P90 and P200MMX is about 75 elo. The hash-tables has increased about 4 times I guess this could give 10 -20 points. So the cpu difference could be about 60 elo. Bertil > >> I also checked the difference with 256k >>cache and 512k the difference was just over 1% (from 0-3%). >> We are just now testing P2-450, Celeron450 and AMD to find out the speed >>improvement vs P200MMX. Several programs are about twice as fast but Nimzo98 >>is nearly 3 times faster! (P2-450) >> >>Bertil >>>Enrique >>> >>>>Inc. Program >>>>87 Hiarcs 6.0 >>>>82 Rebel 9.0 >>>>70 Rebel 8.0 >>>>91 MChess Pro 6.0 >>>>69 Genius 5.0 DOS >>>> >>>>Does anyone have a more detailed list of this type of information? >>>> >>>>How do the non-commercial programs fair in this regard? >>>> >>>>Is there any sort of curve associated with this from previous years (i.e. is the >>>>rate of this increase decreasing from previous years/earlier processors, or is >>>>the rate of increase remaining somewhat constant)? >>>> >>>>Has anyone done any work in increasing the time per move in order to emulate >>>>future processor types? >>>> >>>>Robert Hyatt has mentioned in a earlier post that on his quad, Crafty is running >>>>at approximate 1600 Mhz equivalence. Does this mean that Crafty run there is >>>>about 210 points higher (8x faster = 2^3 or 3x rating point increase) in rating >>>>strength than Crafty running on a P200 MMX? Or is the rating difference >>>>considerably less than that? >>>> >>>>Thanks >>>> >>>>KarinsDad
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.