Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Checkers: Las Vegas and Chinook

Author: Dave Gomboc

Date: 22:06:47 09/17/02

Go up one level in this thread


On September 17, 2002 at 15:43:49, martin fierz wrote:

>On September 17, 2002 at 01:30:27, Dave Gomboc wrote:
>
>>On September 16, 2002 at 14:25:06, Dave Gomboc wrote:
>>
>>>On September 16, 2002 at 03:45:03, martin fierz wrote:
>>>
>>>>oh - too bad! i'd have been interested - i'm sure chinook would have solved this
>>>>faster than my engine, but i'd have liked to know how much faster. i've found a
>>>>problem in my evaluation which hurts in this particular position - i'll have to
>>>>try and fix it. thx for trying anyway!
>>>>
>>>>aloha
>>>>  martin
>>>
>>>I will ask again when it is done.  The code is being upgraded to allow the use
>>>of the 10-piece tablebases.  (Not all of them are generated yet, but it can use
>>>the partial set that has been computed.)
>>>
>>>Dave
>>
>>Oh, one more thing, Martin.  I asked about the processing power of the SGI
>>machine that was used in 1994-95.  Jonathan told me that he got about 100K
>>nodes/second, and we (back-of-the-envelope :-) figured that the ballpark
>>Intel-based PC equivalent to that hardware is about a Pentium (I) 150 MHz.  So I
>>underestimated with single-double digit megahertz, but Pentium III 1 GHz is way
>>off the mark too.
>>
>>Dave
>
>did i say 8x150MHz? my memory :-(
>here's something from the chinook website:
>
>"In the Challenge, Schaeffer had found a first-class vehicle for his driver.
>With its sixteen CPU's, each of which can perform 150 million instructions per
>second, Chinook would be able to consider, in a single minute, 12 million
>possible directions in which the game might evolve."
>
>so according to that it's 200kN/s. i have no idea what kind of processors were
>in that challenge, and what kind of speed a single one of those had.
>
>here's something about a pentium I:
>
>"Byte magazine (May 1993) noted that the Pentium had a MIPS (million
>instructions per second) rating of 112 at 60 MHz or 1.85 MIPS/MHz. This compares
>very favorably (about 2-1/3x time higher!) with the 0.8 MIPS/MHz rating of the
>earlier 486DX, although we must keep in mind that MIPS is not an accurate
>pedictor of real world CPU performance."
>
>MIPS is not a good indicator, but the challenge would have had
>16x150 MIPS = 2400 MIPS, which, according to that pentium tech talk above, would
>be equivalent to a 1.28GHz P1. if i throw in a factor of 2 for MIPS inaccuracy,
>i still get 600MHz.
>of course, that's only the "processing power", which in a 16CPU environment gets
>hurt by parallelization...
>
>aloha
>  martin

It's surprising, but Jonathan told me that the SGI chips sucked for checkers
compared to say today's AMD Athlons (I mean clock for clock, he got a lot less
out of the SGI chips... and I'm not sure exactly why that was).  Also there is
the parallelization issue that you mentioned.

Regarding how far the 10-piece databases have been computed, I don't know if
that is a secret or not.  You could always email him and ask. :-)  He doesn't
have them all yet, but he does have enough that he's fixing up code to use them,
so it must be a decent chunk of them.

Dave



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.