Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka @ 64 bit

Author: Vasik Rajlich

Date: 01:36:52 03/27/04

Go up one level in this thread


On March 27, 2004 at 04:27:42, Russell Reagan wrote:

>On March 27, 2004 at 04:14:02, Vasik Rajlich wrote:
>
>>Rybka 1.3
>>---
>>Athlon 3400+, 32 bit win XP: 1,201,030 nps
>>Athlon 3400+, 64 bit win server '03, 32 bit .exe: 1,201,030
>>Athlon 3400+, 64 bit win server '03, 64 bit .exe (win DDK compiler): 1,506,580
>>
>>The speedup is around 20%.
>>
>>There is a small hopefully-temporory caveat, something is choking the UCI output
>>under the 64 bit OS. When Rybka sends all of the UCI "info XXX" commands to the
>>GUI, the nps rate drops by about 20%! The above is with all UCI "info XXX"
>>commands disabled.
>>
>>Vas
>
>Is Rybka bitboard based or array based?
>
>Also, do you have a comparison with Rybka running on a 32-bit machine? It would
>be interesting to know how much of a speedup you get in total when moving from a
>32-bit Athlon to the 64-bit Athlon. I have a 2GHz Athlon. If you would like for
>me to see how fast it runs, just send me the executable and I will run it. Then
>we could compare the 32-bit machine vs. 64-bit machine speedup as well.

Rybka is bitboard based, including rotated bitboards.

Actually I took the bitscan from a post of yours here some time ago. (I used
"Gerd's" algorithm - Eugene's was the fastest, but I was concerned about the
cache footprint.)

A slightly older (and very slightly slower) version of Rybka gets 797,650 NPS on
a pentium 4, 3.06 GHz. The speedup to AMD64 @64 bits is almost 2:1! I don't have
exact Athlon XP numbers, but something in the neighborhood of 600 KNPS on 2.2
GHz XP.

Ok, I can send you the executable on Monday morning. It would be interesting to
get some exact numbers.

Vas



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.