Author: James T. Walker
Date: 07:07:27 10/31/98
Go up one level in this thread
On October 30, 1998 at 18:16:55, odell hall wrote:
>Hi CCC
>
>
> Yesterday I upgraded to a Cyrix 233 16megs of Ram, I ran the benchmark test for
>rebel 10 and got a rating of 2511, however the internal rating that rebel 9
>assigns itself has actually decreased from 2550 to 2533!! I have noticed That i
>am searching almost three times as many nodes per second than I was with my pent
>120, Nodes per second range from 60,000 to 80,000. Can anyone explain the
>conflict? I might add that the internal rating for rebel10 went from 2555(pent
>120) to 2600 (pent233). Although the benchmark test suggest a much greater
>rating increase. Lastly could someone tell me how many rating pts are affected
>by hash tables? I have read on the internet somewhere that hash tables are not
>that import as far as performance with certain programs. I know that it means
>alot for fritz 5.
Hello,
I thought you might find the following table interesting and possibly relevant
to your question about hash tables.
The position is from the Rebel Benchmark test (Position #3)
r4rk1/5ppp/p2pbb2/3B3Q/qp2p3/4B3/PPP2P1P/2KR2R1- w Rxg7+
Pentium II 333 Mhz
Hash table Time Positions
1 Meg 6:45 35 Million
4 Meg 1:58 10.5 Million
8 Meg 1:36 8.6 Million
10 Meg 1:23 7.4 Million
13 Meg 1:23 7.4 Million
20 Meg 1:32 8.2 Million
60 Meg 1:32 8.2 Million
100 Meg 1:32 8.2 Million
As you can see the difference between 1 meg and 4 meg is very significant. With
no hash tables at all would be even worse. So some hash tables are definitely
a bit help. Of course this is only one position.
Jim Walker
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.