Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: DID YOU TAKE OFF THE 100 AT THE END?

Author: Vincent Diepeveen

Date: 10:51:16 05/06/01

Go up one level in this thread


On May 06, 2001 at 09:38:05, Uri Blass wrote:

>On May 06, 2001 at 09:12:17, Vincent Diepeveen wrote:
>
>>On May 05, 2001 at 21:44:48, stuart taylor wrote:
>>
>>>On May 05, 2001 at 08:32:31, Jorge Pichard wrote:
>>>
>>>>According to this performance rating of this match, the previous Deep Fritz had
>>>>a performance rating of 2683. If the newer version of Fritz 7 gain an additional
>>>>45 rating points in strength, plus the advantage of being able to use faster
>>>>hardware for the actual Kramnik match, then we could conclude that Fritz 7 could
>>>>well be rated over 2750.
>>
>>So at a 8 processor Xeon according to this PR nonsense it would
>>be rated like 3000 :)
>
>I do not see how did you get 3000
>The 2683 performance of Fritz also was not tournament time control but at 1
>hour/game.
>
>It is clear that Fritz like other known programs has no chance against kramnik
>and I do not see the point of playing against him.
>
>Deep Fritz is a known identity and I do not believe in the ability of chessbase
>to do big productive changes in Fritz after Deep Fritz.
>
>It will be more interesting to see Kramnik playing against a program that he
>even does not know the name of it when the program can be changed between the
>games when the only thing that he can know is that he is playing against a
>computer and that the organizers decided about the exact program that he plays
>before every game.
>
>These are the only conditions when a match against kramnik may be interesting.
>I believe that even at these condition the public will see kramnik as a
>favourite.
>
>Uri

Oh well we completely agree on the outcome of a match,
nevertheless a 8 processor Xeon is 5.6Ghz in total.

Versus a K6-2 450 is just 450Mhz and has very little RAM usual.

At SSDF programs that run on 2 times faster hardware,
Let's quote the list:

I see fritz 5.32 at K6-2 450 rated at   2547
and the same program at a 200MMX rated  2477

Now we can argue long about relative speeds, but for
Fritz the k6-2 is about 2 times faster. whether it's 1.8 or 2.1
is not that interesting.

The difference is a full 70 points.
For junior 5.0 the difference is 100 points, though
probably a good reason to explain it would be that junior is in C++
and fritz in assembly.

Anyway, let's take the average and round it downwards. 80 points.

Now i do not know how fast or slow a K6-2 with 128mb RAM is for
Fritz, but i know it's dead slow For DIEP. L2 cache is at busspeed
even (100Mhz only!)

Now compare that to the huge shared hashtable at a 8 processor Xeon.
Even if we don't take that into account, then it's not hard to imagine
that with this new xeon core the machine is 14 times faster. Can be 10
can be 16. For me it's not comparable the speed. More like 20 times.
For fritz it's less as that as it's in assembly but it also profits
from the new intel processor bigtime, so 14 times looks like
a cool rounded down estimate.

So if we take deep fritz rating which by hand has been tuned down:
  1 Deep Fritz  128MB K6-2 450  MHz   2650

14 times faster if we take the 2 log out of that. then that's 3.58

If i multiply that with 80 rating points then
  2650 + 3.58 x 80 = 2650 + 287 = 2937

Closer to 3000 as to Kramnik's FIDE rating :)

Now of course it's all a rude estimate. I do not believe speed
brings something for Fritz, but if we extrapolate speed and
search depth then OBVIOUSLY kramnik should lose the match with
BIG numbers.

No one on earth has 2937 right now.

So from my viewpoint there are 2 explanations
  a) Kramnik loses nearly all games in the match and Fritz
     does have 2937
  b) Kramnik doesn't lose the match but wins it bigtime
     and the ratings as calculated by SSDF say in fact nothing
     about human versus computer relationship.

Now b is very obvious IMHO. If we just take into account that
at SSDF only computers play each other we could already know this.

Anyway i did the above calculation again (i remember Bob already
did the calculation quite some times) to show that there are more
factors to consider as a lineair extrapolation of hardware!

Deep Fritz  128MB K6-2 450  MHz          2650    34   -32   470   66%  2537
Fritz 6.0  128MB K6-2 450  MHz           2626    24   -24   897   66%  2512
Junior 6.0  128MB K6-2 450  MHz          2594    22   -21  1109   64%  2490
Chess  Tiger 12.0 DOS 128MB K6-2 450 MHz 2578   27    -27   691   62%  2492
Fritz  5.32  128MB K6-2 450  MHz          2547    26   -26   741   59%  2485
Nimzo 7.32  128MB K6-2 450  MHz          2547    24   -24   857   59%  2485
Nimzo 8.0  128MB K6-2 450  MHz           2539    30   -30   546   58%  2486
Gandalf 4.32f  128MB K6-2 450 MHz        2529   29   -29   584   52%   2518
Junior 5.0  128MB K6-2 450  MHz          2528    26   -25   750   57%  2476
  10  Hiarcs 7.01  128MB K6-2 450  MHz         2526    37   -37   361   48%
2539
  11  Hiarcs 7.32  128MB K6-2 450  MHz         2525    27   -27   679   56%
2481
  12  SOS  128MB  K6-2 450  MHz                 2524   23   -23   925   53%
2501
  13 Rebel Century 3.0  128MB K6-2 450 MHz    2514   31   -31   504   50%   2516
  14 Goliath Light  128MB K6-2 450  MHz       2496   30    -30   546   46%  2527
  15 Crafty 17.07/CB  128MB K6-2 450 MHz      2487   24    -24   857   47%  2505
  16 Nimzo 99  128MB  K6-2 450 MHz             2486   26   -26   731   51%
2481
  17 Fritz 5.32  64MB P200  MMX                2477   19   -19  1338   56%
2437
 18 Chessmaster 6000  64MB P200  MMX         2473    61   -53   184   76%  2277
  18  Hiarcs 7.32  64MB P200  MMX               2473   24   -24   844   59%
2407
  20 MChess Pro 8.0  128MB K6-2 450  MHz      2471   32    -32   474   45%  2507
  21 Fritz 5.0  PB29%  67MB P200 MMX           2459   23   -22  1005   66%
2342
  22 Hiarcs 7.0  64MB P200  MMX                2458   21   -21  1112   55%
2419
  23 Nimzo 99  64MB P200  MMX                  2446   23   -23   885   51%
2438
  24 Junior 5.0  64MB P200  MMX                2433   20   -20  1185   49%
2440





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.