Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: k-7 Athlon

Author: Gregor Overney

Date: 22:10:16 07/06/99

Go up one level in this thread


On July 06, 1999 at 21:42:53, Alain Lyrette wrote:

>
>On July 06, 1999 at 20:05:41, Gregor Overney wrote:
>
>>On July 06, 1999 at 12:50:15, Alain Lyrette wrote:
>>
>>>Up to now the k-6 was generally the best cpu for chess programs cause of its raw
>>>integer strenth.So now that the k7 is almost on the market i taught that we had
>>>a winner here.Well although it seems like the k7 will beat the pentiun3 by a
>>>large margin in floatting point applications.the integer part of the cpu is
>>>similar to the k6.We WILL see an improvement but only because the cpu can
>>>actually go faster in mhz.(600 mhz and above).
>>
>>
>>"... k-6 was generally the best CPU for chess ...". - This it not correct. What
>>is the SpecINT95 (CINT95) for a K-6?
>>
>>To give you an idea about "raw integer strength", I added a tiny list of systems
>>with only one CPU that are available since months on the market (source from
>>http://www.spec.org):
>>
>>                          SpecINT95        SpecINT95_base
>>Alpha 21264/500 (DS-20)      27.7              23.6
>>PA-8500/440 (N4000)          34.0              30.8
>>PIII/550 Xeon (512 L2)       23.6              23.6  (Intel motherboard)
>>PIII/500                     20.6              20.6  (Intel motherboard)
>>
>>
>>From AMD's recent statement:
>>
>>K-7/600                   SpecINT95_base estimated = 26.7 (?)
>>
>>
>>Remember, the difference between SpecINT and SpecINT_base is the level of
>>optimization. SpecINT95_base is the geometric mean of eight normalized ratios
>>when compiled with "conservative" optimization for each benchmark.
>>
>>How does the K-6 outperform a PA-8500/440 in integer strength again? When
>>comparing CPU's, try not only to focus on Intel's offering. There are still
>>others out there.
>>
>>Gregor
>Gregor,                                                           1)Almost all
>the commercially available programs runs on either dos or window 95/98.Both the
>alpha chip and the pa-8500/440 would have to run these programs on EMULATION
>taking a 35-40% hit.                                               2)99% of the
>people interested in chess programming are the ownwers of either an intel,and
>AMD or a cyrix cpu.Of those,mhz for mhz,the amd is the strongest in the integer
>world.FPU,which is useless for chess,is another matter.
>                                  3)yes the alpha is faster for the rare
>programs that runs on NT.
>4)i'm sure that Deep Blue could outperform AMD too,,,you see my point....

Yes. I get your point.

Your point is to make a statement that the K-6 is the fastest CPU for chess.
Let's step back a little and look at the performance of this K-6 compared to an
old Pentium III Xeon 550 with only 512 KB L2 cache (I do not even consider the
version with 2,048 KB L2 cache).

What is the SpecINT95 of the fastest K-6 that is sold to the general public
(your 99% of users)? I am certain the value will be lower than 23.6. This makes
the Pentium III Xeon 550 MHz a good candidate to be the _fastest_ CPU for Chess
(assuming "raw" integer performance is the main issue here) if we only consider
AMD and INTEL and leave all others out of the game.

BTW., did you ever here of Crafty? It runs well on multiprocessor systems. Check
ICC. It even runs on non-Intel/AMD systems.

Gregor



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.