Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: What are the different Hardwares participating in the WMCCC ?

Author: Chris Carson

Date: 13:14:32 08/18/00

Go up one level in this thread


On August 18, 2000 at 12:42:44, Robert Hyatt wrote:

>On August 18, 2000 at 10:54:32, Chris Carson wrote:
>
>>On August 18, 2000 at 09:21:39, Robert Hyatt wrote:
>>
>>>On August 18, 2000 at 00:16:04, Peter Kappler wrote:
>>>
>>>>On August 17, 2000 at 21:59:37, Robert Hyatt wrote:
>>>>
>>>>>On August 17, 2000 at 17:22:18, Jorge Pichard wrote:
>>>>>
>>>>>>We only know about Dr Hyatt computer since he annouces it 5 days ago. But why
>>>>>>Chessbase does not announces what hardwares they will be using and the same for
>>>>>>other with good possibility of winning such as Chess Tiger 12.9 and Shredder 5.
>>>>>>They need to bring the best hardware possible that can match Dr Hyatt Alpha
>>>>>>system, since Crafty has proven to be playing a la par with the very best
>>>>>>ptrograms using comparable hardwares.
>>>>>>
>>>>>>Pichard.
>>>>>
>>>>>
>>>>>I don't think our alpha is going to turn out to be very fast.  I think we
>>>>>will be doing maybe 400K nodes per second.  A shame, but there was little time
>>>>>to try to get a "real good alpha"...
>>>>>
>>>>>It will still 'play chess' of course. :)
>>>>
>>>>
>>>>
>>>>What kind of Alpha are you getting, and what would you have considered a "real
>>>>good alpha"?.
>>>>
>>>>--Peter
>>>
>>>The 21264A running at 667mhz or faster would be a "good alpha".  The 21264 at
>>>500 is an "OK alpha".  IE we aren't going to be hurting at all, but we could
>>>run probably twice as fast.
>>
>>Here are the Specint2000:
>>
>>Alpha 500       300
>>Alpha 667       424
>>PIII  1G        438
>>Alpha 833       511
>>
>>To bad you could not get the 833.  :)
>>
>>Best Regards,
>>Chris Carson
>
>
>SPECINT doesn't tell the whole story in the case of Crafty.  Alpha is doing
>2x the work of the intel in the SPECINT calculations, since alpha=64 bit ints
>and X86=32 bits.  And then there is the issue of hardware popcnt, etc, that I
>use a lot, but which are not even tested in SPECINT.
>
>My numbers are generally like this:
>
>For the X86, my NPS will be clock/2 roughly.  For the Alpha, my NPS is roughly
>clock*1.5, if it is a 21264A, more like clock if a plain 21264.

Thanks.

Best Regards,
Chris Carson



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.