Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 8 way processor

Author: Robert Hyatt

Date: 18:39:22 08/22/02

Go up one level in this thread


On August 22, 2002 at 20:45:33, Eugene Nalimov wrote:

>Several days ago I run Crafty on 3 different dual systems. I measured only nps
>reported by 'bench', not time to solve something, so resulting numbers show how
>efficient dual support in the system is implemented.
>
>PIII/800:    1.9x
>P4/2GHz:     1.86x
>Athlon/1.67  1.4x
>
>I.e. Athlon was much faster than P4 using single CPU, but became slower when
>both CPUs were used.
>
>Thanks,
>Eugene

It definitely varies.  I once had a dual 300 that was exactly 2x as fast
as using one cpu.  I didn't try to study the chipset to see what was going
on however, but it was supposedly some sort of server-specific MB that
was a good performer...

what are the details about the athlon?  IE FSB speed, etc.  Obviously going
faster on the FSB is not so great if memory stays at the same speed...



>
>On August 22, 2002 at 20:15:15, Robert Hyatt wrote:
>
>>On August 22, 2002 at 18:04:21, Rajen Gupta wrote:
>>
>>>On August 22, 2002 at 17:16:18, Robert Hyatt wrote:
>>>
>>>>On August 22, 2002 at 15:18:37, Gian-Carlo Pascutto wrote:
>>>>
>>>>>On August 22, 2002 at 15:17:55, Rajen Gupta wrote:
>>>>>
>>>>>>thanks: sorry for the obvious quesion but how fast is a 4 way compared to a
>>>>>>single or a 2 way(i assume a 2 way is about 1.75 times a single)
>>>>>
>>>>>4 way is about 3 times faster than a single
>>>>>
>>>>>--
>>>>>GCP
>>>>
>>>>
>>>>I don't think he was talking about chess.  He was talking about number-crunching
>>>>in general I assume.  So the right answer is nearly 1.7 for a dual, nearly 4.0
>>>>for a quad, and nearly 6.8 for an 8-way box.
>>>
>>>sorry: i was actually talking about chess
>>>
>>>rajen
>>
>>
>>In that case the answer varies significantly.  3.0 (as GCP gave) is a good
>>rule of thumb.  But then the other numbers I gave are too high.  IE a dual
>>is about 1.7x faster than a single before considering chess.  It would be
>>about 1.7x faster playing chess if memory was not an issue.  As a result,
>>a dual may well be slower than 1.7x for chess, as it depends on the memory
>>system.  dual servers may well have interleaving, while dual desktop machines
>>typically don't to keep cost down.



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.