Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Questions re P4 3.03 with HT

Author: Robert Hyatt

Date: 15:00:55 12/05/02

Go up one level in this thread


On December 05, 2002 at 13:44:57, Matt Taylor wrote:

>On December 05, 2002 at 11:50:22, Gian-Carlo Pascutto wrote:
>
>>On December 05, 2002 at 11:06:39, Christopher A. Morgan wrote:
>>
>>>Would there be any
>>>benefit when running single chess application on P4 3.03 with HT versus P4 3.0
>>>GHz without HT?
>>
>>Not unless the program supports multiple CPU's. ('Deep' version)
>>
>>--
>>GCP
>
>Hyperthreading is vary analogous to a normal SMP system, but you won't see quite
>the same gains (and there are other drawbacks, too). Basically hyperthreading
>assumes that both programs do more than one thing and it can therefore extract
>more parallelism from that. If one program is working with the FPU, the other
>program can use the ALU in parallel. If both programs want to use the FPU,
>they'll compete for the same (limited) resource, and it will slow them both
>down. However, total throughput should still be equal or more.
>
>-Matt


The most important part is that while one "thread" is waiting on memory, another
thread
can run like blazes.  This commonly happens in threaded programs and SMT stuff
looks
pretty good so far to me after just a bit of testing.



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.