Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A question about the pentium4

Author: Bo Persson

Date: 04:17:52 04/20/01

Go up one level in this thread


On April 20, 2001 at 03:25:19, Uri Blass wrote:

>I talked with a friend who told me that he intends to buy pentium4.
>He is using his computer mainly for chess.
>
>I told him that PIII is better for chess programs.
>
>He did not believe it and when I told him that people found it about the chess
>programs of today he told me that the reason is probably the fact that the
>programmers of chess programs are from the old generation and do not know how to
>use the pentium4 effectively.
>
>He asks how is it possible that the pentiumIV do things faster when these
> things are irrelevant for chess programs.

Do you mean, why didn't Intel optimize the processor for chess programs?  :-)



>
>I have also some questions
>
>1)Is it possible to write chess programs that are optimized for the pentium4 and
>not for the pentium3 so pentium4 is faster than pentium3
>with the same megaherth speed?

Yes, the performance will definitely improve as soon as the chess programmers
can afford to buy new machines. Highly optimized programs are generally tuned to
the machine (or processor family) used by the developer.

On the other hand, the processors are *not* run at the same MHz!

>2)Is it truth that the pentium3 is faster for integers when the pentium4 is
>faster for using variables that are not integers and using integers is always
>better even on pIV?

It seems like the first version of the P4 is somewhat crippled by the current
manufacturing process, so that Intel's designers could not include all the
features that they had originally planned. From what I have seen, the P4 cannot
fully use all its execution units (like the dual, double-clocked integer units),
because of a limitation in the number of instructions that can start each cycle.
So while it could theoretically *execute* 4 integer instructions per clock,
other parts of the CPU can't keep up with that.

This is an imbalance between architecture and implementation, and might improve
in a future revision.

The P4 is at its very best for programs using its new SSE2 instructions, where
it can execute several floating-point instructions in parallell. This works
extremely well for some graphics intensive programs, but I don't see how we can
use it for chess programs.

>Uri


Bo Persson
bop@malmo.mail.telia.com




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.