Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Opteron > 970, Good article

Author: Ricardo Gibert

Date: 14:02:42 06/24/03

Go up one level in this thread


On June 24, 2003 at 14:02:02, Andreas Guettinger wrote:

>On June 24, 2003 at 10:15:48, Dan Andersson wrote:
>
>>http://www.haxial.com/spls-soapbox/apple-powermac-G5/
>
>I very good laugh indeed. The guy in the article compares G5/Xeon results using
>VeriTest compiled on gcc3.3 with Intel P4 results and another Dell WS with
>VeriTest compiled on Intel C++ 7.0 compiler.
>
>
>Compiler:
>Intel C++ 7.0 build 20021021Z
>        Microsoft Visual Studio .NET 7.0.9466 (libraries)
>
>The AMDs may be faster but you cannot compare benchmarks compiled with different
>compilers.

I understand why you think what you do. You want to see how different CPUs
perform running the exact same code. Allowing different binaries to be run on
the different CPUs seem to lose control over a variable you prefer to be held
constant.

This makes some sense, but it has a serious problem. Consider a particular CPU
maker discovers that by making a certain design change, which is a *trade off*,
they can realize an overall improvement in CPU performance. Unfortunately, to
get this improvement, a recompile is necessary, otherwise the improvement will
not appear or even appear as a handicap. What does he do? If you are right, they
should perhaps forego the improvement. Is this what you want?

In your views favor is that it may be a better indicator of real world
performance, since in the real world, we all tend to use the same binaries
optimized for the most widely used CPU, which is Intels.

>
>regards
>Andreas



This page took 0.01 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.