Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Processor's

Author: Aaron Gordon

Date: 18:39:26 06/17/04

Go up one level in this thread


On June 17, 2004 at 14:35:11, Vincent Diepeveen wrote:

>On June 17, 2004 at 12:47:51, Dan Honeycutt wrote:
>
>>On June 16, 2004 at 18:45:02, Aaron Gordon wrote:
>>
>>>
>>>In Quake3 for example.. my Epox 8RDA with a Duron 600 running at 600MHz got
>>>identical FPS to a Celeron 2.4GHz, this was with the same ram and videocard.
>>>
>>
>>I ask from ignorance, but it would seem FPS for Quake is a considerably
>>different animal than NPS for chess - much more memory involved w/FPS.  Can you
>>make a valid assumption on how a processor would do for chess based on how it
>>does for Quake?
>>
>>Dan H.
>
>You cannot.
>
>Better watch aceshardware.com, diep is getting tested there by Johan de Gelas.
>
>He is a very acurate tester. Note he doesn't always mention how many threads he
>tests nor the memory. Default all hardware there has all memory banks filled
>with cas2 memory. Filling all memory banks makes a huge difference for stronger
>chessprograms nowadays.
>
>Slate here is not in objective business, but in overclocking and bad comparision
>business.

What does Slate have to do with this conversation? Or in all your spectacular
brilliance did you forget who wrote the message?

Also, show some results where all filled banks of memory gain a 'huge'
improvement for 'stronger' chess programs. Filled banks won't gain a huge
improvement for the weaker ones? :P They won't gain an advantage for *any* chess
program.

All tests I've done that have full banks vs 1 bank, via, nvidia as well as an
Intel 440BX chipset show that adding banks does not change the latency at all,
only bandwidth. Interleaving was of course enabled. I've tested 0, 2, 4 bank,
etc. Search posts I've made here long ago.

In most cases the increase in performance is extremely small in memory intensive
applications. In chess programs there was absolutely no difference in
performance when improving memory bandwidth such a small amount. I noticed more
of an increase (which was still miniscule) when decreasing memory latency. This
was done via lowering memory timings (2-2-2-5 for example vs 3-4-4-11) and
increasing FSB speed. Also, this was only at VERY large hash size. Anything less
than 256mb had absolutely *0* difference.

I notice Vincent you do next to NO testing of your own. You make ridiculous
claims/statements and then make everyone else work to prove what you said is a
heap of bull****. Why not do some of this testing yourself, like I and other
people here have? Do you not know how? Too lazy? Rather run your mouth and look
like a fool? I find it hard to believe you don't have enough money to just test
some hardware for a short period of time. Perhaps you lack friends to help with
testing?

I am 'poor' and yet I find ways of testing the hardware. I have friends with the
hardware to help test for me, or with me when the hardware required to run the
tests is not available to me first hand. Get some professional 'help' with your
mental condition, start treating people like you would like to be treated and
you'll notice a world of difference in the way people respond to you.



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.