Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A quote from Aceshardware

Author: Aaron Gordon

Date: 15:45:33 06/19/04

Go up one level in this thread


On June 19, 2004 at 06:33:17, Vincent Diepeveen wrote:

>On June 18, 2004 at 08:51:34, Aaron Gordon wrote:
>
>>On June 18, 2004 at 06:39:17, Vincent Diepeveen wrote:
>>
>>>On June 17, 2004 at 21:39:26, Aaron Gordon wrote:
>>>
>>>You asked for example :
>>>
>>>Why filling all banks of a P4C/P4EE cas2 is helping a lot for DIEP you can read
>>>at www.aceshardware.com, hardware reviews from the different processors and
>>>comparing them against each other.
>>>
>>>The speedups reported by Johan thanks to SMT/HT only could get reproduced by
>>>testers with all memory banks filled. If you fill 1 bank only, you don't get
>>>like 21% speedup out of HT at all.
>>>
>>>See www.aceshardware.com
>>
>>Looks like he agrees with me:
>
>>His quote: "No surprises here, as DIEP is more or less immune to faster memory
>>subsystems."
>>
>>Link to page:
>>http://www.aceshardware.com/read.jsp?id=65000313
>>
>>If you notice in the benchmarks, DIEP speeds up 1kn/s (148 to 149) going to
>>lower latency non-registered / non-ecc memory on the same CPU. Now double memory
>
>I said something about:
>  cas2 versus cas3
>
>I said something about filling all memory banks.
>
>Now you say here registered CAS2 memory filling all slots is not so bad?
>
>Why am i not amazed?
>
>You are not contradicting *anything* i said so far.

I've told you time and time again that filling banks of ram does nothing to a
chess programs performance vs 1 bank. Zzzz?

>All the memory used has CAS2 timings (yes especially the registered memory his
>machines have, the manufacturers *really* optimize it well).
>
>All banks are filled.
>
>So that proves my point.
>
>>bandwidth, you'll notice no difference. Only difference I've seen is from
>
>Again and again. Bandwidth is *not* important.
>
>Just random access latency is.
>
>If you want to calculate a random access bandwidth from that... ...fine with
>me...
>
>>decreasing memory/bus latency. Even then it is EXTREMELY small and one could say
>>it is within margin of error.
>
>In case you didn't notice... ...the mainboards are also different...



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.