Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Effect of RAM speed on engine play

Author: Bo Persson

Date: 12:54:21 06/07/03

Go up one level in this thread


On June 07, 2003 at 15:19:19, Tom Kerrigan wrote:

>On June 07, 2003 at 10:21:38, Bo Persson wrote:
>
>>On June 07, 2003 at 08:47:52, Peter Stayne wrote:
>>
>>>I believe I saw a thread a long time ago on the Winboard forum where Robert
>>>Hyatt was discussing this, but I can't find it.
>>>
>>>Given equivalent CPU capabilities, would an engine on a machine with one of the
>>>new 800MHz P4 QDR buses get a large or small or non-existent boost in speed over
>>>standard DDR at say, 266MHz? or even SDRAM at 100/133?
>>>
>>>My 533MHz QDR operates at the same frequency as even old PC133 SDRAM memory, but
>>>it delivers 4x as much data at each interval. Is the base frequency of the RAM
>>>more important, or the amount of data overall?
>>
>>For a chess program, the access time (latency) is more important than the
>>maximum throughput.
>>
>>I have 2 P4s, with different memory systems:
>>
>>P4 1.7 GHz, dual channel RDRAM, 2.5 GB/s stream throughput (SiSoft Sandra)
>>P4 2.5 GHz, 533 MHz SDRAM, 1.9 GB/s throughput
>>
>>
>>Guess which is the fastest for chess programs!  :-)
>
>Sure, CPU performance is roughly 90% of a chess program's performance. (Maybe
>more?) Memory latency is the remaining 10%. So if you have infinitely fast
>memory, you'll only run a few percent faster, but any change in CPU speed will
>result in a significant difference.

Yes, sure!

My faster box has about a 50% increse in CPU speed, and the NPS is also 50%
higher, even though the memory is supposed to be slower. Memory bandwidth is not
significant, even though Intel pushed HARD for the RAMBUS version earlier.

>
>-Tom


Bo Persson
bop2@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.