Author: Vincent Diepeveen
Date: 15:01:35 01/20/01
Go up one level in this thread
On January 20, 2001 at 09:17:12, Dan Andersson wrote: >And RDRAM might be suboptimal for chess programs anyways (atleast until the next >implementation anyways, Samsung I think)), so this might be a lose-lose >situation. For most programs it will slow down, but for DIEP i might be real lucky with RDRAM single cpu that is as i get 8 probes of 16 bytes out of RAM which is 128 bytes. when running parallel of course RDRAM is a disaster. at 1.5Ghz losing 150 clocks for a single lock is a bit much.... What i do not know is how much faster diep gets because of the 4 integers a clock the P4 can handle. All other things are not favouring it like the small caches and the huge branch misprediction penalty. It might get 25% faster but also 2 times slower, i don't know! for sure newere releases of the P4 with 'bugfixed' architecture will kick the butt of any P3 for my program. Whatever speed they get however, the 100Mhz basic speed of the rdram is going to be the slow factor in the end. 8kb L1 data cache and 256kb l2 cache are not going to make up for the slow latency of RDRAM >Regards Dan Andersson
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.