Author: Robert Hyatt
Date: 07:38:16 09/06/02
Go up one level in this thread
On September 05, 2002 at 10:59:17, Vincent Diepeveen wrote: >On September 04, 2002 at 21:19:05, Robert Hyatt wrote: > >The K7 smp chipset has about 600MB/s bandwidth. >I do not know numbers for P4 duals here. It can't be much >more unless L2 cache plays a role here. > >Interesting to know is the L2 cache size of the Xeons which >were tested by Nalimov. I suspect 512K, but he can answer. The quad 550 GCP is using has 512K L2 xeons. My quad 700 has 1024K L2 xeons. That might explain my slightly better speedup numbers over the run he made, I don't know... > >>If anyone has the time, and a dual-cpu machine, would you run the following >>position to depth 13 using first one cpu, then restarting, and running it again >>with two processors? Everything else at default values. >> >>2r2rk1/1bqnbpp1/1p1ppn1p/pP6/N1P1P3/P2B1N1P/1B2QPP1/R2R2K1 b - - 0 1 >> >>That is kopec position 22, one of my favorites. I am only interested in >>two numbers, the raw NPS for 1 cpu, and the raw NPS for two cpus. I don't care >>about the times or anything, just the NPS... >> >>Please include your cpu/speed/vendor/etc... >> >>Vincent thinks that the 2-cpu test will slow way down in terms of NPS. I >>can't reproduce it on my machines here. Eugene can't reproduce it on Intel >>boxes, but the two AMD machines he has tried produce 1.4X the nps using two >>that it produces using 1, while my machines produce about 1.9X the nps... >> >>Thanks...
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.