Author: Dr. Gregor Overney
Date: 22:57:36 11/01/98
Go up one level in this thread
Thanks for the numbers. Did you enable the 1MB cache? How many other programs/utilities did you run simultaneously when making this benchmark? - It seems to be pretty slow for a PowerPC chip at 350MHz and such a large L2 cache. Remember, who ever made this "normalization" was running it on an older Linux box. It is roughly 85k nodes per second. - On NT4, an old p6/200 (L2=256 kBytes only) scores 1.17. A dual p6/200 on NT4, scores 2.15 (using mt=2). And those are really not the fastest systems these days. Using a PowerPC RISC chip myself, I am a little bit surprised that you could not crank out more performance. Did you (or who ever) use full optimization when compiling Crafty? Is 15.4 so much slower than 15.20? Unfortunately, I do not use C on my PPC and therefore cannot try it out myself. But I think, that Crafty should perform better on a PPC since it performs very well on other RISC based CPUs with large caches, such as the Alpha 21164/566 with 4MBytes L3 (96kBytes L2). Btw., according to an interview in 97, Bob's favorite machine to run Crafty is the Alpha. Gregor On November 01, 1998 at 15:47:26, Frank E. Oldham wrote: >On a prototype, with 350MHz G3, 1Mb of 175MHz L2 cache, and 100MHz system bus, >MacOS 8.5, the pre-built crafty 15.4 executable generated: > >Crafty v15.4 > >White(1): hash 16M >hash table memory = 12M bytes. >White(1): hashp 10M >pawn hash table memory = 10M bytes. >White(1): bench >Running benchmark -- this will take approximately six minutes. >...... >Total nodes: 56873894 >Nodes per second: 157983 >NPS relative to Pentium Pro/200: 1.86
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.