Author: Robert Hyatt
Date: 10:59:55 03/28/01
Go up one level in this thread
On March 27, 2001 at 17:42:40, Christophe Theron wrote: >> > >I had overlooked that it was 450PII single against 800PIII dual. > >For a factor 2 speedup the rating increase is 53 ELO, read my message. > > > Christophe That is why the scientist in me cringes at some of these posts. If I told you I put unleaded premium gas in my car, where I was using unleaded regular, and my gas milage went from 20mpg to 24mpg that would be interesting. If I forgot to tell you that I also replaced all the filters, spark plugs, plug wires, distributer cap/roter, etc, that would make that interesting statement very confusing. Which of the changes made the biggest difference in the milage? As a result, I will _always_ give 1 vs 2 vs 3 vs 4 processor tests using the _exact_ same _everything_. Same memory, same hash sizes, same positions, same exact processor (ie you can't even compare a PIII XEON with a PIII coppermine XEON. They are different even if the mhz is the same). I would like to see that same standard of reporting followed by everyone so we don't have to read the fine print, and interpolate results to make them fit known values. IE when Bruce and I were both actively working on the parallel thread stuff in Crafty and Ferret, we compared noted pretty frequently. We used the exact same test positions, same time limits, etc. So that we could see how we were doing. If I do tests using real games, while Bruce uses the Bratko-Kopec positions, it is _very_ hard to compare the results and conclude anything other than "they don't quite match..."
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.