Author: Robert Hyatt
Date: 13:18:07 07/17/04
Go up one level in this thread
On July 17, 2004 at 11:40:01, Sune Fischer wrote: >On July 17, 2004 at 11:04:51, Robert Hyatt wrote: > >>That's not my problem. IE he ran on a 64 bit processor. So _hardware_ >>advantage is easy to compute. If he didn't compile it correctly, there's little >>I can do about that. > >I'm not saying it is your problem and I'm not saying it's unfair either. >That whole discussion doesn't make much sense to me as I see the job of getting >access to a huge machine as simply part of the contest. > >Still, worth noting that Omid has some "easy" ways of getting a big jump in Elo >for next time, whereas you probably cannot be expected to squeeze that much more >efficiency out of Crafty's parallel search. > >> But the raw _hardware_ advantage is easy to compute. Also >>the 1.4 is wrong, because gcc is worse than microsoft's compiler by at least >>10%. > >I think 1.4 is a rather pessimistic estimate. If you had a good compiler then it >would probably be closer to 1.5-1.6, so in some sense the 10% has already been >factored out. The correct number is around 45%, gcc 32 bit vs gcc 64 bit, or msvc 32 bit vs msvc 64 bit. I believe AMD has already published these numbers in fact... And Microsoft's compiler is at _least_ 10% faster than GCC. More can be said by Eugene if he wants... I was simply pointing out that there was absolutely no way a quad is 4x faster than a 1-cpu box. And my quad system was actually running about the same speed as the quad 2.2ghz box I used at the last CCT event... > >But anyway, I consider 10% to be within the error margin with these types of >ballpark numbers, they tend to vary quite a bit when you measure IIRC. > >-S.
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.