Author: Carsten Kossendey
Date: 19:26:15 01/24/98
Go up one level in this thread
On January 24, 1998 at 20:11:39, Jay Scott wrote: >Then I tried MacChess 4.0. Searching to depth 7 took it less >than one second and didn't give an accurate speed, so I went >to depth 9. It reported ~250K nps. From my experience, that's >at the lower end of its normal range--it's what you'd expect >in this position. > >Considering chip performance alone, I'd expect crafty to run >rather faster on a G3 Mac than on a PPro 200. Memory bandwidth is what matters for Crafty. This is why it runs that fast on the PPro (with 1:1 cache clock) and on PeeCees in general (due to 66/75 MHz memory buses). >But Bob reports: >>same ballpark as pentium pro 200... > >MacChess is definitely not as smart a program as crafty, but >I don't think it's 3.7 times dumber. :-) MacChess' NPS rates are generally a factor 2 or 3 faster than anything else. There is *no* knowledge in this program apart from piece/square tables and a little pawn structure eval, I suspect. >And of course it's >optimized for the PowerMac, while crafty is optimized for an >Intel box. I can't prove anything, but from the numbers I >suspect that crafty's taking a big hit on the Mac from some >compiler deficiency or sneaky implementation detail. The pain is hundreds of megabytes being pumped over a painfully slow memory bus. Compile it under Linux with gcc on a PowerMac and then take an IBM RS/6000 (with teh same processor) and AIX or Linux. The RS/6000 will easily outperform the Mac due to its memory bandwidth being like 60% higher, or something. Of course if you have a G3 upgrade card with selectable bus timing, you can try to run it at 66 Mhz (or even more) as well, which will help a lot. > Jay
This page took 0.01 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.