Author: Moritz Berger
Date: 02:38:31 06/01/98
Go up one level in this thread
The fact that Hiarcs/Mac uses only 256 KB hash tables makes it hard to predict the memory performance (one weak spot of all Power Macs compared with the latest 100 Mhz 440BX Intel systems) for other programs that use hundreds of megabytes for hash tables at these machine speeds. Another sad thing is that the differences in CPU architecture across the Power PC family (e.g. G3 and 604) don't encourage programmers to get the most out of the RISC benefits by hand tuning assembly code for that platform. So some part of the Power potential is likely not used to its full potential. General comparison of G3 and PII machines is not trivial. In most Mathematica and Photoshop detailed benchmarks I know, the G3 is up to 2 times faster on some and is also faster on the average, but the PII scores also significantly better than the G3 on a few tasks. So it's really hard to predict for which chess program which kind of speedup or even slowdown would have to be expected - leaving aside for a moment the fact that e.g. Nimzo, Shredder, Junior, Fritz, The King, CSTal, M-Chess, Genius, Virtual Chess, Rebel and Zarkov are not available for the Mac at all. Finally, there's more X86 than just the PII. Rebel runs best on AMD, and they are also moving forward towards the K7. Intel will introduce 450 Mhz PII CPUs on July 26th AFAIR, there's still no 450 Mhz G3 machine you can buy. And outside mainstream, nobody really dares to challenge the Alphas ... Moritz On June 01, 1998 at 03:21:26, Richard A. Fowell (fowell@netcom.com) wrote: >>Questions: >>1. Why is the difference in scores not reflected in the SpecInteger >>measurement (12.? v.s. 11.?); and which is more relevant to the strength >>of Hiarcs6? > >Actually, the HIARCS 6.0 benchmarks I have >(details at http://www.gambitsoft.com/chess/games/hiar6me.htm) > give "only" a 1.56 factor speedup from PII to G3 at the same MHz) >------------------------------------------------------------------- >Mac G3 /266 266MHz 750 39,934 nps 150 nps/MHz Sys 8, no >extensions >Pent II /266 266MHz PII 25,604 nps 96 nps/MHz > >>3. How much is the performance of Hiarcs affected by the size of the >>backside cache? as well as the speed? > >Well, I have a good benchmark if some HIARCS Mac owner has the >hardware and the inclination. >> >It's really not necessary, anyway - absent the >fact that the Mac version has a fixed hash table size of 256K, >the engines are identical, and my ca. 30-problem benchmark >shows that the chess speed differential between two hardware >platforms doesn't vary much for HIARCS. > >> >>6. Any guess as to what Hiarcs 6 could be rated at if the Byte score is >>correctly construed to effectively be analogous to a "1050MHz PentiumII? >>Approx... > >Well, this would take a recompile of HIARCS, alas. >*IF* the smaller HIARCS Mac hash size is not an issue, >my benchmark indicates the 300MHz G3 is "only" a "467MHz Pentium II".
This page took 0.02 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.