Author: Günther Simon
Date: 07:24:49 05/26/05
Go up one level in this thread
On May 26, 2005 at 07:11:26, Steve Glanzfeld wrote: >On May 26, 2005 at 05:03:01, Robert Hollay wrote: > >>Just out of curiosity: how much stronger would be Rebel on a today's top PC >>then it was on those 1998's one? > >As a very rough estimation, you can calculate the speed doublings for typical >CPUs (for example: 500 MHz -> 1 GHz -> 2 GHz are 2 doublings), and multiply with >30...50 Elo. It is not clear which value is more near to the truth, against >humans. Among comps, even values up to +80 Elo for twice the comp speed have >been used with good correlation. Against humans, more speed has less impact. > >Of course, that means the performance gain with exactly the same software >version on faster hardware. > >Hardware experts will be able to explain that a modern 2 GHZ CPU is more than 4 >times faster than a 500 MHz P3, but that's why I write "very rough estimation." >That uncertainty is unimportant because we have to guessimate :-) the Elo gain >per doubling, anyway... I just want to add that it is vice versa, at least for PIII to PIV machines. A PIV 2Ghz e.g. is much less than 4 times faster than a PIII 500Mhz(in kn/s). I don't remember though, why the PIII has a relative advantage... (Here, I could figure out e.g. that a PIII 650 is just ~ 3.3 times slower than a PIV 2.66 Ghz instead of the expected 4.1 times (if linear) - of course both machines are well tuned and configured). Guenther > >Btw. the improved Rebel program is available for free, as an Engine for WB./UCI >capable GUIs, under the new name "ProDeo," from > >http://members.home.nl/matador/prodeo.htm > >It is a Top-10 engine. > >Steve
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.