Author: Vincent Diepeveen
Date: 06:40:52 07/18/02
Go up one level in this thread
On July 18, 2002 at 09:38:58, Vincent Diepeveen wrote: small note. Rambus is 1.5 slower in latency than SDRAM. So i am very sure that would *not* speed you up. the bigger bandwidth you don't need of course. that's because sdram is 32 bytes cache line length versus rdram has 128 bytes. that's how they managed to get a theoretic better bandwidth :) >On July 18, 2002 at 09:11:09, Thomas Mayer wrote: > >unless you coded your thing in MMX, which would make it more than >factor 1.7 slower for the P4, as the P4 only has 1 MMX decoder, >versus K7 has 2, then you sure are about 1.7 times slower as well :) > >>Hi Vincent, >> >>> divide by 1.7 for chessprograms to have the real speed of it in AMD K7 Mhz. >> >>You know that this depends on the engine... As much as I know about Diep (And I >>do not know much... :) your 1.7 could be quite possible... I have tested here >>with a PIV-1600 (SDRAM) -> and for Quark the factor was 1.53, so a PIV 2400 >>would be around the same speed then my current Athlon XP 1800+ (1566 MHz.) It is >>possible that the PIV 2400 with RAMBUS could outperform my Athlon, but I doubt >>that due to the structure of RAMBUS which is not so perfect for chess... When >>discussion comes to speed / price Athlon is still the best choice... >> >>Greets, Thomas >> >>P.S.: Besides, I have also tested on an PIII-1266 here - it was just 1.13 slower >>then my Athlon, so in speed per clock it is a real monster... to bad that intel >>has given up the PIII-structure... But for other engines this might differ >>also... a possible reason that this PIII-1266 performs so well might be, that it >>has the big on die cache with 512 KB...
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.