Author: Vincent Diepeveen
Date: 07:27:22 01/29/05
Go up one level in this thread
On January 29, 2005 at 09:26:58, Evgeny Shu wrote: >On January 29, 2005 at 09:10:24, Vincent Diepeveen wrote: > >>On January 29, 2005 at 08:42:36, Evgeny Shu wrote: >> >>>On January 29, 2005 at 08:28:11, Vincent Diepeveen wrote: >>> >>>>On January 29, 2005 at 08:20:07, Jason Kent wrote: >>>> >>>>Your machine will look like a dual machine. The OS doesn't know better than that >>>>it is 2 processor machine. >>>> >>>>>It looks like by the third quarter of this year, both intel and amd will be >>>>>selling dual cores. Are they basically handled as two processors under task >>>>>manager, and software? I'm guessing this is going mean that to get the most out >>>>>of your cpu, you will have to buy all the Deep versions. Maybe that is why SMK >>>>>decided to seperate the programs? >>>>>Jason >>>> >>>>I won't comment on SMK, nor on chessbase, but it's obvious that if the intel CEO >>>>says that intel will produce ONLY dual core cpu's within a year and nearly >>>>nothing else, that only parallel software will work for you. >>>> >>>>Paying extra for something that just uses a normal single cpu, is a very bad >>>>thing. It means basically you have normally spoken a crippled software program, >>>>as > 90% of all cpu's will be dual core. >>>> >>>>Vincent >>> >>>So what we have here is on every "normal cpu"=dual core we buy a half of a chess >>>program ? Maybe a license for every cpu next instead single vs multiple ? >> >>Maybe no license at all. Maybe better is 1 product that supports n cores, >>without limiting n. >> >>Vincent > > > >What about chess programs with 64 bit support , is it another product ? 64 bits support basically depends upon the program. For example for diep it speeds it up roughly 15%. Not because diep is profitting from 64 bits much, but 64 bits also means there is 8 extra registers. There is 2 modes in x86-64 currently: a) x86 which means 8 registers of 32 bits b) x86-64 which means 16 registers of 64 bits More registers is more interesting than being 64 bits for majority of commercial software. Basically you profit from both bit. A major problem is that it's only GCC which we can benchmark now. It is possible the speedup from other compilers is far bigger. Some programs have a few 64 bits functions which are a bottleneck. In such a case the speedwin is bigger. According to specint2000 we can see from crafty that the speedwin when moving from k7 to alpha 21264c the speed win is then roughly 33%. However 21264c also executes at 4 instructions per cycle versus k7 doing 3 at most. So it's not always as clear what the exact speedup is. >So to make it easy : >64 bit - single/multi , 64 bit uci - single/multi , >32 bit single/multi , 32 bit uci single/multi . > >Only 8 separate products , you can't be confused :) Count at 2 executables for engine: 32 bits version and 64 bits version. the 64 bits version does not run at 32 bits processors. >Of course different compile for AMD and INTEL cpus , but that can be included in >one product I think.. intel cpu will profit less from 64 bits. 64 bits with 16 registers means that instructions get longer and data types also in general go with the bandwagon. That means in short that intel has a major problem as they use very tiny L1 caches which already is too tiny for commercial chess products. Vincent
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.