Author: Vincent Diepeveen
Date: 17:25:41 10/20/03
Go up one level in this thread
On October 20, 2003 at 15:40:43, Frank Phillips wrote: >On October 20, 2003 at 15:14:25, Gerd Isenberg wrote: > >>On October 20, 2003 at 14:57:26, Frank Phillips wrote: >> >>>On October 20, 2003 at 13:44:15, Vincent Diepeveen wrote: >>> >>>>On October 20, 2003 at 13:32:53, Michael P. Nance Sr. wrote: >>>> >>>>at the dutch championship yesterday there was a small discussion on this. >>>>fritz programmer was also involved (frans morsch) and the conclusion was soon >>>>that this was the hardware of the future. >>>> >>>>you don't have to worry. it's a matter of a short time before there will be 64 >>>>bits versions of the different programs. >>> >>>bitboards ? >> >>even pure 32-bit programs profite a lot from 64-bit mode, and even none >>bitboards programs may use some 64-bit stuff for hashkeys or some pawn >>bitboards. Note that int or even ms-long is still 32-bit. Most important feature >>is the doubled register file size. >> >>Gerd > >Yes. > >I imagined the '64 bit versions' related to the program data structure, not the >chip. > >I guess most hash stuff is already 64 bit (or > 32bit) at the moment. > >Frank The biggest win will be not 32 to 64 bits so much, but for majority of programs going from 8 to 16 registers and using 16384 entries for the branch prediction table is going to kick major butts. That last already in 32 bits. See diep being tested in 32 bits at A64 2.4Ghz it's 50% faster single threaded than a P4-Xeon MP-EE 3.4Ghz 2MB L3 cache, which is not even at intels OWN road map, nor getting sold at all. Probably the biggest proof that it was an emergency try of them. So the 64 bits is just a small part of the speedup. it's already 50% faster in 32 bits for me :) See the extensive review at aceshardware.com
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.