Author: Yen Art Tham
Date: 09:33:43 08/23/02
Go up one level in this thread
On August 22, 2002 at 17:13:47, Robert Hyatt wrote: >On August 22, 2002 at 16:25:10, Yen Art Tham wrote: > >>On August 22, 2002 at 13:43:06, Robert Hyatt wrote: >> >>>On August 22, 2002 at 13:38:17, Rajen Gupta wrote: >>> >>>>in practical terms how much faster would an 8 way processor be for the most >>>>scaleable commercial chess programme which i believe is shredder? >>>> >>>>rajen >>> >>> >>>If it turned out to be 1.5X faster than a 4-way that would probably be >>>a huge plus... memory is a problem on 2-way and 8 way boxes... >> >> >>In what way is memory a "problem" on a dual? >>Please elaborate. >> >>yat > > >Memory bandwidth is severely limited on a normal PC. On a dual, you >simply add a second processor, which then competes for memory bandwidth >with the first processor. As a result, a dual machine is generally not >2x faster than a single. > >For the quad (4 cpu machines) the manufacturers use a more sophisticated >chipset that provides 4x the memory bandwidth by doing 4-way memory >interleaving. As a result, on a quad, each CPU sees about the same >bandwidth potential as a single-cpu machine sees. > >On an 8-way box, the chipset still relies on 4-way interleaving, but you >have 8 processors. IE it will perform about like 4 dual-cpu machines, >which is not quite as good as you would expect from 8 cpus... Would the problem be solved with the Opteron which has integrated memory controller? 2 cpu, 2 memory controller.
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.