Author: Robert Hyatt
Date: 13:25:09 09/03/03
Go up one level in this thread
On September 03, 2003 at 15:04:09, Gian-Carlo Pascutto wrote: >On September 03, 2003 at 15:00:51, Robert Hyatt wrote: > >>The SMP box will _always_ be faster for any application that shares _anything_ >>at all. Because sharing will _always_ be faster on the SMP box, period. >> >>I don't know how it can be explained any simpler than that. > >Meh. This applies to Crays, apparently. It doesn't apply to real >systems since none I ever worked with seemed to have a direct >connection to memory. Are you saying SMP Intel boxes aren't really SMP? > >-- >GCP They _are_ SMP in the strict sense. Up to 4 processors. Beyond 4, they are somewhat of a kludge. Check out the original "fusion" chipset to see how the 8-way boxes were cobbled together, and no, those are not quite SMP. But for 2-way and 4-way, yes. Bus vs crossbar is a design choice. Bus doesn't scale beyond 4 very well. Crossbar scales to 32-64 but becomes very pricey. NUMA scales way on out but has a routing bottleneck.
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.