Author: J. Wesley Cleveland
Date: 13:00:27 05/08/02
Go up one level in this thread
On May 08, 2002 at 14:47:09, Bernhard Bauer wrote: >On May 08, 2002 at 11:23:12, Dan Andersson wrote: > >>The clock speed will increase on this CPU. And th IPC also. And the memory >>access will be outstanding. And there will be a really impressive NUMA >>architacture. And NUMA is the big brother of SMP. Or rather SMP is the deformed >>mutated less than viable simile of NUMA. And it will feature full SIMD >>compatability with theintel extensions. And it will run existing code faster. >>And, and, and ... In short, it seems to be way superior to anything ever seen >>before on the x86 scene. It's a beautiful idea of an chip and I am confident >>that AMD will be able to deliver the silicon. This architecture will severely >>disadvantage the current Intel chips. Lets hear it for healthy competition. >> >>MvH Dan Andersson > >I mostly agree. My question is: Will existent SMP code run on a 8 proc opteron 8 >times faster? Or do you need a separate hash table for each proc? >Up to now I don't know of a NUMA chess program. The neat thing about the AMD NUMA implementation is that you can treat it as SMP as the added delays for non-local memory are small compared to DRAM latency.
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.