Author: Robert Hyatt
Date: 11:24:41 05/31/05
Go up one level in this thread
On May 31, 2005 at 09:42:12, Gian-Carlo Pascutto wrote: >On May 30, 2005 at 22:05:06, K. Burcham wrote: > >>My point was, it has been my experience that the bench tests are more accurate >>than the hype. > >Bench tests have already been published for the dual cores.... > >Here is one for example: http://www.sudhian.com/showdocs.cfm?aid=667&pid=2543 > >The results are exactly as expected: dual-core is equal to a very efficient >dual. > >-- >GCP The only minor glitch is that on a true dual opteron, there are two memory banks, one for each CPU. A good NUMA code will take advantage of that. On the dual-cores, the two CPUs share one memory bank with the expected memory access conflicts that can occur and there is nothing that you can do there since there is a fixed bottleneck. Fortunately it is not a serious bottleneck with decent cache on the opterons already...
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.