Author: Robert Hyatt
Date: 08:27:23 05/07/04
Go up one level in this thread
On May 07, 2004 at 03:08:11, Gian-Carlo Pascutto wrote: >On May 06, 2004 at 20:02:50, Robert Hyatt wrote: > > >>>point #2 is that the speedup on 4 CPUs on average is 3.17 in this test, which >>>might be one point for bob in the duel with vincent; although i suspect that the >>>speedup depends on the hardware architecture - i will leave this question to the >>>parallel computing experts though... >> >>The opteron is not as good as a real SMP box. Such as my quad xeons. Opteron >>is NUMA which definitely has "issues" to deal with. > >Rerun the test on one of the Xeons - I know you have enough :) > >I doubt the speedup there will be higher. > >At least we now have some real data, please give us some more :) > >-- >GCP You've had real data in the past, as has Vincent. But for current Crafty, I doubt the xeons are any better, since specific NUMA issues were addressed in the version of Crafty I used to run these results. I doubt they are any better or worse than the xeon numbers. I was simply pointing out that the Opteron doesn't have any parallel architectural advantage over my dual/quad xeons... In fact it is at a disadvantage that requires some programming to fix. That's how I got involved with AMD in the first place, the NPS speedup was not good on Opterons. Now it is almost perfect.
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.