I have another dual core single proc server that is currently running 2003 server with 2 clients, if my email upgrade project that these servers are for gets delayed any longer, I may load Linux on that one and try out the SMP client on there to. With less cores but the same memory maybe it won't get the 1400 WU's.
That should be a good data point to have for the 1440 pointer question.
Also, I've noticed using the system monitor on Linux that NONE of the cores stay at 100% but they are in constant flux between 50-90%. Anyway I could take advantage of those extra cycles and run another client? Will the standard Linux client run on 64bit?
I have no idea what's really true, but I've thought that it might just be the granularity of the system monitor's updates that's causing that result. I checked mine again just now and (ordering the list by CPU time) it's showing 217, 121, ~118, and 96 minutes of CPU time for the 4 folding core executables. The next highest thing on the system is gnome-panel at 27.2 seconds, then gnome-terminal at 25.89, and metacity at 19.74 and on down to 0 seconds. So I think it's doing reasonably close to all it can do. We can get rid of some of that cycle usage from the gui software, but I think that's about all. notfred
's remark in one of these threads would seem to indicate that may be a coming possibility for us.