More of a nuisance than anything. Pentium IV, 3.0 GHZ, 512 MB RAM, HT, BOINC 4.45 (now 5.1.8) WinXP Home. Attached to CPDN, SETI, Rosetta, and Predictor.

Downloaded a Sulphur and started running it. After about 24 hours, time to complete stabilized at about 1400 hours, projects swapped in and out, and all was well with the world. After a computer reboot, time to complete went about 2400 hours and never came down again (as it had done a couple of times previously).

Well, this caused the scheduler to go nuts. It went into Earliest Deadline first and both logical CPUs ran the other projects until completed. Then the CPDN got a CPU and ran until the last WU was run on the second CPU. When that CPU ran dry, it downloaded a couple of work units and promptly went into Earliest Deadline first mode and ran those new work units. This went on until I got tired of watching it. As I understand the scheduler, it is pretty much exactly what it is supposed to do (some time ago, there were admitted problems with dual CPU scheduling... I guess this is one of them). While this was going on, I installed 5.1.8 -- it is behaving the same as 4.45 did.

In order to work around this, I told the other 3 projects \"No New Work\", let them finish, turned HT off in the BIOS, and am running CPDN only for a while. Time to complete is now down to 2350 or so. If I can believe the percentage complete, 22 hours of CPU time changed the percentage complete by 1.98%. So a rough estimate to complete would be about 1200 hours (which is roughly consistent with the original estimate).

Now to see if running the benchmark in 5.1.8 kills it.