That last little bit


strongly suggests that in terms of result throughput even running 4 cores of the same is counter productive... [repeat repeat] that's why I run a mix. I think I go for upgrading the Q6600 to a Q9650 which is really getting nice price here... and romance the thought that I'm doing better ;>)

I just got one of those monster WUs on my P4 -- 38.27 hours. Validated just fine, but guess I skipped over all the beta WUs for DDDT while this was crunching.

26 hours for me. it was nearly complete, then reset itself when it was at about 89% back to 2%. has this happened to anyone else? its the same project =/

There is a large variation in run times of the individual docking attempts, so it's very difficult to cut the jobs to a fairly regular size. In the last 10 or so HFCC jobs I've seen spreads between 3.5 and 11 CPU hours, wallclock or elapsed shows bigger numbers depending on the time BOINC can use. Would I run these same jobs on my laptop, they'd have run between 8 and 25 hours.

The current average is 7.6 CPU hours, so somewhat over the 7 hours mean target. The techs test a few jobs in production for each batch on the AutoDock using projects, so they have an indicator how long they take and then set the slicer accordingly. That's no guarantee when dealing with non-deterministic calculations... try try try retry to find the lowest bonding energy.

Much of this is caused by hard booting, avoid that. An additional possible remedy is to set the BOINC startup delay. The WCG kit I think puts it to 1 minute so all services can start up before BOINC kicks in... the ever present AV/Security software always on the prowl to see a bug where there is none.

Of course, it obviates to say that I never see this, following the above practices, but do have the occasional fail when Dick and Ing concatenate on my computer via the keyboard.