Windows Xp Pro. I've been folding for about 6mo, but still haven't gotten a handle on the point system. It seems rather arbitriary at times.
that's a good platform for folding. (I use it on my rig at home.) I'd have to agree wtih you with the points seeming a bit arbitrary at times. They're roughly indicative of folding time, but not precisely.
Since there are legitimate reasons to dump WUs.. such as WU corruption or client config issues, if the client had an option to dump current WU & notify Stanford before dl'ing new WU(s) - would that solve the assign&reassign WU problem?
Sometimes, a unit gets corrupted. Actually, dumping a unit is pretty easy. (at least on the console version, if you don't care about EMIII, etc.) Just delete the "work" folder and the queue.dat file.
Funny thing is, I can't quite remember what the original problem was.
Oh, I guess it was that a particular protein was very slow on a particular machine. I suppose a few fixes are in order:
1) turn on -forceasm flag -- I believe it will force any CPU-specific optimizations which will speed up the process. If your rig isn't stable with those settings, then your overclock is too fast, anyway, and you probably haven't done sufficient memory and CPU stability testing.
2) turn off the -advmethods flag -- This flag is often used in giving preference to some of the harder proteins. So, turning it off might help avoid those particularly big ones.
3) If a protein is really slow, try to finish it anyway if you can. It's not too fun to throw away 18 hours of folding ....
4) If a protein is slow to the point of stalling, or you can definitely tell you'll never get it done in the quota period, then delete the work folder and the queue.dat file. (Just deleting one or the other will generally result in it trying over or downloading the same WU.) It's better to throw away a WU early in its cycle than later, as you hate to throw away a lot of computational time.
-- Paul