Will FAH produce bad data on *unstable* machine?

[drshivas]

Will FAH return *bad* data if it is running on a machine that is suspect of being unstable? Or does the FAH client detect such a contition and error out (like p95 torture test errors)?

I have a box that won't get past 12- 18hrs of prime95 at certain speeds, YET it will run folding fine 24/7. Does this mean that FAH could be creating invalid results, while not throwing an error (like one of those console errors that stops the client)? Or is the machine ok with running FAH (at least) at its current speed?

FYI, there is a recent discussion here about using folding as a stability test:

http://forum.oc-forums.com/vb/showthread.php?s=&threadid=235542&highlight=folding+stability+test

What do you all think?

[Cowboy Shane]

Make sure that you don't have the forceasm flag turned on if you suspect your system is unstable. If assembly options get turned off, then you know your system is turning in bad units, but really, you are best off just slowing your system down until you are sure it's stable. Bad work units don't hurt stanford much, but they certainly don't help.

[Sterculus]

Usually FAH will exit if some error occurs that corrupts the WU bad enough that it is useless. If there aren't any errors in the log anywhere, then it's good.