Regarding F@H, EUE and overclocking/unlocking

[Plimogz]

Here's the situation. I have a 720BE X3. It unlocks to 4 cores pretty well; in fact, I've run 16 hours of prime95 blend and 6 hours small FFT on it as an X4; never crashes, no remotely obvious problems with it.

however, it fails Intel Burn Test, usually between the loop 15 and loop 30, or thereabouts.

Now, I haven't used it to fold as an X4 because of it failing IBT but I would like to and I probably will, assuming that Folding@Home can detect unreliable results.

So can I expect some kind of heads-up from F@H if the 4th core is in fact thrash? or could errors just slip through the cracks utterly undetected? Or put another way: how good is Folding@Home at identifying computing errors?

edit: I've been looking up what kind of CoreStatus codes the client might throw up if ever it should fail a WU and that's got me somewhat reassured that it should be able determine if the fourth core is flaking out.

edit2: BSOD - ah well, I suppose it wasn't meant to be. Weird thing is, this makes me want a quad-core more than I ever did before finding out about potential unlocking.

Update: Woot! some persistence and the lowering of clocks has yielded nice results: X3@X4 with 720BE default clocks passed 50 loops IBT, Crunched an SMP WU and now frames which were taking 9.5 minutes with 3 cores at 3.4GHz are only taking 6.5 minutes w/ 4 cores @ 2.8Ghz. Good times.

 

[Drsignguy]

In a nut shell, Folding@Home is extremely sensitive to unstable machines. Even if you think your rig is stable, F@H will let you know it really isn't. You have to know that you have done all you can to get your machine stable. This is not advise just for the F@H projects but everything else that may be sensitive as well, like windows! 🙂

 

[Plimogz]

A'ight. Thanks for the reply Drsignguy.
It's good to hear that F@H should be sensitive enough to determine if something is going awry. Now I guess I can consider "folding stable" a valid benchmark and endeavour to achieve that without worrying that I'm unwittingly submitting eff-upped results to Stanford.

Besides which, I'm thinking the psu may be the real culprit here; as it only has a 4pin 12v atx connector instead of the preferred 8. Perhaps the difference between this lesser connector being enough or not lays somewhere between 3 and 4 cores... I wonder.

What is certain however is that with all the time I've spent trying to figure out this extra core I could have earned enough to just buy a couple of authentic quads and be done with it already!