I used to participate in the F @ H project. Havn't for a while noe though. Just wondering something. You know how you can run Prime95 to see how stable your PC runs when overclocked, and when its unstable it makes errors in Prime95? Well shouldn't the same apply to F @ H, seeing as it often stresses out your CPU a lot, especially if you leave a computer on for a very long time, dedicated to F @ H... And what if the CPU is overclocked... It would create errors, like in prime95. So couldn't an overclocked / stressed PC ruin a F @ H project, giving back false results? If so, how would this effect the entire project as a whole? If so, F @ H seems like a giant waste of hardware, bandwidth, and electricity to me.
Running 24/7 @ 100% shoud'nt affect how stable a computer is, if it does then there's a problem (better cooling, windows needs restarting again []). Each work unit has to be verified, ether the same work units are sent to multiple computers or there's is a fast work to check if the result was correct (carried out by the server, or sent to a diff computer to check!) or even both, which would make sense as they obviously cant trust the nodes they send work to!
Indeed, there could be errors in the computation process if the CPU is overclocked and unstable. But the main point of F@H is to harness the power of everyday, non overclocked desktop computers. Redundancy is provided when multiple computers do the same work unit, so any errors can be identified. This is probably why they prefer the work unit to be completed in a time period. F@H in my opinion, is more useful than something like P95 or einstein@home.
