KTE, is the Large FFT test in Prime a good way to test CPU/IMC? After all the description says it stresses everything. On the otherhand P95 is older than Phenom and thus might actually NOT stress the CPU/IMC/core enough.

I guess Memtest is too much RAM based rather than a good way to test IMC as well.

Im wondering if we get it idle stable for ~10 hours (and load too of course), we should consider it stable? I mean, a reboot for example in 10hours isn't that of a disaster. On the otherhand, it might be foolish to think instability only occurs after a certain amount of time. It might happen as well after 3 minutes after all.

I want to know though, if Small FFT's run on CPU 100% stable, would that mean settings for the CPU alone are fine? So we got to fine tune different stuff like IMC? If not, wouldn't that basicly mean there must be some ratio/link between IMC and CPU for both Voltages and speed?

The longer I think about it the more stupid it becomes. By simple CPU multiplier clocking you can get Small FFT's stable, idle it crashes. Though as mentioned earlier, every 'program' uses it's own algorithms. If you'd manage idle and load stable, even a simple mouse click could cause an instability if it activates a different part of the CPU...

Or maybe it's a build in 'CnQ' thingy that causes crashes... Im out of clues really. Worst part is, we all always considered Prime95 to be one of the most ultimate stability tests which is being rendered completely useless now, and I really hate that because I dont know anymore how to test stability. Of course your guidelines for using Everest benchies for and such it helps a little. But I dont see how that would say whether the CPU, IMC or whatever is instable. A crash could still be caused by anything.