Prime95 is a good stress test for temps and some stability but it's not the only thing you need to test with on these CPUs as some seem to run P95 fine and lock using IE. I'd say a couple of hours on P95, a couple of 4x SuperPi 32M runs, a couple of video benches, some mixed usage (surfing while running 1x P95 and 1x SuperPi) etc. And even then don't expect it to be 100% stable!

For a quick stability test while OC'ing I still like 4x Prime95 large FFT's as that's what seems to crash mine the fastest. Once each app gets to step 3 I move the clock up again. By then the cores have reached maximum temps ...