I agree with Slappa when it comes to determining which test is a greater stressor of stability with regards to phenom II chips.
I can pass linpack (which is IBT etc) at 3.9GHz or 4.0GHz, tested multiple times through a few different combinations that will let me hit 4.0 with some sense of stability. Some would call that stable, and fair enough it is probably at least somewhat stable. However, running prime95 will crash my system fairly quickly (relatively speaking) which means it is not completely stable, even though it passed IBT.
To 'fully stress' my system, I prefer to use prime 95 if I had to choose one single program. As Slappa has said, once you start reaching the 3 hour mark in Prime 95 it puts a lot of stress on the NB. And while it could be just my RAM, motherboard or some particular interaction/combination of the two with the CPU, I find getting a nice stable NB to be the sole challenge of my OC's past ~3.8GHz. To be fair, I am now actually trying to get my system to pass 24 hours of Prime 95 with my new OC, as opposed to only having tested up to 4 hours with linpack. I'd test longer with linpack, however as every time prime 95 has crashed earlier given an equivalent period of testing time I'll stick with prime 95 for now.
On a side note, the oc listed in a link in my sig is quite outdated, and I've found that by loosening my RAM timings (5-5-5-15~960-980 as opposed to the 4-4-4-12 timings in my link) I can achieve 3.9GHz with a lower vcore (1.5v vs 1.52v) and a much higher NB clock of ~2400. Seems like a worthy trade off to me. It's passed 12 hours at that OC, but it's crashed before 24 hrs. I simply have not updated my sig rig, because I'm still trying to achieve something I'm happy with. If 24 hours proves too difficult, I'll just bite the bullet and settle on 12. And for the record, IBT/linpack only increases my temps by 2-3 degrees at most (usually two) on average over prime 95. Then again I never really have to worry about temps; pumping 1.55v into my system with 20 degree ambient, it never exceeds 50 degrees while running IBT.
Now the point for my rambling side story is that in reality, no system is ever 100% stable, no matter what form of testing you use. The only thing you can prove is that it is stable for however long you tested for, using which ever conditions you used, for that particular test if it actually passed. I could pass 24 hours of Prime 95, and stop it, yet it might crash at 25 hours, who knows? TBH I thought it was fully stable at 12, but curiosity got the better of me and I thought 'if it is really stable, it'd pass 24' and found to my dismay that it crashes at roughly ~18 hours.
Add to the fact that there are testing programs that would make linpack/IBT and prime 95 cry in terms of stress/heat, and you throw in yet more programs to consider. You could use 5 different programs that each utilises a different testing method, run them all for 24 hours, or heck even 72 hours, and it might crash at the 73rd hour. You simply cannot prove that an OC is 100% stable. You can prove it is almost completely stable to a probability of 99.99999% or whichever number you chose, and maybe it is actually 100% stable, but you can never be certain and you cannot prove it. And truth be told, you'd probably never encounter a problem if your system could pass such a ghastly array of stress tests. Once again however, you could still not be 100% certain. Case in point: I know my current OC is not 100% stable since it crashed at 18 hours in prime 95. However it's passed the 12 hour mark more than 3 times now, and so I feel confident in using it 24/7. I've installed absolutely everything on it bar the OS while OC'ed, and I've yet to run into a single issue caused by hardware instability (drives for win 7 is a different story). To me this is acceptable.
So to wrap it up:
Will 3 hours of prime 95 guarantee stability? The answer is no. The fact is however that nothing will, no matter what you stress with or how long you stress it for. So you have to settle for a realistic measure of stability for an OC thread such as this, one that suggests fairly convincingly that the system is stable, yet still be reasonable in terms of testing time. Otherwise everyone would just post their highest OC’s possible, and while it is fun and useful to do so, you couldn’t actually use your system under those conditions most likely. Following Slappa's suggestion, I've found 3 hours of prime 95 to be a great tester of stability with my setup (greater than linpack at least), without taking too long a testing period as to be unreasonable. Once it passes 3 hours for me with a configuration, usually it'll last a lot longer, up to roughly 12 hours at least, in my experience. Your mileage may vary, of course.
The only other practical solution is to allow multiple stressing configurations/tests as qualification for addition to the list. While this may make it easier for some, it would then be difficult (or perhaps impossible) to judge which conditions (such as length of testing) under a certain program is equal to say, 3 hours of prime 95 in terms of stability. And with that said, it is not actually very practical at all when you really consider it, so having one measure of stability (for the purposes of an OC thread) is the best route.
And in my opinion, taken from Slappa,
is that 3 hours of prime 95 is probably the best balance for stress testing you'll find for phenom II chips. Simply because it puts a greater strain on the NB I've found, and OC'ing the NB becomes incredibly difficult when you start hitting ~3.8GHz or higher.
My two cents, or perhaps it was two dollars worth
