Best programs to test stability of overclocks?

[Aurum9]

For CPU I torture test P95 which I think is the best program for it.
However GPUs aren't so easy. I run a benchmark and my OC seems stable but in game it can crash. Is there any program that can definitively tell me if my OC is stable?

Side note (that is longer than the actual post): Sometimes when I boot my PC my bios gives me an Overclocking Failed error that is fixed by rebooting the PC. Is this as a result of my CPU, or the XMP profile on my RAM? I don't think GPU OC's are tied into your BIOS since I use afterburner. Similarly, this may be related, but the PC also sometimes fails to wake from sleep, BIOS monitor shows a number 40 and screen shows no signal. I have to restart if this happens.

 

[Snipergod87]

Prime95 and OCCT are decent tester's but running a load like a game which is more realistic changes the CPU's workload and can show instabilities. Also running a OC stablility tester for a few hours is what you need to do to really validate the overclocks.

 

[JackNaylorPE]

Problems w/ P95.

1. Most folks use the older version so as not to damage there CPU.

2. What's the point in proving your OC is stable if the test doesn't include the instruction sets contained in P95s newer versions and modern applications. Kind alike testing the ability of your SUV off road by driving down the highway in Miami.

3. P95 hammers the CPU with sequential single tasks which is hardly reflective of today's multi-tasking environment. I have had 24 hour P95 stable OCs fail running the multitasking benchmark in RoG Real Bench.

4. P95 artificially limits your PC by presenting a voltage / temperature laod that your PC will never see again in its lifetime . Now if ya built ya PC to run benchmarks and get ya name on web site OC Leader Boards that's fine; but if ya bought / built ya PC to run applications, then an application based benchmark like RoG RB will present more load than any set of applications you can throw at it. So what might be a 4.8 OC limited by temps observed under P95, might be a 4.9 or 5.0 OC under RoG RB

And yes, GPU does present a bit of a dilemma. First I test w/ Furmark as I want to see the impact on GPU temps as well as coolant temps and how long they take to stabilize. Then I may throw all the 3D mark and Unigine tests at it cause they quick ,,,, but in the end, it always comes down to the game in question.

I have 5 profiles stored in MSI AB....

a) Maximum stable core (stock memory)
b) Maximum stable memory (stock core)

These are to help determine whether any particular game is being tripped up by core or memory.

c) Maximum core / memory combo which yields not the highest OCs but highest fps.
d) A dip down from the above where a particularly game or games trips on above
e) A bigger dip down for anything that has "Battlefield" in the title.

 

[BigBoomBoom]

Realbench is better, it uses more variety of realistic test.

 

[Vellinious]

For light load CPU testing: Realbench for an hour
For heavy load CPU testing: OCCT for an hour
For GPU stability testing: That's tough...there really is no end all beat all for GPU stability testing. I use FS Ultra, graphics test 1 and 2 and loop them for an hour.....if it passes that, it's usually pretty good. Usually.....