Quote Originally Posted by virtualrain View Post
Thanks for the overview!

It may be that your tool is better at uncovering CPU instability quicker, but it's nearly impossible to prove and I still am skeptical of any such claims. As you say, CPU errors are random.

At any rate, I will add this to my overclocking tool chest again. I used it previously on my current OC and found that I was OCCT stable for an hour when PRIME95 would fail in about 15 seconds. I also found that none of the monitoring worked in x64 due to unsigned drivers.

I really agree with using multiple tools... PRIME95, OCCT, 3dMark, Memtest, Games, etc.

Side story... I remember back in the 80's and early 90's that the way you could test if your PC was "IBM Compatible" was to run Microsoft Flight Simulator (most current version of the day). I guess MS Flight Sim was one of the first stability tests!
I also used to trust prime95 small FFT for CPU only testing. Now, that OCCT in CPU mode loads the CPU as much as Prime95, I use both of them. But who wins, I won't be able to answer it, maybe none would be able to do.

However, Small FFT is no longer enough to say an overclock is stable or not. Small FFT/CPOU test will let you adjust the vcore, vPLL and vFSB.

The RAM mode of OCCT gives overclock stability testing a big boost. People often wonder why Prime95 small FFT is stable while games crash in few minutes and F@H bugs. RAM and NB stability account a big part in an overclock. Many people know it well and use Prime95 Blend test for that, or even memtest.
OCCT in RAM mode really outperforms Prime95 Blend/large FFT and memtest (memtest is still the only program testing the whole RAM integrity because it runs in a dos mode, but it won't heat the RAM or NB as OCCT or Prime95 Blend). Prime95 Blend can be stable for 10h when OCCT in RAM mode will fail in few minutes, usually in less than 20mn, and rarely above 2h. Anandtech now also use OCCT in their tests rather than Prime95, must be a reason there, right?

OCCT in mix mode will alternate CPU and RAM tests, so it will be much more efficient in detecting instabilities than Prime95.

If you read some of the other threads in Intel section, you'll see that many people noted it, Blend mode in Prime95 eating banana comapred to OCCT in RAM mode

And yes, I do agree that people looking for ultimate stability should use many benchmarking programs not only one