There is not "one best" pick one, or use them all, they all do the same basic thing, exercise the cpu the a limited but stressfull way to raise operating temps while checking the validity of some calculations. I like Orthos just because it was new when I was looking for a tool and the blend test (might be present in the others not sure) allows some testing of the memory system as well. An of course you need Memtest86+ to validate the memory setup before you even start with testing the cpu/system.
IMO the best stability test is to use/run what you always use/run for applications and games for 2 weeks and if no issues, you are stable. What else could you want ? .
For testing cpu stabiltiy while finding an OC setting, eh, take you pick. Be sure to run 3dmark as well to ensure your cpu/MCH-northbidge/video card will work at the new settings. The score is not as important as is completing the test without a problem.
I use for my personal machines, in the order I run them. after making the easy settings and start getting to the "edge" of the OC:
TAT to check the cpu thermals.
memtest86+ memory test
ORTHOS but runs with both the small FFT and a Blend run.
3dmark forget which but not the recent one, need to fix that now that I have a dx10 card.
I install speedfan and rivatuner for the temp monitoring capabilities and enable logging and write the files to my desktop.
After I am happy with where I am, I then grab drinks/snacks etc and sit down and play DAoC until I fall out of the chair from sleep deprivation. If no issues, done.
edit: of course then after two weeks I start wondering if I can squeeze just a little more out and screw everything up. Take notes and keep a log.
anyone know how to set the priority in version Prime95 version 25.6?...I know in previous versions there was a password in the ReadMe file but this time I can't find it anywhere
I noticed Kyle (I believe) using rthdribl for testing graphics so dug it out.
It seemed to give the same load temps as the hardiest graphics tests so today I tried clocking with it (after fitting MX2 paste to my gfx card).
As long as you are watching you will see if there are any graphics glitches.
I found the resulting core/shader clocks to be highly stable but it did let me set my ram 20MHz higher than is stable while gaming.
A pretty good test for gfx clocks is to use Crysis benchmark with a very high config, 1600x1200 res or higher and run at least 6 time demos, the more the better but I havent personally found any benefit to running more than 10 but everyones case air flow and surrounding air volume etc etc are different so find what suits best.
To set the number of loops, edit
\Crysis\Game\Config\benchmark_gpu.cfg
and change the line
demo_num_runs =
If you feel like adding some AI calculations into the mix, use the following
demo_ai = 1
I lose about 3fps (6% drop approx) with AI on.
Something handy to know when stability testing with Crysis:
Crysis sometimes crashes with clocks that are too high.
When this happens, it seems you have no control of the PC but ALT + F4 will kill Crysis and if you leave it for a minute or so, you will eventually get control back.
I sometimes have to press the windows key after, before it lets me click on anything on my desktop.
This saves having to do a hard shutdown or quick shutdown.