prime95 99% of the time throws a rounding error if its going to do anything but bluescreen for me. When i start a stress test its hands off after i push that start button. most ill do is run music or talk to someone on aim.
Stability is always a relative thing for overclocking. The only true stable is Stock. and even at stock you have random bit flips in memory which becomes proportionally more probable with more memory. The whole point of 12/24 hour stress testing is your running somthing thats HARDER and more sensitive to memory errors then you ever will.
I have in the past seen certain overclocks that were for instance stable with super pi and prime95, but unstable in linx
and likewise stable in linx and unstable in prime95, I have also seen stable in every cpu stress test i could throw it at, but run 3dmark(at those days, 2003 was the latest) and forget about it. This is why i have developed this stress test method i use as my final test:
I know its old, but i run RTHDRIBL with the skull, and maximum effects to drive the frame rate down. I have to extend the window across two monitors to bring it under 60 FPS (if your video card can run it faster then your monitor, its not stressed enough)
Then ill run linx using a test large enough to use up the rest of my avail ram. if that combo is stable for 12~24 hours i consider my computer stable enough for what i do with it(mostly gaming)
RTHDRIBL is also really neat for detecting graphics card stability issues, you see bright green/blue flashes whenever there are errors in the processing
I dont know, maybe im a nutcase, but thats what i do when i build a new system or change a system-changing part such as a psu or mobo. If its somthing simple as new ram or new cpu ill do the usual cpu/ram tests and then if i notice anything that could be stability issues while using the computer ill do the full oneEdited by ryuji - 2/1/12 at 9:23am