Originally Posted by gordan
So what you propose is no repeatable or measurable stability test at all. Taking that view to it's obvious conclusion, you might as well just test whether the card will idle without crashing the machine. I'm not saying that something that is stable for 24 hours in OCCT and CudaMiner is necessarily stable under all conditions. There are parts of the card that those tests don't stress at all. But since those are the only tests I'm aware of that return any conclusive stability information, it is at the very least a good starting point.
Hardware is designed to last for years. If software is capable of generating enough load to damage the hardware in hours or days, then the hardware was faulty to begin with.
Well, since I use my computers to game, run benchmarks, edit photos and videos, and do small CAD renderings, I use those same games, benching, editing, and rendering programs to test for MY systems stability, pretty common sense really, because those are what I need them to be stable with.