Just found this program - BurnTest - Basically stability testing. Seems to be far more stressful on the CPU than Prime95 - my temps on this are 15 degrees higher at full load than on Prime95.
I use it regularly, never bothered with Prime95 - after all, 30 minute stability tests with rock-solid answers and higher temps through better stress is far better than 24hrs of 'can I count that as stable yet?'.
the load difference is to do with the percentage of CPU being utilised. for example, the cache hit-rate is a huge factor is the CPU utilisation rate, Intel Burn test is using LinPack to calculate, LinPack is a very efficient algorithm calculating the matrix inversions, so cache hit-rate is very high, therefore CPU spend less time waiting for memory and more time doing stuff. there is, possibly, more stressful algorithm than LinPack, but since this is more stressful than Prime95, and no program can even reach Prime's level of stress because of branching and other stuff, it's safe to say as long as you pass the BurnTest, your can consider your CPU's temperature to be within range.
Actually Folding@home, or more specifically GROMACS, stresses a CPU far more than Prime95 ever will. That's precisely why StressCPU2 was created. It uses hand-coded tight assembly loops to stress the FPU and SSE units of CPUs.
Different types of test will give different temps... Temps aren't directly related to CPU load. I have been using F@H on both overclocked CPU's and GPU's lately to perform stability tests
interesting. thanks. i was encoding when i typed last post, and my temperatures are a good 7 or 8c lower than Prime95's highest value. to me, F@H also never achieved high temperature may be because GPU2 client cutting into the schedule.