Prime95 Stability

[LordShadow]

Hi,

Im currently overclocking a new Conroe rig.
I'm using prime and Superpi running at the same time as a rough guide to stability, or Im using dual prime.
What are the main differences between the Large FFTs and Small FFTs tests?
Is it that the Small FFTs stresses mainly CPU?
Im stable on small FFTs but not on large FFTs, does this suggest the instability may lie with the chipset rather than the CPU?

Thanks, LS.

 

[joeyjojo]

The way I understand it, small fft's are between 8k and the size of the L1 cache, large are bigger than the L1 cache but smaller than L2. This is so, on small, they fit in the L1 cache on the cpu, with large they stress the L2.

What I do is run SP2004 on custom. Got it using between 8k and 1024k. It's like running small and large.

Both stress mostly the cpu. For some reason, one gets the cpu hotter than the other. I'm not sure that one will stress the chipset more than the other, never had a system with a northbridge.

 

[manveruppd]

It's large FFTs that generate more heat, presumably because the on-core cache rather than just the processor itself get hot. Don't know which is the best one to run to test stability though - I usually run Blend since if I'm overclocking the RAM will be OC'ed too and also needs testing.

I'm not sure that one will stress the chipset more than the other, never had a system with a northbridge.

Eh?

 

[ted34]

The problem with blend is that you dont know what failed - the cpu or the ram so you dont know which to add volts to. Best is memtest for memory and sp2004 orthos set to priority 9 on small ffts for dual cores