# of h for p95 torture test to be "stable"?
 

Anyone? I have my machine clocked @ 8.5x400 and it's largeFFT stable (prime95 x64 v25.6) for 18 h 21 m at which point a rounding error occurred. Guess I'm wondering at what point (18h, 24h, 48h etc.) you call a system "stable"? I used to say 12 h was fine, but I over slept and it ran longer than I anticipated and gave the error.

Here are the specifics on the system:
[CODE][b]D.F.I. LP LT P35-TR2[/b] (BIOS: LP35D317)
[b]Intel X3360 @ 8.5x400=3.40 GHz[/b]
CPU VID=1.11875 SB 1.05V=1.070
NB Core=1.370 SB Core/CPU PLL=1.075
CPU VTT=1.280

[b]Corsair Dominator DDR2-1066[/b] (TWIN2X4096-8500C5DF)
2x 2Gb @ 5-5-5-15 @ 960 MHz (5:6) @ 2.100V[/CODE]

Any new PC I get I usually run it for week. Just so that it can experience at least once the expected weekly temperature and load variations and whatnot.

Recently on Core2 (e.g. C2Q) architecture, I run 4 custom small FFTs test (32k-160k) for ~24h and afterwards I switch one of the instances to blend test for another 6-24h (it is better to use 4 instances of v24.14 for such testing). If the system doesn't fail during this time, I consider it stable.
Getting back to your specific case - your system is not prime stable, I would fiddle some more with voltages, and if that doesn't help lower the CPU frquency e.g. to 3.2GHz :smile: