Prime95: different test = different CPU temp?

P

Purostaff

Weaksauce
Joined
Mar 9, 2008
Messages
73
I've noticed that the Small FFT test was 10 degree Celcius higher than the other two tests (In-place large FFT and Blend). I've tested with 1.5 vCore (68C vs 58C) and 1.388 vCore (62C vs 52C) to confirm. Now I've read some write ups on prime95, and they reported very little to no differences in temp for the three tests mentioned.

I believe most people accept the 'blend' test as gold standard for stability testing, but I'm just OC'ing my CPU right now and want to leave out RAM as much as possible until I get new sticks to play with.

Anyways, is the 10 C temp difference normal for C2Q?
 
Did you happen to notice that the blend and large FFT's don't stress the CPU fully?

Small FFT's stress mainly the CPU, Large FFT's mainly the Ram, and the blend is a little of everything.

This is why you have different temperatures on your core...

EDIT: also remember to run separate instance of Prime 95 for each core...meaning four instances of Prime 95 running at the same time.
 
MaxTheLimit said:
Did you happen to notice that the blend and large FFT's don't stress the CPU fully?

Small FFT's stress mainly the CPU, Large FFT's mainly the Ram, and the blend is a little of everything.

This is why you have different temperatures on your core...

EDIT: also remember to run separate instance of Prime 95 for each core...meaning four instances of Prime 95 running at the same time.
Click to expand...

Yup, I do aware of what each setting does. I just want to clear up some temperature discrepancies with an article I read.

So, which temp readings should I (or all oc'er out there) go by when OC'ing? I'm sure I would not be stressing the CPU anywhere near the 'small FTT test' with normal/gaming usage. Surely, 'blend test' would give me lower temp readings, but would that be safe even if temps for 'small FTT test' are higher than recommended maximum?
 
I thought this was all well understood... It says what each type stresses in the dialog box that starts them. ???

"Small FFTs" for CPU overclock stability testing and "blend" testing to make sure the FSB, RAM, etc are all happy as well.

On current versions of Prime95 do you still need to make a shortcut to it with the "-a1", "-a2", and "-a3" switches (and also launch it by just double clicking on the .exe) in order to get all 4 instances up for quad testing? Now that I have a quad (in my home server), that is starting to get annoying. For my e6600 game machine it never bothered me, though (only needed an "-a1" plus the .exe).

Also, you need to set the affinity (for CPU 0, 1, 2, and 3 in my case), not just launch 4 instances. If "let the program decide" is checked, your out-of-luck. I am using a copy I got about a year ago, so I suppose a few things could have changed for newer versions.
 
Just download the newest version( 25.6) and you can run all 4 cores(or 2) in 1 instance.
 
Purostaff said:
Yup, I do aware of what each setting does. I just want to clear up some temperature discrepancies with an article I read.

So, which temp readings should I (or all oc'er out there) go by when OC'ing? I'm sure I would not be stressing the CPU anywhere near the 'small FTT test' with normal/gaming usage. Surely, 'blend test' would give me lower temp readings, but would that be safe even if temps for 'small FTT test' are higher than recommended maximum?
Click to expand...

certain streamed programs can fully (or at least almost) utilize your 4 cores. More and more programs will take advantage of them so for the sake of seeing what a streamed program will have on your temperatures you still want to run a Small FFT's test to see how the temperatures look when your CPU is fully stressed.

The newest beta of P95 does stress multi cores but there are reports of it giving false errors that aren't due to overclocking problems. compared to Orthos, or even OCCT sometimes it gives errors where they don't...
 
It's unlikely that any normal program would stress the CPU to the level that the Small FFTs test does, because few real-world programs would do minimal memory access and no hard drive access, etc. and just keep running calculations as fast as the CPU can go. However, if your CPU is stable under a stress test like that, then you can be pretty confident that it won't have problems day-to-day. If it can't cope with the toughest tests, then maybe it will be okay with day-to-day usage, but maybe it won't - something is obviously running very close to being out of tolerance. Best not to risk it if you want a stable system.
 
Exactly. It's like a worse case scenario. The point of stability testing is testing under the worst case scenario for each piece of hardware involved.
 
You must log in or register to reply here.
Back
Top