CPU Thermal Stress Testing

Malovane

Distinguished
Jun 17, 2008
177
0
18,680
I just built a new system based on these specs:

AMD 9850 BE 2.5Ghz CPU
Foxconn A7DA-S 790GX Mainboard (with 750 Southbridge)
4 GB (1x4 chips) OCZ Platinum DDR2-1066 (currently running at DDR2-800)
Sapphire HD 4870 512MB graphics card
Corsair 550W Single Rail PSU

This is all fitted in a Silverstone Kublai KL-02 Case with the following cooling components:

Thermalright IFX-14 on the CPU Fitted with 2 60 CFM 140mm fans
2 Silverstone 120mm 110CFM intake fans + 1 Silverstone 120mm 110 CFM exhaust fans
Thermalright HR03-GT with HR-11 to be installed shortly, so using stock cooler for now.

The operating system is Vista 64 SP1 Home Premium.

My ultimate goal is, of course, to overclock the living hell out of this thing. However, I want to make sure that I can stay within thermal tolerances, no matter what load is on the CPU. I am, sadly, not a wealthy man, so I want my components to stick around a while (though I will move to a Deneb CPU when that line comes out and is vetted).

Currently, my thermal test setup includes one instance of Prime95, one instance of SuperPi, and 4 instances of folding@home. This pushes all 4 cores to 100% fairly nicely. I am wondering though if this is actually stressing the CPU in the correct manner for the most thermally grueling experience.

I've gotten the following stats so far:

Idle: 27C Min, 28.5 Max
Full Load: 32C Min, 34C Max

The temperatures have been confirmed through several programs, which left me scratching my head a little bit. Of course I'm pleased with the temperatures, but they seem awfully low for air cooling, and made me wonder about my test method.

So the question is this: What is the best software combination and/or method to test the maximum thermal output of a CPU? Am I doing a decent job with that, or is there something you would do differently?

I would appreciate any advice you guys have to offer. Thanks. :)
 

Grimmy

Splendid
Feb 20, 2006
4,431
0
22,780
Sounds like you have the older version of prime.

This is off the OC guide that has the update prime to work on quad:

P94v256.zip

Just replace the executable, and when it comes up with 4 worker threads, just chose the FPU stress test, and that should increase the temp a bit.
 

Malovane

Distinguished
Jun 17, 2008
177
0
18,680


Thanks Grimmy, that helped out a lot. After about 5 minutes it seems to have hit it's peak at 38C, which is a pretty big difference. :)
 

Malovane

Distinguished
Jun 17, 2008
177
0
18,680
Thanks again Grimmy. :)

System now appears stable at 3.3Ghz @1.45V and max load temp of 49C after an hour and a half of Prime95-64, with an idle of 32C.