How long is long enough?

jadesaber2

Limp Gawd
Joined
Nov 8, 2005
Messages
298
So I upgraded my system recently. Went from
AMD Opteron 170
ASUS A8N-E
X1950Pro
2GB PC3200

to

Q9300
Gigabyte EP45-DS4P
HD4870
4GB PC2-8500

Got myself OCed to 3405 MHz last night. Ran Orthos on all cores for 8 and a half hours. Temps stayed below 70C in a 25C room with a Xigmatek HDT S1283.

My question is, is 8.5 hours long enough to be sure that my setup is stable? Or should I let it run all day?
 
That's probably a decent degree of stability, but I've seen Orthos/Prime fail at 15 or 16 hours. Whether you continue or not just depends on the amount of certainty you're comfortable with.
 
If you're just gaming, 8 hours is probably fine. If you are running distributed computing projects like folding@home, I would do 24 hours just to be super sure you're stable - wouldn't want to be having calculation errors on a DC project. I usually run about 1-2 hours while I'm finding maximum overclock, and then once I settle in on a 24/7 speed that I like, I leave it run for 24-36 hours and then forget about it. I like OCCT as a test utility since it stresses all 4 cores. The intel utility is great too.

I have no basis for this other than my Q6600 @ 3.2ghz doesn't really break 60c unless it's really hot in the room, but 70c seems hot for a 45nm quad. Although, there's no danger at that temp.
 
try the intel burn test utility, it's a much better stress test

Not really. It's good if you want to test your cooling, but not necessarily for stability. IIRC, the intel burn utility does not do any calculaion checks like prime does, so if there is a calculation error, you'd never know it unless it was severe enough to completely lockup the computer.
 
8.5 hours is fine, unless you are getting paid to run telemetry for n.a.s.a. i usually go an hour or 2 tops, but i am only gaming.
 
Not really. It's good if you want to test your cooling, but not necessarily for stability. IIRC, the intel burn utility does not do any calculaion checks like prime does, so if there is a calculation error, you'd never know it unless it was severe enough to completely lockup the computer.

Actually, IBT does perform calculation checks. It produces some numbers for each run it does, and if the numbers from all the runs don't match up, the system isn't completely stable. It's a fair amount better than Orthos and Prime for stress-testing since it will weed out errors much earlier than anything else. 20 runs of IBT is enough to test pretty much any overclock.
 
Personally.. if it was me, I would not want to hit even 60c for 3.4 Ghz.. and that is on a Q6600.

At 3.76Ghz, it doesn't go over 62c with my setup.... air cooled. That is about the max temp I dare to let it have.

As for being stable.. it sounds like it is plenty stable, especially for gaming.
 
Personally.. if it was me, I would not want to hit even 60c for 3.4 Ghz.. and that is on a Q6600.

At 3.76Ghz, it doesn't go over 62c with my setup.... air cooled. That is about the max temp I dare to let it have.

As for being stable.. it sounds like it is plenty stable, especially for gaming.
try running 20 cycles on IBT... you won't be pleasantly surprised....
 
Actually, IBT does perform calculation checks. It produces some numbers for each run it does, and if the numbers from all the runs don't match up, the system isn't completely stable. It's a fair amount better than Orthos and Prime for stress-testing since it will weed out errors much earlier than anything else. 20 runs of IBT is enough to test pretty much any overclock.

Good to know. I just tried 20 runs on my setup and they all came back good. Does IBT work with AMD processors too?
 
Good to know. I just tried 20 runs on my setup and they all came back good. Does IBT work with AMD processors too?

I think it does, although I can't confirm anything. The only AMD system I have is my laptop and I wouldn't risk running IBT on that.
 
Back
Top