How long do you need to run temperature tests?

[Moromann]

Hi. I just overclocked my Intel i5 2500k from 3.3GHz to 4.5, and I have run Prime95 for about an hour. The temperatures seem very to be very stable at 60-65 degrees, but I just read somewhere that you should run the test 12-24 hours to get a real result. Is that really necessary, or do you only need to run the test that long when you're overclocking more heavily?

Here is a printscreen of the current results: http://oi50.tinypic.com/28kh8ir.jpg

 

[Kreij]

Many people run the test for 24 hours because it gives you a better idea of the long term stability of the system under heavy load.
While one hour let's you know that the OC is stable, it doesn't give you an indication of how the system will perform under hours of stress ... like a long gaming session, especially with a CPU intensive game.

I have not OC'd anything in quite some time as I see no need.
When I did overclock I usually ran a stress test for about 5 minutes and then went back to gaming.

 

[Aquinus]

When I did overclock I usually ran a stress test for about 5 minutes and then went back to gaming.

That's basically what I do. I wait to see where temperatures settle at and if its reasonable and Prime95 doesn't crash it in 10 minutes, I'm right back to whatever I was doing.

 

[Kreij]

Unless you are going to be doing heavy benchmarking on the rig and want to ensure long term stability in a short period of time, I think an hour is sufficient for assessing system performance.

You are unlikely to stress the CPU, in the manner that Prime does as it runs, in your everyday use of the computer.

So ... quit worrying and get back to gaming.

Seriously ... you need to feel comfortable with the OC, so run the test to the extent you are happy with the results.

 

[BarbaricSoul]

That voltage seems a bit high for 4.5ghz. I use less than that for 4.7ghz. Try lowering the voltage some more.

 

[Norton]

You can also spend some time grid computing with World Community Grid (WCG) @ 100% load if you want to test your stability, temps, etc.

If you're not stable in WCG your rig will produce errors in the results submitted, BSOD's, or both. Great option for longer term stability testing

and... rather than just generating heat and using power you can contribute to a cause that can ultimately help us all.

http://www.techpowerup.com/forums/showpost.php?p=1364692&postcount=1

[/end sales pitch]

 

[BarbaricSoul]

You can also spend some time grid computing with World Community Grid (WCG) @ 100% load if you want to test your stability, temps, etc.

If you're not stable in WCG your rig will produce errors in the results submitted, BSOD's, or both. Great option for longer term stability testing

and... rather than just generating heat and using power you can contribute to a cause that can ultimately help us all.

http://www.techpowerup.com/forums/showpost.php?p=1364692&postcount=1

[/end sales pitch]

What he said.

 

[bmaverick]

About 3 days. This ensures that the temps are stable and the thermal paste has settled for the best performance.

 

[Yo_Wattup]

That's basically what I do. I wait to see where temperatures settle at and if its reasonable and Prime95 doesn't crash it in 10 minutes, I'm right back to whatever I was doing.

+1

You gotta keep in mind that stress tests are artificial. You can stress your cpu for days and have the computer crash 5 minutes into a gaming session.

My temps usually get no higher after about an hour of prime 95

 

[Drone]

From a physics point of view heat dissipation requires less than 15 minutes.

As for stresstest ... I dunno .. I've seen some cases when system could prefectly run for 14 days but collapsed on the 15th day. Average test requires I think 12 hours or something. There's nothing special in 24h test because if it didn't fail for 24 hours it still means nothing.

 

[Jetster]

An hour or two if fine

 

[vrdublu]

An hour with LinX 0.64 will tell you everything you need to know, best software for stress testing by far.