20 runs of intelburn on high adequate to test for stability?

frenchy0312

frenchy0312

frenchy0312

Associate
Joined
21 May 2010
Posts
115
Hi, I've just been fiddling with my AMD 955 C2 rev. I have it currently set at 1.325 volts with the multiplier at 18x giving 3.6Ghz. Now I am using intelburn to test it for stability and its done 17 of 20 runs so far on the high setting and hasn't crashed yet so I'm wondering if you think, once it has completed 20, you could consider it stable?
 
A pc is never truly stable, there are many interpretations of stable. 20 passes of linx could fail prime in an instant and vice versa. Passing prime for 8 hours might crumble after 10 mins of gaming.

In short, all the stability tests are indication only and 20 passes of intel burn is therefore an indication that your overclock is stable. Like Martin, I would also leave it running prime overnight on highest priority to give that extra piece of mind.
 
w3bbo said:
A pc is never truly stable
Click to expand...

Poppycock I'm afraid.

A computer is a mathematical machine. If it EVER miscalculates, it is not stable.

Even under an overclock, the PC should run and run and run any stability test for as long as you want to leave it and never error. If it does error, even after 3 days, it's not stable.
 
Martini1991 said:
I don't consider any overclock that hasn't passed 8 hour prime blend stable.
Click to expand...


what a load of rubbish. you could do it for 3 days and it still not be stable and have lock ups or BSOD. you can run prime95 for days then run occt or intelburn and get erros with there programs


run the test for as long as you want. then go back to your normal day to day stuff it it locks up or crashes then its not stable back to testing. dont spend your life testing or you will never get to use your computer.
 
koooowweeee said:
what a load of rubbish. you could do it for 3 days and it still not be stable and have lock ups or BSOD. you can run prime95 for days then run occt or intelburn and get erros with there programs


run the test for as long as you want. then go back to your normal day to day stuff it it locks up or crashes then its not stable back to testing. dont spend your life testing or you will never get to use your computer.
Click to expand...

That's why you shouldn't rely on one stability program. I use occt, linpack, hyperpi, prime95 and memtest. Run them all at the same time for 24 hours. If it can take that.... then you can be PRETTY sure that any crashes or BSOD are the fault of software, and not instability.
 
My Q9550 passes 20 run's of Linx at 4.2Ghz but fails prime after 15 hours. 4.13Ghz is rock solid and passes 24 hours of prime. Depends how far you want to go to call it stable i suppose.
 
If you run prime long enough almost anything will fail... you can prime something for a week straight and it pass and then run it again and have it fail after 5-6 hours.

I usually use 20 passes IBT to get an idea of where is stable when I think I've found it I'll give it another 20-40 passes on IBT and 2-3 hours on OCCT - and then probably ease it off 1 notch if it passes that if I'm really worried about stability.
 
Memphis said:
Poppycock I'm afraid.

A computer is a mathematical machine. If it EVER miscalculates, it is not stable.

Even under an overclock, the PC should run and run and run any stability test for as long as you want to leave it and never error. If it does error, even after 3 days, it's not stable.
Click to expand...

To add to this...

If you lock a cat in a box and never open the lid, how do you know if it's dead or alive?

If you overclock your PC and it never crashes, how do you know that the overclock is stable or unstable? Only when it crashes do you know that it's unstable... :P

But aye, 8 hours Prime95 is my test of choice.
 
You should use intel burn test to quickly test stability over 20 runs. If it passes that then you can try prime/orthos for the many hours it takes.

Otherwise its pointless to try 24 hours of prime and for it to fail after 5-6 hours, when with intel burn test it would have failed in about 3 minutes.

Personally I dont use my PC to do any massive number crunching or folding, so I am happy with 20 runs IBT stability. But I do use memtest to test memory stability, and if I get any BSODS or wierd instability errors then I go back to the drawing board. So far I have had none of those.
 
IBT confuses me.

When testing do you use the extreme setting?

Im a noob overclocker, but I soon realised that IBT, is no where near as good for overclocking as Prime is.

Prime is gentle and lets you tweak, fine details.

IBT, is brutish smashes stuff, doesnt tell you whats broken, and leaves you with the mess to clean up.
 
I've never done any more than a couple of hours of prime and 20 runs of IBT on max, plus some furmark when OCing the GPU. Now I'm busy enjoying the performance of my 4GHz i5 which has never crashed during gaming or anything else. Some people are ridiculously OCD about stability and stress testing. Run a stress test long enough and you will get an error eventually even at stock settings. Why some people seem to want to prove this statistical inevitability and spend more time stress testing than actually using their machines beats me :)
 
Generally these days, i'll 10-20 run IBT/LinX, 1- 2 hr prime blend, then go onto serious testing.

What i class as serious testing is 24-48 hours of folding, if it can do that without crashing the folding client or BSOD, then its stable.
Why i do that is, generally my overclocks where allways prime stable, but where sometimes flaky when it came to folding, so i now use folding as my main stability test, as that is what my computer is used for, well that and gaming.

But as others have said, stress testing is only an indication, you can never trully call any PC 100% stable, we use to many different bits of software these days, the hardware may be stable, but a software fault may cause a crash also.

But I would say about the original question, 20 runs on high/max is the first step towards stability, if it passes that, then its still always a good idea to do a long Prime Blend run also. But it all depends on how much time u want to take really.
 
Small tests basically the Cpu heavily, and uses very little else. So its a good way to find out if the vcore is an issue or not.

Blend checks everything, from vcore, memory, chipset stuff, so is allways best to run this at some point for the overall view.
 
I read somewhere that you need to run Prime95 Blend test for 10 hours as each FFT size is used for 15mins(default) and it would take Prime95 about 10 hours to cover all of the FFT sizes available.
 
Well I have my i7 980X now overcloced to 4.1GHz.

I ran 20 runs of IBT with highest settings and another 20 runs with LinX 0.6.4 highest settings and finally 5 hours of each Prime95 Small FFT and Blend mode.
 
Memphis said:
Poppycock I'm afraid.

A computer is a mathematical machine. If it EVER miscalculates, it is not stable.

Even under an overclock, the PC should run and run and run any stability test for as long as you want to leave it and never error. If it does error, even after 3 days, it's not stable.
Click to expand...

Really? I would never have guessed a pc did math! :rolleyes:

Like I said, a pc is never truly stable. PC's will miscalculate. Even a so called 'stable' pc will still throw up an error every now and again. There are simply too many variances to throw into the overall equation for a pc not to be. Just a small power fluctuation can result in even a pc at stock speed to fail prime/linx/ibt or any other program.

I have had pc's that have folded for months on end, passed prime for 24hours which have thrown up an error after months of trouble free use.

So I will reiterate - A PC is never truly stable.
 
You must log in or register to reply here.
Back
Top Bottom