How Can You Tell If Your Overclock is Unstable?

[dn7309]

I overclocked my 4770k, ran Prime95 and Aida64 for two days straight, each is 18 hours. Display no errors.

Then today I used Google chrome for 2 hours, and got a windows 8 error message saying that windows has encounter an error and needs to be restart.

 

[Racan]

You can't.

 

[Insert_Nickname]

I haven't had problems with anything that has survived half an hour of Aida64 stress-test and Prime95.

It isn't necessarily your OC that's the problem, it could be a software error, or worse you could have damaged your hardware with excessive stress testing. 36 hours of full-load can damage your hardware.

 

[dn7309]

I haven't had problems with anything that has survived half an hour of Aida64 stress-test and Prime95.

It isn't necessarily your OC that's the problem, it could be a software error, or worse you could have damaged your hardware with excessive stress testing. 36 hours of full-load can damage your hardware.

Many people say you need to stress up to 24 hrs for maximum stability. Their rationale is if it can handle prime95 for that long it can handle anything :hmm: Not sure if thats overkill.

 

[Insert_Nickname]

Many people say you need to stress up to 24 hrs for maximum stability. Their rationale is if it can handle prime95 for that long it can handle anything :hmm: Not sure if thats overkill.

To quote Anandtech's own Ian Cutress:

Overclocking also introduces this strange notion of ‘stability’ – how stable is your overclock? The word ‘stable’ means different things to different people, but the basic assumption is that the system should be stable for everything you do. Intel and AMD ship their CPUs at a voltage and frequency which keeps them stable no matter the situation. Some users attempt to match that stability by stress testing their system, whereas others are satisfied for a gaming stability with no need for transcoding video stability. Testing the stability of a system typically requires some form of stress test, and again users will select a test that either emulates real world (video transcoding, PCMark8, 3DMark) or attempt to find any small weakness (Prime95, XTU). The downside of this latter testing philosophy is that a bad stress test has the potential to break a system. Personally, I shudder when a user suggests a system is not stable unless it passes ‘72hr Large FFT Prime95’, because I have seen users irreparably damage their CPUs with it.

My stress tests here at AnandTech typically consist of a run of the benchmark PovRay (3 minutes, probes CPU and memory) and a test using OCCT (5 minutes, probes mainly CPU). If there is weakness in the memory controller, PovRay tends to find it, whereas if the CPU has not enough voltage for video transcoding, OCCT will throw up an error. There are outlier circumstances where these tests are not enough for 100% stability, but when my systems are stable with these tests, they tend to devour any gaming or non-AVX transcoding for breakfast.

Couldn't have put it better myself...

 

[VirtualLarry]

36 hours of full-load can damage your hardware.

Personally, I shudder when a user suggests a system is not stable unless it passes ‘72hr Large FFT Prime95’, because I have seen users irreparably damage their CPUs with it.

That's complete horse(patooie).

As long as you don't push your volts and temps too high (beyond "safe" levels), you won't damage anything.

Us DC'ers run our CPUs months at a time, sometimes years with loads equivalent to Prime95 loads, and it doesn't harm our CPUs one bit.

 

[Ken g6]

What Larry said, without the profanity. Some even run Prime95 itself, for its original purpose, finding Mersenne primes.

Have you tried [thread=2312638]Wand3r3r's WHEA overclock utility[/thread]?

 

[Insert_Nickname]

That's complete horse(patooie).

As long as you don't push your volts and temps too high (beyond "safe" levels), you won't damage anything.

Us DC'ers run our CPUs months at a time, sometimes years with loads equivalent to Prime95 loads, and it doesn't harm our CPUs one bit.

Notice the word "can". If you don't have adequate cooling of the VRMs f.x.

 

[coffeejunkee]

You can't.

Basically this.

But to elaborate a bit: you use 'real' software to test, not the fairly useless stresstesters which don't really test your oc. What works for me is running lots of stuff at the same time: open a bunch of tabs (preferably with flash stuff), utorrent, play video, play music and do some handbrake encoding during all that. And keep checking for whea errors.

 

[ShintaiDK]

You can't.

^^ This.

Prime95 and Aid64a also only uses a fraction of the instructions in a CPU.

 

[Kenmitch]

You didn't get into much overclocking details.

I know in the past I've had similar issues using a offset voltage. Stable during stress testing with random bsods doing stupid stuff like web surfing,etc. I leave power saving options enabled for the most part. It's the light load hitting of the CPU that causes it. Enough load to ramp the clock but not enough to ramp the vcore. Most of time a little more vcore fixed the issue.

 

[wilds]

I was having a strange freeze up bug at 4.6 GHz when my vcore was too low. On synthetic benchmarks such as LinX or Prime95, nothing froze or blue screened. In certain games and 3dmark the system would lock up.

Going back down to 4.5 GHz or increasing vcore fixed this. Heavy games such as GTA IV, Metro, and Watch_Dogs can be very useful for stability testing.

 

[Kougar]

Different workloads stress different things. As Ian points out Prime95 isn't a blanket stress test, just because a chip is stable under Prime doesn't mean it won't crash doing something else.

I have a degraded 4770K that passes hours of AVX Prime95, it can even pass 4 runs of Memtest86+. But at stock voltage and stock frequency, it will crash every two-ish days running Folding@home throwing all manner of BSoD types that all point back to memory subsystem / cache errors.

Maybe if I ran Prime95 or Memtest for 2 days they would finally detect an error too, but it just goes to show that testing for one full day is still not a guarantee of absolute system stability.

 

[Sattern]

There aren't any ways to check if your overclock is unstable.

In most cases you just have to be sure your hardware can handle it, otherwise if you use the wrong settings you will find out when your computer crashes.

 

[Executor_]

No one else uses IntelBurnTest? Once I read that it was similar to the tests Intel uses to verify stability, I was convinced. I never run it for very long since the temps it hits are ridiculous, well above the maximums I ever encounter otherwise.

If you're really a glutton for punishment, run it at the same time as FurMark. If your system survives 5-10 mins of that without melting, it's probably stable.

 

[ShintaiDK]

No one else uses IntelBurnTest? Once I read that it was similar to the tests Intel uses to verify stability, I was convinced. I never run it for very long since the temps it hits are ridiculous, well above the maximums I ever encounter otherwise.

If you're really a glutton for punishment, run it at the same time as FurMark. If your system survives 5-10 mins of that without melting, it's probably stable.

Intel Burn test got nothing to do with Intel. And its simply a frontend for Linpack made by AgentGOD.

 

[jkauff]

The only software I use on a regular basis that stresses my system is Handbrake (I don't game). If I can get through a Blu-ray encode with no Windows errors, I consider that OC stable.

 

[GLeeM]

or worse you could have damaged your hardware with excessive stress testing. 36 hours of full-load can damage your hardware.

LOL, only 36 hours?

It's the light load hitting of the CPU that causes it. Enough load to ramp the clock but not enough to ramp the vcore.

Perhaps this is the reason I have not had a BSOD on my i7 920! It does not run "light loads". It has been running hard full load stress testing for 44,520 hours straight (except for update restarts and the rare power outage). Temps have been 72-85*C for all those hours.

 

[WhoBeDaPlaya]

I use IBT to get a very quick gauge of stability (very useful during initial rough tuning).
After I hone in on my settings, my setup has to pass 100x IBT passes, ~half a day of Prime95 and a couple of passes of x264 stability test v2.

That last one is great - I've had settings that were completely IBT and Prime95 (AVX2) stable eventually crash on Handbrake jobs. But if it passes x264, no worries.

http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics

 

[Madpacket]

As Kenmitch stated the instability is caused by light loads with inadequate vcore. This is why many overclocking guides recommend disabling c states, power saving etc.

 

[toyota]

running crazy loads for hours seems stupid to me. every crash I have ever had after my oc was stable under a 20-30 minutes of stress testing has happened at idle. so what good would non stop load testing do? in every case I have had to go back and set my voltage to auto setting and I will never crash again. I just seem to have the strangest luck with any adaptive or manual core voltage.

 

[Ed1]

Right , with a heavy load the voltage stays pretty stable but with normal use you have the transition from idle to mid load were it can be a issue .

One thing, if your chip is real good or your not OC real high set the LLC not to high that you have to use big - value, this only compounds the transition to load were voltage doesn't raise fast enough at light loads .
bumping Vcore a 10mv helps if your not to low on idle voltage . would be nice if MB bios supported a min voltage to start vid table, kind of like what you can do with fan profile .
Adjust min, ramp amount and a max voltage per your multiplier .

 

[dn7309]

running crazy loads for hours seems stupid to me. every crash I have ever had after my oc was stable under a 20-30 minutes of stress testing has happened at idle. so what good would non stop load testing do? in every case I have had to go back and set my voltage to auto setting and I will never crash again. I just seem to have the strangest luck with any adaptive or manual core voltage.

So what is the difference between Adaptive and Auto setting then? doesn't both setting auto adjust voltage base on load? So why should I set mine to auto instead of adaptive or the other way around.

 

[WhoBeDaPlaya]

Right , with a heavy load the voltage stays pretty stable but with normal use you have the transition from idle to mid load were it can be a issue.

That's why I dislike adaptive Vcore overclocks. Call me old-fashioned, but I tend to stick with fixed Vcore OCs.
Idle power usage is no biggie for me, since I'm usually running either encode jobs or DC.

 

[schmuckley]

O Bless Mr. Cutress.
My test for stability=
2 runs of Hyper-Pi 32m
+ maybe 3 Intel Burn Test./.
or..Crysis:Warhead up til "the corner"
Usually if it's unstable there will be glitches at "the pit" flashback area.
all in all,less than 1 hour