Whats your testing methodology?

Soldato
Joined
18 May 2010
Posts
23,646
Location
London
Over the years I've attempted at OC my 3570k with varying degrees of success. I've generally managed to get it stable for 8-10 months until it starts giving my problems, either games crash to desktop or PC start to hard reset in games.

I've mostly used Prime 95 small fft to overclock. Originally my CPU was at 4.5Ghz.... was stable for about 10 months in Win 8.1 prior to upgrade to Win 10 and then started crashing.

Again, settled on 4.4Ghz was stable for about 8 months and then recently started to hard reset in Dues Ex.

Ideally I want a set up fast enough so I don't need to OC but before I pull the trigger on a CPU upgrade, I'll have another go at overclocking.

What's the quickest - yet thorough way to test stability on CPU during OC?

At the moment 4.4Ghz seems to require about 1.31v
 
Asus realbench for 1 hour and then BF1 (the most demanding game I play) together with a few chrome tabs (youtube, twitch, etc) on the 2nd screen for a good 3-4 hours. No crash? Good.
 
Asus realbench for 1 hour and then BF1 (the most demanding game I play) together with a few chrome tabs (youtube, twitch, etc) on the 2nd screen for a good 3-4 hours. No crash? Good.

Cool. Will give that a shot when I get home.

I imagine you set it to run the test 'infinite' and just quit it after an hour?
 
Yes, it's very demanding but in a realistic way.

I run it with half my ram cause if I use full RAM it won't start.
 
Well..... mostly stable. 2hrs of stress testing in Real Bench and 2hrs of bench marking using Real Bench and it passes, but freezes and crashing in the The Division. So I'm still upping the vcore. Currently at 1.31v in bios and it goes to 1.33v I think with LLC under load. Package temps about 77c. So that's all good as far as I understand.

But it makes me wonder, do you guys ever get to the point where the CPU is 100% stable? Or am I not understanding what OC'ing is about? Do you guys get to a point your happy with a CPU that can game for hours with the odd crash or are you too looking for rock solid stability?

Saying that my old QX9650 went 5 years at 3.7Ghz and 1.375v. I only removed the OC as the machine went from a gaming PC to a workstation for web browsing. That never crashed at all.

---

As far as I understand it from Googling my CPU can safely take around 80c in temps and from 1.35v-1.4v in voltage. So I'v still got a little left in the tank.
 
Testing methodology?
Intel Burn Test - Max load, 20 passes.
P95 mixed - 8h
3D Mark time spy and fire strike "stress tests" - 1 each
OCCT on CPU - 1h
FurMark - 2h
If all of those are "stable" or at least seem to be, I begin a real life testing: hell lot of gaming :)
 
Testing methodology?
Intel Burn Test - Max load, 20 passes.
P95 mixed - 8h
3D Mark time spy and fire strike "stress tests" - 1 each
OCCT on CPU - 1h
FurMark - 2h
If all of those are "stable" or at least seem to be, I begin a real life testing: hell lot of gaming :)

Wow! Is it really necessary to test for so many hours? How often do you find your CPU can get through all that and then still have issues in game?
 
Wow! Is it really necessary to test for so many hours? How often do you find your CPU can get through all that and then still have issues in game?
Hah! I found my CPU at 5.1GHz to be fine for few days on stress tests, games etc, and instantly BSOD on game launch on the other day. Little CPU usage spikes can cause such. So I have a reason to make so much stress tests. But then again, I buy a CPU once every 3-4 years, so realistically I spend maybe 72h in total on each new CPU to find a good, stable OC.
At the moment I just run it at 4.6GHz, and been rock stable, not a single BSOD on 1.175V 4.6GHz. Ever.
 
Probably 2 hours in each stress test is ok. Don't see the point in 8hrs of prime 95 if experience shows that you can still be game unstable.

Such a shame there isn't a benchmark out there that actually tests your instability. It seems when you test stability what your really saying is I'm 2hrs Real Bench stable, not necessarily 'stable'.
 
See this is what frustrates the hell out of my when it comes to OC. Recently put back my OC at 4.5Ghz and used Real Bench to test it. It can run Real Bench stress and benchmark for two hours each no problem. It did 4hrs of gaming on Tuesday straight and have been gaming on it no issues in BF1 and the Division. Yesterday however the PC would just blue screen whilst loading the Division. Kept bumping vcore but it just kept blue screening.

How can it possibly be stable one day and not the next! :mad:

Gone back down to 4.4Ghz at the moment.

It seems any stress test you run all your saying is 'I am stable for x amount of time in this test', not necessarily stable in anything else.

Do I really need to run P95 for 8 hrs to get a good indication of a stable OC?
 
I used IBT before but because it puts out so much heat I use Handbrake and games, I don't care if it crashes as my work isn't important so I'll just tune it to that and when I find it is stable up the Vcore another .05v to be safe. Use HWinfo64 and you can see WHEA errors too which is great.

See this is what frustrates the hell out of my when it comes to OC. Recently put back my OC at 4.5Ghz and used Real Bench to test it. It can run Real Bench stress and benchmark for two hours each no problem. It did 4hrs of gaming on Tuesday straight and have been gaming on it no issues in BF1 and the Division. Yesterday however the PC would just blue screen whilst loading the Division. Kept bumping vcore but it just kept blue screening.

How can it possibly be stable one day and not the next! :mad:

Gone back down to 4.4Ghz at the moment.

It seems any stress test you run all your saying is 'I am stable for x amount of time in this test', not necessarily stable in anything else.

Do I really need to run P95 for 8 hrs to get a good indication of a stable OC?

Is it possible that the PSU isn't up to the task? Could also be RAM issues.
 
It seems any stress test you run all your saying is 'I am stable for x amount of time in this test', not necessarily stable in anything else.

It's worse than that. The stress test is just saying I am stable for x amount of time in this test that one time.

There is a possibility that the Division, being an ubisoft title, is just garbo and crashes for the lols. Could also be that your windows install has crapped itself due to updates or just windows things.

I do remember when I had a 3570K, I ran OCCT in AVX mode for about 18 hours at 4.5Ghz and was fine for a couple months then started having crashes in BF4. Ran OCCT and it was no longer stable. I needed to sort out my watercooling loop at the time so I decided to repaste the CPU, reset the bios and fresh install windows and overclock again as if it was a totally new chip. It was OCCT and BF4 stable again at the same voltage and was totally happy for the rest of the time I had it.

Overclocking is always weird. With my 5820K its super weird.
 
Is it possible that the PSU isn't up to the task?
Yup. Before, I used CX750M with my i7 6700k. CPU would crash, always, on 4.8GHz, not to mention !@#$ ton of coil whine coming from PSU, GPU (1070 back then) and CPU.
Then bought SuperFlower Leadex 850W Platinum. 5.1GHz mostly stable (80%), and no coil whine from anything.
 
I didn't like OCCT. To complex, too many options and variables. I'm a noob. Just want something simple and reliable. :(

I'm gonna go with 8hrs of Prime 95 over weekend and see what happens.

This is pushing me closer and closer towards an upgrade.
 
I used IBT before but because it puts out so much heat I use Handbrake and games, I don't care if it crashes as my work isn't important so I'll just tune it to that and when I find it is stable up the Vcore another .05v to be safe. Use HWinfo64 and you can see WHEA errors too which is great.



Is it possible that the PSU isn't up to the task? Could also be RAM issues.

My PSU is from 2008 but it's a 850W Thermaltake something or other. I've got 16Gb of Samsung Green, which I run at stock.

---

Edit, PSU is a Thermaltake Toughpower 850W from 2008.
 
Last edited:
OCCT is very simple. Linpack tab, tick AVX compatible linpack if you want the most intense stress (you PC will be completely unusable). By default the test type is infinite and 90% ram which are what you want.

PSU could be deteriorating with age. Not power output but voltage regulation or jitter.
 
OCCT is very simple. Linpack tab, tick AVX compatible linpack if you want the most intense stress (you PC will be completely unusable). By default the test type is infinite and 90% ram which are what you want.

PSU could be deteriorating with age. Not power output but voltage regulation or jitter.

How long should I run it for, to give me some confidence in stability?
 
As long as possible to be as stable as possible. Rubbish answer but otherwise it would be disingenuous. 18 hours was easy for me as it was overnight plus my working day
 
Back
Top Bottom