Do you actually stress test your CPU?

[Recon-UK]

Well back with my Phenom II i could pass Prime95 for 12 hours just fine, then if i tried Bad Company 2 it would BSOD between 15-20 minutes.


I usually just subject my CPU's to programs i actually use, like rendering video in Vegas and some intensive games that get all threads working.

 

[CAPSLOCKSTUCK]

i just run my everyday programs and game.

I set my o/c in BIOS and High Performance in windows and leave it there, which is usually between 4.4 - 4.6 ghz so i suppose i am constantly stressing it as the stock speed is only 2.9 ghz.

I'm sure the old Xeon will cope OK

 

[agent_x007]

I run 10-20 min. of AIDA64 + a benchmark, and then use it as it was stable.
If it crashes, I adjust settings and try to replicate error from before.

 

[BiggieShady]

I stress test the overclock in prime, but have no patience for an hour run ... I just wait for temps to stabilize and call it a day if it didn't derp before that
... however I often wake in cold sweat from my nightmare yelling loudly "What if my OC is actually unstable!"

 

[manofthem]

I haven't OC'd much lately so I haven't stressed tested much. Usually, stress tested for a little bit while finding a stable OC, then play some games and see how it holds up.

For the ultimate test, let it crunch BOINC 24/7 and if that runs fine, it's fine.

 

[alucasa]

My workload stresses CPU automatically.

So, no. By using my PC for the work I do, I stress-test my CPU every single day for hours and hours.

It's called rendering.

 

[Kursah]

Interesting topic.

Depends on the build. But I stand behind a reputation of building bullet-proof PC's, and to verify they'll do that I run OCCT's series of tests.

I voted on your sarcastic first option, though it doesn't do a good job of explaining what I do and why I do it...

Some folks take stability seriously some don't (until it's an issue and they're on here pissed their system keeps BSOD'ing and can't figure it out...). I like to find a happy medium of testing and real-world use. If your system is stable for what you do, then it's stable for what you do and that's all.

I'm not saying OCCT is the end-all for stress tests...but for me it's a solid suite of tests that get the job done and have for years.

I don't use Prime or IBT, though sometimes I'll expand and use Aida64, Intel XTU, Asus ROG Realbench, etc. Frankly though, that's more depending on the system and situation. 95% of my situations, OCCT does it all and allows me to vet out instabilities.

If a system is stock, I'll still run it to see how it can cope with heavier than normal loads...

If a system is OC'd, then obviously stability should be tested so that's what it get's used for!

For how long it takes, depends on the response from the first test...if the temps are good and OC is stable on the first CPU test, then move onto GPU, then PSU. If all pass...then we're good...if not then it's time to adjust and retest. Some folks get away with not testing stability much or very long, I've been bitten and I'd rather take the time to do it right by methods I've honed by building and overclocking for almost 2 decades now. It works for me, and that's all that really matters in the end. My systems are stable, my customers are very happy and overly willing to pay me extra to make sure their systems are stable.

Onto timeframe, I start with 1-hour tests. That's long enough to have a consistent heat-soak and time for things to fluxuate in voltage, temps, and stress levels. If it is stable there, I'll run another one overnight that could range from 3-8 hours depending on how the system responded from the initial test. If all that is passed, then it's time to do what the system was built for...usually gaming or rendering, heavy multitasking, etc. The best stress test in the end is using your system for what you intended...but that's not always the most helpful when issues crop up, nor is it ever good timing. So if we can reduce the likelihood of BSOD's and crashes when merely using the PC, why not test stability?

 

[m1dg3t]

Guess I'm crazy? Using p95, I run a full 24h on each small/large/blend. AVX on when applicable and error checking/reporting 'on'? Been a while since I did any stress testing.

Obviously it takes some work to get it 24h stable, don't just make adjustments and expect it to be stable. I've been testing small/large with no issue only to have it fail blend, times like that the fun really starts LoLoL

Edit: This method has worked well for me over the years and I find myself using less volts overall and better temps. AUTO volt settings usually result in over volt, I find and apply all the stock minimum voltages and work from there...

 

[slozomby]

no real need to "stress test" for more than a few minutes to see if its stable and get good temp readings. but I'm not looking for any kind of record for daily use OC. once its "stable enough" that a few minutes of testing doesn't crash the system. I play <insert current game here> for a while and if it doesn't do goofy stuff then I call it happy.

 

[Fluffmeister]

I did stress it about 7 years ago, I can confirm it's been stable ever since.

 

[Jetster]

I just do a burn in for 20 min when its new then see how far I can OC it. Then its just a mild OC until I get rid of it

The one answer you don't have there. Burn in just to check that its a good install.

 

[Arctucas]

I stability checked my current build overclock with:

3 hours IBT, at Maximum

5 hours of RealBench, using all RAM

8 hours of x264, normal 16 threads

13+ hours of HCI Memtest

 

[HammerON]

For the ultimate test, let it crunch BOINC 24/7 and if that runs fine, it's fine.

Yep. I do test a over clock before crunching with AIDA64, ROG stress test or Intel's XTU. But the main stability test is being able to crunch 24/7.
Since crunching is one of my "daily programs", that was my choice in the poll.

 

[slozomby]

please get it stable before crunching.

 

[FilipM]

50 runs of LinX with Max Memory for 24/7.

 

[m1dg3t]

I just do a burn in for 20 min when its new then see how far I can OC it. Then its just a mild OC until I get rid of it

The one answer you don't have there. Burn in just to check that its a good install.

Ah yes, the all important burn in! I run it stock for a week or two before I get down to it. I never really push my chips, I just tweek for the highest stable frequency using moderate volts.

 

[ithehappy]

Yeah I do, but only after upgrading my system or something like that. Just need to find at what temp the CPU is running for a custom set voltage after overclocking that's it. Prime and Realbench both of them. But after finding a sweet spot never again I run them, utterly meaningless for everyday usage I might say. But to test you OC capability, I suppose you need these type of software to run, for a brief period of time.

PS: Couldn't vote on the poll, didn't understand whether you asked that on daily basis or one time's point of view.

 

[jsfitz54]

Anything wrong with using Intel's test?
https://downloadcenter.intel.com/download/19792/Intel-Processor-Diagnostic-Tool

 

[cadaveca]

Anything wrong with using Intel's test?
https://downloadcenter.intel.com/download/19792/Intel-Processor-Diagnostic-Tool

Intel XTU is better.

After so many years of doing OC, I don't need P95 or IBT or even AIDA64 to know if my CPU is going to be stable or not. The one thing I sometimes have issue with is IMC stability, and that's with just using XMP @ 3200 MHz+. None of these "stress tests" actually tests that particular thing properly, and I have tried everything. Memory testing is easy, use Memtestx86, find first mem multi that isn't stable, how it fails (which test) will tell you about what tweaking is needed, done.

CPU testing is equally simple... find the lowest voltage a multi will POST at. Take that voltage, use the multi below that, and add 0.025V. Done.

Then do regular stuff, render, play BF, whatever... if it crashes, and you've only tweaked one thing, you know what you need to do... CRAM MOAR VOLTAGE!!! j/k. Back the clock off a bit.

24/7 OC testing doesn't take much, or much skill. tweaking for benchmarking for HWBOT... now that's something entirely different. Go ask those guys if they use P95 or IBT.

 

[AsRock]

I run the system though OCCT after that i never bother again unless there is a real issue going on. OCCT typically runs a system over what your normal gaming will ever do.

Prime \ IBT i don't bother with.

 

[R-T-B]

I do Prime95 testing. Why? Because I like to err on a conservative overclock heatwise. If it can run an hour of Prime95 without heat-throttling issues, my overclock is not too hot running. I like that.

 

[INSTG8R]

Nope, not anymore I am happy with my 4.4 "stock" I don't think it requires any stress testing-

 

[fullinfusion]

Do you actually stress test your CPU?

you need to change the pole from 10 times to 100% stability... I myself go for the lowest voltage at a high clock... and use AIDA64 and select just the FPU setting to heat the cpu up to it's MAX..

If It passes that run after a few hours it's 110% stable imo and have yet to have a BSOD.. and btw I didn't pole because it's lacking other options

Intel XTU is better.

After so many years of doing OC, I don't need P95 or IBT or even AIDA64 to know if my CPU is going to be stable or not. The one thing I sometimes have issue with is IMC stability, and that's with just using XMP @ 3200 MHz+. None of these "stress tests" actually tests that particular thing properly, and I have tried everything. Memory testing is easy, use Memtestx86, find first mem multi that isn't stable, how it fails (which test) will tell you about what tweaking is needed, done.

CPU testing is equally simple... find the lowest voltage a multi will POST at. Take that voltage, use the multi below that, and add 0.025V. Done.

Then do regular stuff, render, play BF, whatever... if it crashes, and you've only tweaked one thing, you know what you need to do... CRAM MOAR VOLTAGE!!! j/k. Back the clock off a bit.

OC testing doesn't take much, or much skill. tweaking for benchmarking for HWBOT... now that's something entirely different. Go ask those guys if they use P95 or IBT.

Um sure... and Yeah what version? I notice variables in a few different versions

ANY stress that can utilize "JUST FPU" imho is better for stressing then the rest so, if it passes that then its a 24/7 pass as far as I'm concerned

voltage don't always mean moAR the better...

 

[trog100]

one good test i like to do is run a "mild" cpu stress test (not prime95) in the background whilst doing normal browsing and the likes..

i forget all about it and if everything is okay for two or three hours without any lock ups i assume the machine is stable enough..

but some kind of stress testing is part of any overclocking or voltage meddling.. it has to be to find acceptable limits as regards what the hardware can do.. a good basic rule is find its real limits and then knock it back a bit (5% - ish) for every day use..

apart from benching there is little to be gained by pushing things too far for evey day use..

trog

 

[P4-630]

I have stress tested a few times with AIDA64, I have found that's more realistic than Prime or other stress tests.