When stress testing your CPU overclock do you test for 24 hours or do you think it's an arbitary test length so go for...

When stress testing your CPU overclock do you test for 24 hours or do you think it's an arbitary test length so go for a shorter duration of testing?

about three fiddy

intel burntest

Wait until the hottest ambient temperature that it will probably be in then keep it at max temp, 15 minutes is usually enough to satisfy me.

>stress testing your daily driver for a solid 24hrs

no thanks, I have work to do

does your mom pay the electric bill?

I test an hour with the OCCT test (not Linpack) and then proceed to test my actual workload. I've got a 4790K and I find that stress tests are incredibly unreliable in determining instability, I've had settings that never crashed in P95/Linpack for hours but running a video game would kill it in

Watching anime, while fun, is not work.

this, this 24 hour stress test is a meme. I do it for say 5-6 hours

1 hour max

Memtest and drive tests I'll run over night. CPU tests though I just wait for temps to stabilize and then give it a few minutes. I use that time to make sure any thermal throttling profiles are doing what they're supposed to be doing.

Of course, who do you think bought his Razer gayman computer?

Prime95 blend test for 30 minutes. Ill have hwmonitor side by side while it's running and peek occasionally at temps and to make sure the CPU isn't falling out of its boost clock. (Thermal throttle).

I stopped overclocking my CPU forever ago though. An i7-6700k even at its stock 4.2ghz boost is more than enough. I'm undervolted now though.

I've had overclocked processors fail 17 hours into prime95.

By all means run a short 15-minute test when you're poking around trying to find out how far your chip will go. But when you're going to settle on one setting for long-term use, 24 hours is a minimum to call it stable.

I usually just install it and start working right away, if it craps.out from regular workloads the stress test probably would've killed it faster

None of this should matter if you have warranty. You DID buy an extended warranty, didn't you user?

Actually theres more to it than just one test.

The problems i was running into with my OC were happening literally once every week every Friday evening .

I almost got superstitious over my PC boot looping exactly every Friday evening, but it turned out it was my own fucky clock/voltage settings that were causing boot failure BUT ONLY SOMETIMES

So it requires extended testing.
And if you ever get any blue or black seen or a boot loop EVEN AFTER it worked OK for months it's your OC.

I made two profiles. One that can pass 24h in x264 and another that can pass 4h. So far I haven't had any issues with the higher OC that doesn't make it 24h.

What are you doing that requires a load for over 15 hours? Seriously. Been running 3 years now with a OC that doesn't make it past that and have had no issues what so ever. 24h is a meme. If it makes it overnight you're done.

>>What are you doing that requires a load for over 15 hours? Seriously
You're saying that as if the failure couldn't recur at any time when the CPU was under load. Its not like the CPU accumulates stress when under load and then eventually fails, it has a certain chance of failing with every operation. If that chance is low prime might take a long time to find it, but it is nonzero, and your unlucky number could come up at any point.

That system would occasionally have a game crash-to-desktop or something on the order of once every two weeks or so. For a long time I thought nothing of it, thinking it was just the game being buggy, the fault of the graphics driver, etc. Until it happened at an inopportune time and I got mad. I redid all my stability tests (it passed memtest and failed prime after that long period), and then backed it off a bit, at which point it passed prime for 24 full hours. My game never crashed again.

We'll I'll just give you my anecdotal evidence here running a OC that didn't make it 24h. 23 day uptime, most if it with the load you see here. Haven't had a blue screen that wasn't bad Nvidia drivers.

So either you've been lucky, or the failure wasn't big enough to cause a crash and went unnoticed.

I guess I'm lucky. The OC definitely isn't 24h stable. I kinda just went with it for shits to see how long it would last but I haven't had a problem so I kept it. I'm going to do the exact same thing with a Ryzen build in a couple weeks that will replace this so we'll see if it's really luck.

I do it for 2-3 hours at length.

This is usually good enough for me. When I play games, they don't max out the CPU cores, and I only really play it few hours at a time anyway.

However I use multiple stress test to see if its stable or not. Prime95, Cinebench, IntelBurn, etc.

The stress test are so long so that they keep your thermal interface warm enough for long enough to deform slightly into a a state which suits extreme conditions. Especially relevant for new builds.
Though it also highlights fluctuations and variety in the whole system if you keep it running for a day. You can afford to run that stress test.

Yeah, I just use bf4 for stability testing.

This. It's essentially random if your overclock will fail or not. The only way to make sure there's a low chance of failure is to allow a sufficient amount of time. I'd personally run it overnight a few nights and see if I get anything bad when I think I've settled on a stable clock speed.

i just stress test whenever i'm occupied but also when i'm able to check the pc every couple of house. like i'd do it on a weekend when i have mates around and when we're chilling playing xbox or whatever i'd let the stress test run for the whole day but check on it every hour or so.

there's no point in doing a full day stress test if you're going to be away from the pc since it might blue screen and you won't know till 8 hours later when you get home/wake up or whatever. that's just lost time.

couple of hours*

24 hours sounds excessive. 1 hour is probably good enough. I think by 2-3 hours there's very little chance it's not stable.

What do you use for drive tests?

1 hour is good enough, after 24 you're just testing your cooling system.

>testing for more than 20 minutes
You people are retards

I just read somewhere that you use Intel Burn test for 10 rounds. Takes about 30 minutes.

I OC'ed to 4.5GHz, it failed in the first try, then I upped the voltage by like 0.1V and tried again and it worked.

Never crashed since. Maybe it's not perfect but whatever

honestly, if it doesn't crap opening notepad i say fuck it and put the side back on

15 minutes if I'm feeling like wasting a lot of time, but usually I'll start it and just start my work at the same time

I've had my computer on for 8 days straight so I think I'm done testing now.

In 15-30 minutes it should reach its max temperature. Anything more won't give you any more different results. If it's not during summer, you may want to retest in a summer day when it's hot, just to make sure, especially if you're like me and live in a place where you get both -30 and +30 degrees Celsius throughout the year.

If it starts up, I give it a 15 minute run. I am constantly monitoring the 15 min. run, if no artifacts or instabilities happen, I let it run for like 2 hours and do other stuff, just occasionally check if it runs.

I never do long stress tests simply because it's unrealistic for most use cases. I test my overclocks by using them and seeing how stable they are in what I do. Maybe an hour in p95 or so, but nothing beyond that

Given enough time prime95 will fail even on stock clock processors.

jsut run it for 2h and you should be fine

If using P95 don't use Blend because it mainly stresses the IMC and RAM so it won't stress the CPU that much.

IntelBurnTest exists you know..

one bluescreen every 6 months or so is stable desu

I hope my cpu isn't getting too hot.