How long do you stress test your CPU OC stability?

I put a stress test on and go shitposting on my thinkpad.
If it's still running after 15-20 minutes it's probably stable enough.

I don't.
I used to undervolt because Aussie summers before I had working Ac and good case air flow now I just don't bother

Oc is a meme unless you undervolt these days

Hour in OCCT and Prime95, then a week of regular use.
My ryzen works on 3.9GHZ 1.3v (4GHZ is only stable on 1.4v but going an entire 0.1v for only 100mhz gain is not worth it.)

zoomers
I once got an error on prime95 at like 20 hours
even that little instability will lead to data corruption over time

It will lead to data corruption anyway if you leave your pc 100% use for 20 hours everyday

no

expose it to 3.6 roentgens

Me personally I have a 20 min rule if it passes that then I have a semi stable oc then I stress test it for 40 mins to an hour if it doesn't crash its probably good. After that its real world tests to see how it fares but I mainly play games so I dont think im ever 100% stable.

t. 2500k owner @4.9ghz

One pass of passmark, one firestrike physics test. If it’s unstable it will crash like hell.

Zero. Do you think data centers stress-test?
Or laptop manufactures?

I run my software and of it works then I know is a good chip.

Why the shit would I stress test? Overclock and then start working with it. If it crashes, you overdid it. It's not like your CPU is gonna break. Worst you can get is a bluescreen

Attached: 1561065224201.gif (300x300, 139K)

Depends, if you're just playing games and don't use your PC for anything productive you can stress test for an hour or so, in the rare case it does fail you won't lose anything useful anyway, but for my builds at home I run for a minimum of 48 hours, sometimes up to a week. I have a job though

Undervolting gave my laptop like 2h extra battery life and it always ran cool. Ran like -.1 on the CPU and GPU. Good battery life, good temps, never throttles.

If you have a laptop and you arent undervolting, what the fuck are you doing?

Why the shit would you not stress test? It's not like the cpu is going to break if you do. Worst you can get is a blue screen.
(Also while working worse you can get is lose your work and or get annoyed)

>t. 2500k owner @4.9ghz
Noice, what cooling setup?

Dunno my last one was over a decade ago.
Dont use them anymore as faggot gamers ruined that segment with overpriced tacky shit

>stress test
>stability
Lol

If it boots and registers all RAM it works.
If it loads my OS without failing it works.
If it runs [insert current high-end game] it works
If it crunches a video edit without crashing it works.

If it doesn't work I up the voltage or lower the clock or re-seat the RAM.
I'm not wasting my time and CPU life burning it up with stupid tests.

3 hours prime95 custom blend with 80%+ system memory
2 hours asus realbench

Then multiple 1 hour stressapptest under WSL to see how much i can rise cache/uncore before it hangs and downvolt vccio and vccsa

Attached: RAM-VCCIO-VCCSA.png (1024x600, 162K)

Corsair h60 a buddy of mine didn't want anymore I have taking in fresh air from the front really helped cool it down.

What voltage? I'm a pussypants that won't go over 1.4v, even with this fuckhuge Evo.

Install Gentoo.

My overclock is:
9900K 5.0GHz at 1.3V in bios (1.25 under load)
4x8GB 3733MHz 16-16-16-35 with optimized sub timings

CPU stability test has been 24H on Aida64, 6H in Realbench, and 12H in Prime 95
RAM testing is 12H in Karhu RAM test and 24H in HCI memtest

1.45 with the highest llc setting I forget what else honestly don't do it im more or less actively trying to kill this thing after 4.6/4.7ghz the improvements are pretty damn minimal.

1h with OCCT and then regular usage. I find that stress tests aren't very reliable for modern CPUs, what with all the power saving features, downclocking, boosts, XFR and all that shit. A stress test is basically just a baseline reading for me nowadays, actual stability is only validated through real use with a lot of different programs and all the clock adjustment features going at it. I've had CPUs pass 24h stress tests only to repeatedly fail 10 minutes into a game, shit's basically a waste of time beyond confirming baseline stability and cooling limits.

I just checked real quick 1.43 volts is what I have it under.

>3733mhz
>CL16
kek

I don't. Stress testing is a non-deterministic activity - if your PC crashes, it's unstable. If it doesn't crash, it may crash later.

Therefore, I just use it until it crashes.

t. brainlet

aida64 for stable overclock, prime95 if you want to overtest. 2 hours is good. my profiles don't work anymore :(

~20 minutes but with multiple different applications. Prime95/IntelBurnTest/some games

I find that long gaming sessions or large batch video encodes tend to find instability pretty well. No CPU is perfectly un-crashable, so if you want to know if your setup is stable then just use it and see.

I let Prime95 run for four, or at least I was going to until core three crashed and blue screened about 2~2.5 hours in.

Turns out both the MoBo and CPU needed a BIOS & microcode update. After that was done the machine made it to the four hour mark ezpz.

This poster is delusional, take him to the infirmary.

Y tho?
I like to keep my shit cool.

I can do 4000 16-16-16-35 but I need to pump 1.5V through and it gets pretty warm after a while.

I think I did 6 hours on my 4670k.

Attached: 1403156794425.png (699x854, 222K)

Until it crashes

If you are mostly gaming then should do the job although I had cases where it OCCT or P95 popped errors after 45-50mins. If you do any actual work on your computer then 2-4 hours would be a better idea.

5 minutes on prime95 and a game

>5 minutes on prime95 and a game
I get that you don't like wasting time but 5 min really won't tell you anything

Will this ever get fixed its been here since the start hasn't it?

not terrible, but not good either

MOOOOOOOOOOOOOOOOOOOOT

30min is fine.
If it's gonna fail, it's gonna be within 30min.
I even use OC voltages that are unstable with Prime 95 or OCCT.
Because normal usage won't stress a PC like an artificial test will.

10 minute. I still worry about heat dissipation of i9 overclocked to 5ghz on air-cooling.

i9 had stim IHS returned.

Attached: Stess test.jpg (3693x1925, 1.17M)

I stress test my CPU not for stability but for temp nowadays.

what the fuck is this

>How long do you test your CPU overclocked stability with a Cpu stress test? 1 hour? 4 hours? 24 hours?
After overclocking my CPU and memory I did a night of memtest86, two days if prime95 and a couple of rounds of 3DMark benchmarks, when my clocks were way too high I got system hangup on prime95 and when they were a bit too high I got timing error on 3DMark but no problems on other two programs.

>If it's gonna fail, it's gonna be within 30min.
My prime95 run didn't fail before several hours into the testing

3 runs of CB20 back to back, then i run it till it crashes and tweak settings till it never crashes again

Then something else is causing that.
Potentially a bad power supply or insufficient cooling.

After dialing the initial overclock I start with a series of quick tests (LinX or Prime95). Once I have settings that clear those two, I get y-cruncher up and running, turn on all tests, and let it cook in 5 minute, 10 minute, and 20minute/test intervals. Each test cycle runs a minimum of 12 hours.
If all of those clear, then I call it stable.

Winter time, run prime95 and furmark together overnight. If it crashes, keep doing it again. Haven't needed to test more than a week of sleep.

Every day is a stress test and I'm afraid that today may be the day that the inevitable failure comes

>find the lowest voltage at which it's barely stable in aida or realbench for a couple of minutes
>add 20 mV to it
>test 15 minutes with realbench 2.4x, 10 minutes with realbench 2.5x, 5 minutes with a modified x264 stability test, 5 minutes with prime95, and finally convert a 4k h265 video with handbrake
>if success, add another 20 mV to it
>profit for the next 5+ years

I don't. If windows boots up, it's fine. If it starts crashing, then I dial it down.

That's actually quite significant.

>4GHZ is only stable on 1.4v but going an entire 0.1v for only 100mhz gain is not worth it.
Pussy, I'm running up to 1.43v with my LLC at 4.1GHZ on my 1600.

Stress tests never work for me, I use games for that, BF4 to be precise.

OC could sustain 8 hours of prime95 yet crash 15min into game.

I had the same issue with BF4, where it was always pickier than any prime95 or cpu load test. Overwatch was always pickier for my GPU loads, too.

This to calibrate. 50 minutes after tunning if I care about workloads other than gaming.

I don't. I trust the manufacturer's QC.

I just play games and see if it crashes or not. Why would I bother doing stress tests to achieve conditions that will never happen? Only playing directly is a stress worth doing.

When testing on windows, i first do a few consecutive cinebench runs, and if it passes, i run occt or prime95 overnight, so about 10 hours. I've done this in 2017 when i built my desktop and my r5 1600 has been running at 3.8GHz 1.32V, and my ram at 3200MHz from the stock 2400 without an issue ever since. Did the same on my laptop before that, in 2015 and it ran stably at 4GHz for the whole time i used it.

how's your chip degradation holding on user?

Great I'm still stable and only had to up the voltage by .1 to keep my OC.

Usually just a run of CPU-Z built in benchmark and stress test

Normally Prime95 and linpack for 30 min. If linpack doesn't overheat or crash it, nothing will.

2months.

I run 7 different cpu's at 100% load 24/7 running WCG, that weeds out the weak silicon

Linpack should be more well known.
software.intel.com/en-us/articles/intel-mkl-benchmarks-suite
Here's a custom lininput_xeon64 good for up to 32GB you can try. It performs a lot of tests on small inputs decreasing to 2 tests on the 32784 input. Then it will run a single pass test with the max available ram on your system.
14 # number of tests
1040 4112 8208 16400 32784 60016 55024 50016 45024 40016 35024 30016 25024 20016 # problem sizes
1040 4112 8208 16400 32784 60016 55024 50032 45040 40016 35024 30032 25040 20016 # leading dimensions
64 16 8 4 2 1 1 1 1 1 1 1 1 1 1 # times to run a test
4 4 4 4 1 1 1 1 1 1 1 1 1 4 4 # alignment values (in KBytes)

>a week
Christ how stable do you need it to be?
Just save regularly if you are that concerned