Random BSOD caused by overclocking?

[thegeek2 7]

So, about 10 minutes ago, my computer BSOD'ed then restarted (without managing to create a minidump file). This is the first time it's blue screened, and I've had it for about a year. I built it, and I'm running an i5 6600k overclocked to 4.4 Ghz. When I first established the overclock (also about a year ago), I of course stress-tested it and had no issues, and have had no issues until today. Would the overclock have caused my bluescreen?

[Fulcrous 2,073]

open event viewer.

click on critical events.
expand by right clicking at critical events

show log for most recent critical event.

[thegeek2 7]

Already did. Nothing besides the Kernel Power and Dump Error I can see. Screenshot of the log attached.

No other critical events besides the kernel-power.

[Hally 126]

I'm kinda doubting it is from the overclocking.

If your computer reboots from overheating, caused by thermal paste degradation.. it would usually just reboot immediately, without even giving you a blue screen at all.

Could be just a one time windows thing...

---

Does this actually happen multiple times? or is it just a one time thing.

[thegeek2 7]

Well no it wouldn't be overheating. Overclocks can sometimes be unstable which means bluescreens

 

I'll give Prime95 a spin, see what happens.

It happened once

 

I did have a Graphics driver failure a month or two ago. I updated my driver after that, but I suppose it could be GPU? Thing is it didn't bluescreen that time. Just went blank.

[thegeek2 7]

Well, it made it through 15 minutes of Prime95 with no issues, so maybe it's not the overclock.

[BlackAdder 1,390]

17 minutes ago, thegeek2 said:

Well, it made it through 15 minutes of Prime95 with no issues, so maybe it's not the overclock.

15min is a joke, let prime95 runs at least 2 hours.

[Folterknecht 2,257]

@thegeek2

 

Please upload a screenshot where you clicked on the error in the event viewer that was directly before the PC crashed.

 

And I agree with Blackadder - 15 min Prime is a joke. Only good enough to check temps with "Prime small FFTs" for air coolers. What Prime test were you running?

[Fulcrous 2,073]

Well it still helps to know exactly what went wrong, hence knowing what error occurred in the event viewer would help identify the cause...

Kernal Power errors occurs when the PC shuts down unexpectedly. Please expand that so bugcheck parameters may be identified.

[Haswell 2,399]

10 hours ago, BlackAdder said:

15min is a joke, let prime95 runs at least 2 hours.

24 hours just to be safe. There's technically no guarantee any overclock will be 100% stable, just that the longer you can run the chip under full load the more certain you can be it won't roll over and BSOD suddenly (and that's not considering low-power states yet, it might still crash if the mobo decides it doesn't need that much power while not bringing the clockspeed down first).

Post complete specs, specifically the chip stepping (use CPU-Z), mobo model and any setting you touched in the BIOS that isn't on default. Also check temperatures while running Prime95.

My gut tells me you gave the chip too much voltage on your OC that it's starting to kill itself, or the VRMs on the mobo is on its way out. Or overheating because summer.

Examining event viewer is mostly useless because it doesn't tell you what the crash was caused by, usually. Go to the BSOD event and copy down everything it tells you. Extra points if you can provide the dump file for easy reading.

[Fulcrous 2,073]

I mean... I used this to figure out whether my bsods were indeed caused by ocs via event viewer. 

http://www.overclock.net/a/common-bsod-error-code-list-for-overclocking

[Haswell 2,399]

Notice how most of the solutions involve pumping more power into the chip.

SB had a "safe" voltage threshold of 1.38v for 24/7 usage, anything higher will likely to significantly decrease the chip's lifespan. Skylake has similar thresholds at 1.4v with 1.52v as its absolute max according to the spec sheet. If OP has been cranking more than 1.4v into the chip for a year the likelihood of hardware degradation is quite high.

[Stige 103]

6 minutes ago, Haswell said:

Notice how most of the solutions involve pumping more power into the chip.

SB had a "safe" voltage threshold of 1.38v for 24/7 usage, anything higher will likely to significantly decrease the chip's lifespan. Skylake has similar thresholds at 1.4v with 1.52v as its absolute max according to the spec sheet. If OP has been cranking more than 1.4v into the chip for a year the likelihood of hardware degradation is quite high.

Bullshit. I ran my 2500K at 1.52V since day one and at first it was at constant voltage because.. reasons. After 3 years it still showed no signs of degradation and then I fried it trying to boot to windows with 1.7V @ 6GHz
My 3570K ran at 1.56-1.58V for over a year before I noticed and even after that at 1.52V, showed no signs of degradation when I sold it and bought this 6600K now running at pathetic 1.46V because it won't do more than 4.7GHz no matter what I throw at it.

 

 

[Fulcrous 2,073]

16 minutes ago, Stige said:

Bullshit. I ran my 2500K at 1.52V since day one and at first it was at constant voltage because.. reasons. After 3 years it still showed no signs of degradation and then I fried it trying to boot to windows with 1.7V @ 6GHz
My 3570K ran at 1.56-1.58V for over a year before I noticed and even after that at 1.52V, showed no signs of degradation when I sold it and bought this 6600K now running at pathetic 1.46V because it won't do more than 4.7GHz no matter what I throw at it.

What it means by "safe" is reasonable length of use. Sure you might be able to run 1.52v for some extended period of time, but the lifetime of the chip is cut severely by doing so. I presume you enjoy sauna's because those idle temps would be ~50-60

[Haswell 2,399]

28 minutes ago, Stige said:

Bullshit. I ran my 2500K at 1.52V since day one and at first it was at constant voltage because.. reasons. After 3 years it still showed no signs of degradation and then I fried it trying to boot to windows with 1.7V @ 6GHz
My 3570K ran at 1.56-1.58V for over a year before I noticed and even after that at 1.52V, showed no signs of degradation when I sold it and bought this 6600K now running at pathetic 1.46V because it won't do more than 4.7GHz no matter what I throw at it.

It's hard to notice signs of degradation until your chip starts acting unstable while others work just fine. Harder still if you aren't constantly taxing the chip at full power because because how are you going to notice any difference if you're not using it 24/7?

I mean, I can push my 2500K to 5.2GHz with 1.46v and have it pass PCMark, but I wouldn't want to use that as my 24/7 overclock when I can be stable at 4.8GHz with 1.35v.

[BlackAdder 1,390]

1 hour ago, Haswell said:

It's hard to notice signs of degradation until your chip starts acting unstable while others work just fine. Harder still if you aren't constantly taxing the chip at full power because because how are you going to notice any difference if you're not using it 24/7?

I mean, I can push my 2500K to 5.2GHz with 1.46v and have it pass PCMark, but I wouldn't want to use that as my 24/7 overclock when I can be stable at 4.8GHz with 1.35v.

1.52V is overkill and differences between additional 100-200 MHz is marginal but degradation is real. Especially if you use it high on daily basis. 

 

Currently my 4790K is undervolted, same goes for 2600K. I prefer quite and durability over few seconds in benchmark. If i need upgrade i’ll go for more cores... 

And I agree with you about OC test and DD voltages.

[thegeek2 7]

It's not an overheat issue. This occurred while the cpu was essentially idling while I surfed the internet. I'll run a longer prime 95 test on small FFA? Bios changes: modified more aggressive fan profile, core multiplier set to 44 with default 100 mhz base clock. I've never gone above 72 ish degrees on a prime95 test, less on something like OCCT.

CPU-Z report and screen shot of event viewer kernal power error attached. There was no minidump file created from BSOD, it was around for like half a second so I couldn't catch an error code off of it either. I can post a screenshot of the dump error details as well if you want.

DESKTOP-G27ANQD CPUZ

Oh, CPU core voltage set to 1.3 V.

I'm not pushing it that hard, I think?

The errors directly before the time of the crash in event viewer were steam client failue to connect errors. There was no BSOD event recorded aside from that one kernal power and the dump failure, both of which were recorded on the restart..

[thegeek2 7]

As you can see, literally nothing was recorded for the bluescreen itself. You only start seeing stuff about kernal booting and the like in response to my computer restarting after the bluescreen.

[Folterknecht 2,257]

Enough of this 1.5V nonsense - lets get something done here.

@thegeek2

 

1) Do you find the following errors in your event manager " WHEA Loggers ... Error Type: „Internal parity error“ ?

2) Are you running "Auto Voltages" on VCore, QPI, VTT ... or manually set to certain values?

3) What RAM are you using? XMP??? Did you test your RAM when setting up your OC???? - suggestion - let Memtest run from an USB stick over night with your current settings (testing RAM within the OS is BS). That way we can start to narrow it down.

 

[thegeek2 7]

Vcore is manual 1.3V everything else is auto.

I can't upload anymore screenshots because I appear to have hit my allocated cap, but at least sorting all system events recorded alphabetically and looking for WHEA whatever, found nothing. RAM is 8GB G-skill Ripjaws V, 2 4-gig sticks in two channels. I never tested RAM when setting up OC, but RAM itself has no overclock applied. I have run built in Windows Memory tester which appears to be independent of the OS since it runs on a restart, but I can run it off a USB overnight if that is necessary. Checking xmp status, need to reboot to get bios...

 

XMP is disabled.

 

[Folterknecht 2,257]

Keep XMP disabled (for now), just make sure the RAM isnt running above 2133 or 2400 MHz (1066 or 1200) and timing arnt below CL15

 

And please Memtest86+ for RAM testing

[thegeek2 7]

Not familiar with RAM overclocking, but from what I can tell, settings are default, speed is 2133 mhz, and I have no idea what timing is but I saw something that was set to "15". RAM test results will of course have to wait until tomorrow.

[Haswell 2,399]

Pic uploads >>>> https://imgur.com

To enable BSOD dumps, follow instructions here: https://www.tenforums.com/tutorials/5560-configure-windows-10-create-minidump-bsod.html leave it on automatic dump. Also check that your page file is enabled since no dump will be generated if you don't have a page file on your boot drive.

Use this to read dumps: http://www.nirsoft.net/utils/blue_screen_view.html

RAM timings on default *should* be fine, don't worry about it until memtest86 is done running.

Memory aside, did you check if it's stable while on stock clocks and voltages?

[thegeek2 7]

No long burn in tests, I'm afraid, but it has been running fine overclocked for a year. If it was unstable at base settings, there's no way I could have managed a 25% overclock under any circumstances, right? BSOD minidumps were enabled before last night's BSOD, as was a page file, yet no dump was created. Settings have not been changed. Last night's BSOD was the only one ever to have occurred so far for me.

[Haswell 2,399]

So to be clear, there was only one BSOD so far, the rest are just blackscreen/hang/crash/restarts right?

Memtest86 doen't take long to complete, after that run Prime on blend with stock clocks and voltages overnight to make sure there's nothing else wrong. This way you can establish a baseline of what works and play with your OC accordingly.

If I were to be extremely lazy, I'd just bump the Vcore up a notch and leave it there until it crashes again, then up it another notch, rinse and repeat until I get to 1.38v. If it's still not stable I'd decrease the OC since chances are the chip can't handle the high speeds and/or voltage anymore due to degradation. Chips are only guaranteed to last long with stock clocks and power, estimating lifespan with anything above/below becomes a crapshoot.

Though if you had to use 1.3v for a stable 4.4GHz your chip might just be a bad bin since most people can get there with 1.2v-ish.