Computer crash 5 min after game crash 3 times in a row

Hi guys,

Recently I had an issue with my computer suddenly restarting, maybe once every few days, not often.

Computer: 5 years old (but case,buttons, front fans are maybe 10 years old)
i5-7600K, GTX 1070, 16 GB ram, Win 10. OS on SSD .

I looked into what could be and did some obvious things:

1.Cleaned up the inside of dust
2. Started using a voltage stabilizer (Indonesia has blackouts and brownouts, this is something I should have been using before I know).
3.Switched the power socket so the PC wasn't sharing with lots of other devices
4. Upgraded the Motherboard BIOS (first time in 5 years).
5. Turned off Auto restart, sleep and various other things recommended online in the power settings.
6.Run memtest for a few hours to make sure Ram was OK - no errors.
7. Sfc/ check disk : no errors on hard drive.

The first two times it happened I didn't notice the connection but it just happen a third time and it was the same series of events, I am playing Dragon Age Inquisition, modded. The game CTDs, not unexpected when modding a game, then after about 4-5 min, when I am back in the desktop, after using Task manager to end the DAI running on my system the computer crashes (restarts/shutdowns).

From what I can see of my temp (after I restart the computer) they look fine, CPU cores are around 45 degrees, GPU at around the 50 degrees. Sure they may have been hotter 4 min before, but they are unlikely to drop 40-50c in a matter of minutes I think.

My fans do seem a bit old and slow though.

Could I need a new PSU? I have had this one for about 5 years, I think it is a 800 Gold.

What do you think guys?

PSU would definitely be suspect. Temps do drop and raise drastically in seconds so I wouldn't rule that out either. However memtest probably would have stress tested your CPU and multiple hours without a crash is a good sign.

Maybe also get a GPU benchmark program to stress your graphics card. This would push your GPU and power draw to the upper limits.

Lastly a crash dump analyzer may point you in the right direction: https://download.cnet.com/WhoCrashed...

Thanks Evil Dead, I will run A GPU benchmark to test it.

Oh I forgot to add the critical event that lead to system shutting down was 'Kernel-Power 41'
But I am not sure how helpful that is, as far as I understand it that means "computer shutdown unexpectedly"

That sounds like a hardware fault, rather than software. The most likely culprits, as you've already figured out, are overheating components, power supply failure, or both. If your mains voltages aren't reliable, that seems likely to be the core of the problem, and could well have damaged the PSU.

Since it's easy, however, I'd suggest ruling out a heat issue before messing with that. Download and run CoreTemp, but don't use the standard download, as that's got some kind of advertising thing in it. (I forget what, it's been a long time.) Rather, click on the little More Downloads link, and then download the standalone version from there.... you probably want the 64-bit edition.

Run that program and make sure it's registering sane temps at idle.. typically 40 to 60C. Near the bottom, each core will have a Max heat reading, which is the hottest that core has gotten since CoreTemp started running. Leave it running, and go run Inquisition. Play for awhile, ten or fifteen minutes, and then alt-tab back to the desktop and look at the Max values. Or, if it crashes, immediately look at the Max values before the computer dies.

If those stay under 90C, CPU heat is not your problem. (ideally, you want them under 80C. The lower, the better.) The next thing to check would be GPU heat. The most recent version of Windows 10 supposedly has one built right in. PCWorld has the instructions here. The salient bit:

To see how hot your graphics card is running, open the Task Manager by pressing Ctrl + Shift + Esc, by pressing Crtl + Alt + Delete and selecting Task Manager, or by right-clicking on the Windows Start menu icon and selecting Task Manager. Once you’re in, simply head to the Performance tab and look for your current GPU temperature listed in the GPU section, as shown in the image above. This works only if you have the Windows 10 May 2020 Update or a newer version of Windows installed; older versions lack the capability.

It doesn't track it over time, it only shows the measurement at the second you're looking at it, so you have to leave it running during gameplay, and then alt-tab and look FAST for it to be somewhat accurate. It takes only seconds for components to cool down, not minutes.

Once you've ruled out heat, then we can talk about how to test the PSU.

Malor wrote:

The next thing to check would be GPU heat. The most recent version of Windows 10 supposedly has one built right in.

Wait, it does??

tests

Well I'll be damned, it does! Nifty!

Thanks guys, you're all life savers

OK stage 1 CPU Temp check: All OK, DAI for 12 min Alt tab = Core 0-3 Max temps range from 56c-61c

Stage 2: GPU Temp check DAI for 15 min Alt tab = 81c (Alt Tab and straight way)

........casts eye towards PSU

Okay, your PSU is your most likely culprit. Testing that is harder, because it's possible that the computer may instantly crash as soon as the voltage dips. But you can try looking at it with HWMonitor:

https://www.cpuid.com/downloads/hwmo...

Look to see what voltages your motherboard is reporting. They should typically all be slightly above the rated voltage. The "Min" column is what you're most interested in... if any of those figures are below the headline value, after playing a game for awhile, then that's probably your root cause.

If the Min value goes below what it's supposed to (well, more than a little), then I'd consider that pretty strong proof. But if it crashes instantly, you may never see the bad figure. You can try hooking up a voltmeter to some of your white 4-pin power adapters, which will give you easy access to +5V and +12V.... 12V is the most important. (Here in the US, you can get cheapie multimeters for about $10, and reasonable quality ones for about $30. Not sure what the market is like there, unfortunately.)

Oh, also, you might have a multiple-rail power supply; many supplies split power across several 12V rails to increase safety. This can be a bit of a PITA for builders and troubleshooters, but it does increase overall system safety. (ie, if one rail fails, it can only cook what's attached, not the entire computer.) Can you tell us the exact make/model of your PSU? That will let us look it up.

If you've got a single-rail supply, then looking at any connector with one of the main voltages is good enough, but if your supply is a split design, then you're just measuring one of possibly four rails, in an 800W supply. Testing all of them is likely to be annoyingly difficult.

Hi Malor, just a few questions.

rated voltage= headline value = the 'Value column' ?

If so I can see that some are below that, (I am not in game now)

VIN4 value is 1.088v but Min is 1.080v
VCORE value is 1.168v but min is 0.584v

The other numbers Value is the same as Min

I'd buy a new PSU and pop it in. If you still have the crashes then you can always return it.

It's annoying to put a psu in but I find it's more convincing than checking voltages.

Yes, I work from home so my computer is pretty crucial even if I wasn't a gamer

Obligatory warning to be really careful when poking around inside a powered-on computer, even with a "safe" multimeter.

I don't think I will play around with a PSU, it's 5 years old, that's a good run. I'll just buy a new one and pay one of our IT guys in the office to come over and install it.

Brownypoints wrote:

I don't think I will play around with a PSU, it's 5 years old, that's a good run. I'll just buy a new one and pay one of our IT guys in the office to come over and install it.

Heh, that works. That can be anywhere from a 30-minute to a two-hour job, depending on how the computer is built.

merphle wrote:

Obligatory warning to be really careful when poking around inside a powered-on computer, even with a "safe" multimeter.

I don't think there are any voltages inside a computer that can hurt you, but damage to the computer is possible if you do dumb stuff. I believe, however, that in voltage-reading mode, most multimeters use a very strong resistor, so that almost no power will travel over the probes. At the voltages computers use, I think it would be hard to short anything out or cause any real grief.

In other measurement modes, the danger is likely significantly higher. Personally, I only ever use volts mode, because that's all I understand.

Thanks for all the help guys. I am planning on buying a new computer early next year but looks like the PSU will just be coming 6 months early, no biggie.

Malor, do you think this was the culprit: VCORE value is 1.168v but min is 0.584v ?

Brownypoints wrote:

Thanks for all the help guys. I am planning on buying a new computer early next year but looks like the PSU will just be coming 6 months early, no biggie.

Malor, do you think this was the culprit: VCORE value is 1.168v but min is 0.584v ?

I'm not sure, actually. It's possible that the chip just reduces internal voltage when it's not being stressed. I believe VCore is a derived value, where the chip takes 12V as input and steps it down, so it's quite possible that it may generate less voltage when it's idle. Voltage is the major driver of heat output, so dropping back that far would reduce the generated heat quite a lot. I think it might even be a square function, so it might be generating a quarter as much heat by dropping VCore by half.

The thing I really really watch is the 12V rail or rails, because most of the power in a modern computer comes from there. The other voltages are rarely stressed and, as a consequence, aren't as likely to fail.

That's a normal vcore range, which is controlled by voltage regulator circuitry on the motherboard based on CPU resource usage/demand.