PC will shutdown randomly while playing games.

[zwerewolfripper]

Hi all you can just call me Josh, I humbly request help with an issue I have been having for two days now. I'm at my wits end here.

The issue - while playing games, my pc will randomly shut off not blue screen just no power and restart. The two games it happens so far with where Overwatch which happened first and I thought it was an overheating issue, and the second was with Anon 2205. I haven't played anything else because I was spooked.

The backstory - this as I said started to happen two days ago, after a black out on my street but the thing is I have a UPS unit and was able to safely shut down my pc. I also recently upgraded my GPU, from a GTX 760SC to a GTX 1080 Ti.

The specs -
CPU - i7 5820k
PSU - Corsair HX850
Mainboard - Gigabyte X99-UD3 (I'm aware it's not the best board)
GPU - GTX 1080Ti FTW3
RAM - Corsair Vengeance LPX ddr4 16gb
Cooling - Corsair h100i

I know it's a mash up of parts but I'm slowly upgrading please don't judge me....

Intial speculation - I thought because overwatch has a history of some people haveing overheating issues with it, it was that and that's why I initially thought ahh cramped case it's just hearing up, but I monitored the temp in the case for anno and the hottest it the gpu every got was maybe 45°C so it's running cool and that was while I was streaming. Im pulling out my hair at this point.

Any help would be greatly appreciated!

 

[Maarsch]

Do you have the specific Blue screen error?

When you switched GPUs, did you uninstall drivers and then re-install, or just update? Perhaps using DDU to properly scrub the drivers and then reinstall just the 1080 ones will help.

 

[blockhead78]

1. if you have any CPU/GPU overclocks applied, reset them back to default
2. make sure you're running the latest bios version for your mobo
3. run a DDU driver wipe and re-install latest drivers from nvidia

if still having the issue:
4. stress test the CPU (aida64 or intel burn test) & GPU (furmark) separately and see if either of them cause a crash and check the temps on both

comment on the post with the result

 

[zwerewolfripper]

Ok I have just finished using DDU, and reinstalling my drivers, no I never got any blue screen. I also never had the GPU/CPU overclocked. About to see if I still the shut off.

[strike]**Update**, the drivers appears to have been the issue, no idea how, thanks alot guys, usually it cuts after 5 mins in the menu of overwatch, I left it in the menu while streaming for about 7 minutes and it ran fine, thanks again. [/strike]

So it cutoff again after 6 mins in the overwtach menu. Im running the intel benchmarks now.
I ran the Intel burner, at high stress for 5 times, it came back reading as stable.
I ran the Furmark, the temp rose to about 77C which is the highest i have ever seen it and was stable,.

I did notice however that ever since doing the ddu, the gpu clock at idle is a lot lower, probably due to a conflict with the older driver before using DDU the gpu clock was always at 1500mhz now its at 139mhz. The when running the furmark it would go to abou 1750mhz and no more. However playing a game the card would go past that and it says on the XOC, that the boost clock is at 1750, so Im thinking that the gpu is shutting the system down when running past the boost clockspeed for an extended time. What do you think?

 

[eballer48]

Most likely your GPU/CPU is overheating and it shuts down to prevent damage.

 

[zwerewolfripper]

So I limited the card to temperature of 77C or 80% power target, and it ran in the Overwatch Menu for about 10 mins it was fine, the highest the temp rose to was 65C, I then decided to set it back to 100% power target or 84C and it stayed at 65C but cut out after a minute.

 

[dpkjha]

uh this happened to me before my fix was going to the power settings in windows and resetting them to default

 

[zwerewolfripper]

i tried that, nothing doing.. I have believe its the psu being unable to provide the power for the GPU at 100% power consumption, because so far thats the only thing I can do to stop it from shutting off, is limiting the power draw by the card, to about 85%

 

[Achint2000]

850W PSU is more than enough for a 1080Ti and an i7 7700k. 1080Ti can take about a peak maximum possible of 500W, your CPU cant take more than 150W (91W TDP) and rest of the system hardly takes so much power.

So, it's highly unlikely to be a PSU issue. It's more likely to be a clock speed issue.
Use NvidiaInspector and lower the clocks by about 20 MHz on both core and memory.
Use ThrottleStop to see if your CPU is throttling.
Goto nvidia control panel, click "Manage 3D Settings" and set "Power Management mode" to "Prefer Maximum Performance".

You can also use NvidiaInspector Commandline.

Steps to force default clock speeds:

1. Create a shortcut to nvidiainspector.exe and at the end of the target, type -forcepstate:0,16 and run it. This will set your clocks to automatic.

if that doesn't work, Steps to forcing a user defined clock on P2 State:
2. use forcepstate:0,2 at the end of target,
3. then goto nvidia inspector,
4. click "show overclocking",
5. select "performance level 2 [P2]"
6. click "Unlock Max" on the right side,
7. slide your clocks right upto the default speeds.

That'll FORCE the clocks you set.
If it keeps zapping back, increase with 20Mhz at once and keep clicking apply after each increase.

You can also try forcepstate:0,0 but that sets my GPU clocks to 540 MHz so I prefer P2 instead.

 

[zwerewolfripper]

Ok, so windows 10 wont let me create the shortcut, "\nvidiaInspector\nvidiaInspector.exe -forcepstate:0,16" keeps saying invalid target, not sure how to get around that.

 

[Achint2000]

My bad, wasn't clear in my previous answer. You have to type the parameters after the quotes.

Example: "C:\Example\Example.exe" -parameters /parameters

So it'll be: "\nvidiainspector\nvidiainspector.exe" -forcepstate:0,16

 

[zwerewolfripper]

ahh ok well i did that. This is what its currently set as in Nvidia Inspector https://ibb.co/iPnM8k this is what it use to be set as in
XOC after using DDU https://ibb.co/kKShEQ, for comparison, this is what it looks like under any load https://ibb.co/dLUCEQ
I dont belive the updated clocks where taken, I will use the manual method to set them, So i did the differing methods, and the only change i noticed was in stead of having a idle clock speed of 139mhz, it rose to 1569 mhz which is the stated base on XOC. I will check now to see if i get the issue.

 

[Achint2000]

Gotta say those are some powerful clocks there.

Then, I use a more clean UI with less style when it comes to monitoring, so could you send HWinfo screenshots instead?
Check online, if the GPU clocks you've set are default. Set them to default first.

Then, send HWinfo Screenshots:

Spoiler

Start HWinfo with "Sensors Only" ticked, scroll down almost to the bottom to find GPU sensor values, right click and "Show Monitor" on all the purple highlighted sensors here: [IMAGE:]

Whichever clocks you're using, you have to make sure those clocks are stable.
Start a GPU Benchmark like Heaven/Valley or Furmark, on extreme.
Stress your GPU as high as possible, and wait to see if it crashes.
Run all benchmarks you want, in windowed mode first, and let the sensor graphs be on top.
Wait for atleast 30 minutes and see if it crashes or not.
Pay detailed attention to the GPU Power Graph, as that could be the issue.

If it doesn't crash, your GPU is fine and something else must be wrong.

I've had a recent annoying issue with Shadowplay, it messed up my userprofile and made my system crash every 5 minutes after rebooting. Updating it helped, so disable shadowplay and anything else like fraps/bandicam or any screen recorder, just to be sure.

 

[zwerewolfripper]

Ok currently running furmark, and I noticed it never pulls more than 850 volts I'm only about 5 mins into the benchmark atm but it seems stable and the clock speed for the gpu never goes past 1657 mhz as well. Meanwhile in overwatch it pulls like 1083 volts and gpu clock at 1950mhz

UPDATE - Ok it cut off during the benchmark at about 12mins before the 30 min mark. The GPU Clock was hovering round 1657 mhz, the power temp was stable at 79C, the min voltage was 800V, max voltage was 894V , avg voltage 847V.

 

[Achint2000]

Voltages are in mV... then, try heaven or valley, furmark at times strangely uses lower clocks in my experience.
Try any other game which is as heavy as overwatch and see if that makes it crash.

Those voltages look too low... default 1080ti maximum voltage is around 1.00v or 1000 mV.
how exactly does this happen...? does the screen go blank first? or does it simply reset or does it properly restarts?

 

[zwerewolfripper]

Thats the thing, the benchmark takes that voltage but overwatch takes the voltage your speaking of, so yes it simply resets, also running the other benchmark now.

Ok So i ran it on the Heaven Benchmark, At https://ibb.co/md7mb5 it ran at https://cdn.discordapp.com/attachments/189167909032034304/360886521286623232/20170922_162428.jpg Note the .800V is when its idle its at that

Update - Ran heaven at same as before settings but these settings got to 21 mins before shutting off https://cdn.discordapp.com/attachments/189167909032034304/360892519505985536/20170922_165554.jpg
So im thinking i need to lower it even more??!?!

 

[Achint2000]

Just run it on Default GPU Clocks. Read the manual your GPU came with, or see online (I dont know what clocks a 1080Ti is on) and make sure you're on default clocks or just click that button on EVGA Precision X.

Test all again on default clocks and see if it crashes. If it works on default for 30 mins properly, leave it for an hour and check back on it later. If it works for 1 hour, leave it for half a day. If it still keeps working, leave it overnight. After that, you can be sure everything's fine.

Then, the links from "ibb.co" report a bad gateway so i can't access them.

If it's crashing on default clocks and it's suddenly just "shutting down", it can indicate a serious issue with hardware.
It could be anything, we can't be sure, and finding out what exactly is causing this can take a long time.

UPDATE: Click "Priority" thing in the software. You DONT need temperature to be prior to power here because the cooling design automatically takes care of that. Set "Power Target" to the highest possible.

Then, also pay detailed attention to "GPU VRM Temperature" in HWinfo. In my Strix 970, the VRM isn't covered with heatsink and is totally dependent on airflow for cooling, and VRM reaches upto 110C on my 970, could be a LOT worse on a 1080Ti.

UPDATE 2: Nvidia Geforce Experience. UNINSTALL IT. It's been crashing my system because of it's crappy gamestream service in the updated versions. It simply makes the entire system freeze and not respond at all. This has been the piece of garbage that's been making my system crash every 5 minutes since past whole week...

UPDATE 3: If you're using Windows 10, you must try reset your PC if everything else fails. Win10 has alot of configs and resetting it to default can help!

 

[zwerewolfripper]

Ok guys so update, I redid my os and just sat in the overwatch menu for an hour at defaults without any crash, I also not that the temperatures are no richly lower after redoing the os the highest it goes to is 60c on the power... So I'm thinking it was a software and temperature issue I have not had a crash since redoing my os. And I have been pushing it to try and force a crash.. (pushing within reason)

 

[Achint2000]

"Redoing" your OS. It must've been Windows power plan settings or something...
What exactly did you do? Reinstall Windows?

Forcing a crash... good luck finding your GPU's OC limit.

 

[zwerewolfripper]

Yes I reinstalled windows, it may have been the power settings, but I did reset the power plan settings to default and it didn't fix it before. I really have no idea. All I can say is the temperature never goes past 65c today and no crash. The gpu clock was stable at 1900+mhz for an hour, the power was stable at 1.060mV for for an hour..

That's the thing I never overclocked my gpu. It's set to the default clock. That 1900mhz is the boost that's default with the card.