Question Computer Restarts Itself Without Warning

[Nickamus]

Hello everyone,

Hope everyone is safe and well during these times

First, I apologise for my very lengthy first post.

My issue is the title of this thread. My question is, how do I further identify the issue? Currently I think I have narrowed it down between the Motherboard, GPU, or the RAM. However, due to so many inconsistencies, I can’t pinpoint the issue.

I will appreciate any other suggestions and will test them out as well.

Below are information of my build, a background of the build, temperatures, what games I have tried and tools, the tests I did, the tests I still intend to try, and their results.

1. Build
My current build:

• AMD Ryzen 7 3800X
  • BIOS on “Optimal”
  • Performance automated by ASUS - went to 4.2GHz @ 1.37V. Sometimes even at maximum! Which is 4.5GHz but didn't note down the voltage (but less that 1.5V)
• Corsair H100i RGB PLATINUM
• Asus PRIME B450M-K
• Corsair Vengeance RGB Pro 16GB (2x 8GB) DDR4-3200 CL16
  • BIOS runs it at 3200MHz (16-18-18-36 @ 1.35V)
• Crucial 525GB MX300 SATA M.2 SSD
• Western Digital Blue 1TB 2.5" SSD
• PowerColor Radeon RX 5700XT 8GB Red Dragon
  • “Default” setting on the “Radeon Software” (performance automated by Radeon - can't remember how high it went. Around 2000MHz I think)
• Corsair Crystal 280X RGB
• Corsair RMx 650W (80+ Gold)
• Asus TUF GAMING VG249Q (via DisplayPoint)
• Windows 10 Pro

2. Background
This desktop was first built at the end of July. Probably overkill for 1080p gaming but wanted a minimum 60fps for Godfall, Assassin's Creed: Valhalla and Cyberpunk 2077 at the highest graphical setting...

For the first few months, no issues at all. Everything went well. I never tinkered with anything and let Radeon and ASUS do it’s own thing - I used “D.O.C.P. Profile #1” and “Optimal” in the “EzSetup” at the BIOS and "Default" for Radeon.

However, early September, problems started to happen. Games would crash with no warning until I check my Event Viewer and Radeon Software. In “Event Viewer”, I get the “Kernal-Power” issue and in the “Radeon Software”, I get the “Default Radeon WattMan settings has been restored due to unexpected system failure”. I have done numerous fixes but they always come creeping back.

I don’t know if the fault lies with Radeon (as I am always with nVidia), and because of the current pandemic, I can’t get my mates to test it on their systems. When idle, watching TV, listening to music, or downloading, I get zero issues.

All hardware and software are all up to date.

3. Temperatures and Utilisation
Based on videos and forums, the following results seem to be normal for my chosen parts.

• CPU
  • Temperatures:
    • Minimum (idle): 30°C
    • Maximum: 75°C
  • Utilisation:
    • Not quite sure. I am pretty sure it doesn't go above 90%. Cyberpunk 2077 doesn't go over 50%
• GPU
  • Temperatures:
    • Minimum (idle): 50°C
    • Maximum: 80°C and 100°C (junction)
  • Utilisation:
    • Pretty much between 95 to 100%
• RAM
  • Temperatures:
    • Minimum (idle): 42°C
    • Maximum: need to check
  • Utilisation:
    • I think it's between 50 to 60%

4. Game Testing
Games played before September:

• 3DMark: Crashes at the first test - within a few seconds to 5 minutes
• Divinity: Original Sin 2: restarts computer after an hour of gaming or less
• League of Legends: freezes and resumes but no crashing - detrimental when team fights occur
• Path of Exile: I don’t think it crashed once
• Tom Clancy's Rainbow Six Siege: restarts computer after approximately 2 hours of gaming or less

Issue for September was resolved by taking off one thin piece of Lego and no issues for several hours of gaming on any game (longest running game without closing it was 6 hours with Path of Exile). Yes, I use Lego to hold up my GPU
So the culprit (maybe): Motherboard’s PCIe socket is faulty or GPU is faulty

Games played after September till Cyberpunk 2077 came out:

• 3DMark: no issues
• Assassin’s Creed: Valhalla: no issues
• Cyberpunk 2077: crashes vary from seconds to several hours. Also, I experienced my first crash to the desktop and sent the error reports to AMD/Radeon and CD Projekt.
• Godfall: no issues
• Path of Exile: no issues
• Warzone: no issues

Issue for Cyberpunk 2077 is still undergoing and I’m trying to root out the issue. So far the fix currently is to return the RAM - which leads me to believe that the RAM is receiving insufficient power.
So the culprit (maybe): Motherboard’s RAM socket is faulty or RAM is faulty

Overall it could be:

• Motherboard sockets are bad (since it is a cheap ol’ thing)
• Bad luck with faulty GPU and/or RAM
• Cyberpunk 2077 strains my computer badly

5. Tests (September)
Tests I’ve done to try and solve the issue in September:

• Test 1: Turned off fast boot
• Test 2:
  • RAM back at base 2133MHz and CPU controlled by ASUS
  • For the GPU, on the “Radeon Software”, I left it at “Automatic”
• Test 3: Same as Test 2 but for the GPU, I left it at “Undervolt”
• Test 4: Same as Test 2 but for the GPU, I left it at “Overclock”
• Test 5: Reseating GPU and RAM, making sure all cables are plugged in and not loose, reseating and applying new thermal for
• Test 6: Try the original PSU cables (I bought premium Corsair cables to rule them out)
• Test 7: Try onboard GPU
• Test 8: Unplug all unnecessary devices (only the keyboard & mouse connected)
• Test 9: Try each RAM individually and different sockets individually
• Test 10: Reinstalling drivers
• Test 11: Remove my Lego stand to stop the GPU drooping down

Results:

• Test 1 to 10: Issue still occurs
• Test 11: Issue is gone. Concluded that the Lego was too high and it must be the motherboard with a bad PCIe socket or the GPU had a bad connection because of the Lego.

6. Tests (December)
Tests I’ve done to try and solve the issue in December is exactly the same as September, however, with a set of additional new tests:

• Test 12: Using two dedicated 6+2 PCIe cables instead of one splitting into two 6+2 cable ends
• Test 13: Tried setting RAM at 3200MHz @ 16-18-18-36 1.2V
• Test 14: I've seen people saying to try MemTest86, so will need to try that
• Test 15: Try other games and see if the issue persists (when RAM is back at 3200MHz @ 16-18-18-36 1.35V)

Have not tried the following tests yet:

• Test 16: Prevent ASUS from overclocking the CPU as well - as default it’s all automatic. However, I'm not exactly sure how to do this yet as I thought resetting back to default settings meant turning it off.

Results (so far):

• Test 1 to 15: Issue still occurs. Other games do not cause the same issue as Cyberpunk 2077.

Edit 1 (18/12/2020): Updated the tests done in December
Edit 2 (22/12/2020): Updated the tests done in December

 

[Roland Of Gilead]

Hey there,

Everything you've described points to an issue with the PSU. Random restarts (typically at load) are telling. Your PSU is rather good, but that doesn't necessarily mean it's not failing.

Have you a spare you can swap out with to test? Maybe borrow one from a friend? Alternatively, you could bring to local repair store and get them to swap out PSU and test. might costs 20€£$ or so.

 

[Nickamus]

Hey there,

Everything you've described points to an issue with the PSU. Random restarts (typically at load) are telling. Your PSU is rather good, but that doesn't necessarily mean it's not failing.

Have you a spare you can swap out with to test? Maybe borrow one from a friend? Alternatively, you could bring to local repair store and get them to swap out PSU and test. might costs 20€£$ or so.

I completely forgot about this. I initially thought it was this originally back in September, but everything was fine after taking out a piece of Lego to prevent my GPU to droop. My mate has a 750W old PSU I could use to test, but not sure how old and safe it is.

Thank you for brining this up as I would've completely ruled it out

Could it be Currys PC World has given a faulty case and PSU when I first bought it from them? xD

 

[Roland Of Gilead]

I completely forgot about this. I initially thought it was this originally back in September, but everything was fine after taking out a piece of Lego to prevent my GPU to droop. My mate has a 750W old PSU I could use to test, but not sure how old and safe it is.

Thank you for brining this up as I would've completely ruled it out

Could it be Currys PC World has given a faulty case and PSU when I first bought it from them? xD

Well, I wouldn't think they deliberately give you a bogey one. There is always a failure rate with PSU's. Typically it's very low percentage. But yes, sometimes they can be faulty right out the gate.

The 750w replacement would be okay for the purpose of testing. If that works without issues, then you know the PSU needs replacing.

I'm presuming this was a pre-built from PC world? If it's within the year guarantee they provide, I'd go that route first and see if they can give you a new PSU. If not, then you could try RMA'ing the PSU to Corsair.

 

[Nickamus]

Well, I wouldn't think they deliberately give you a bogey one. There is always a failure rate with PSU's. Typically it's very low percentage. But yes, sometimes they can be faulty right out the gate.

The 750w replacement would be okay for the purpose of testing. If that works without issues, then you know the PSU needs replacing.

I'm presuming this was a pre-built from PC world? If it's within the year guarantee they provide, I'd go that route first and see if they can give you a new PSU. If not, then you could try RMA'ing the PSU to Corsair.

Unfortunately, the 750W PSU is a no go - he stopped using it because of a burning smell. I built the system myself. Parts ordered from various retailers.

Is there any other tests that I can do to narrow it down to the PSU? Last time, I used OCCT to monitor the voltage levels and were all correct - but I may be using it wrong.

Alternatively, I could buy a cheap 600W (unless I can go lower) for purely testing purposes. If it is the PSU, either contact Currys PC World (didn't get a good customer service with my PC case) or Corsair.

 

[DSzymborski]

Unfortunately, the 750W PSU is a no go - he stopped using it because of a burning smell. I built the system myself. Parts ordered from various retailers.

Is there any other tests that I can do to narrow it down to the PSU? Last time, I used OCCT to monitor the voltage levels and were all correct - but I may be using it wrong.

Alternatively, I could buy a cheap 600W (unless I can go lower) for purely testing purposes. If it is the PSU, either contact Currys PC World (didn't get a good customer service with my PC case) or Corsair.

Places like Amazon don't usually have issues of returns for PSU. You can always purchase something, see if it meets your needs by fixing the problem, and returning it if it doesn't.

 

[Nickamus]

Places like Amazon don't usually have issues of returns for PSU. You can always purchase something, see if it meets your needs by fixing the problem, and returning it if it doesn't.

That was my plan. I'm thinking about trying a the £47 EVGA PSU 600W W2 80+ [White Label] but read that it's a very bad PSU xD But if it can keep up with my games while the corsair gets RMA'd, I'd be happy with that. Not exactly modular, but it will do.

 

[Roland Of Gilead]

That was my plan. I'm thinking about trying a the £47 EVGA PSU 600W W2 80+ [White Label] but read that it's a very bad PSU xD But if it can keep up with my games while the corsair gets RMA'd, I'd be happy with that. Not exactly modular, but it will do.

Alternatively, I could buy a cheap 600W (unless I can go lower) for purely testing purposes. If it is the PSU, either contact Currys PC World (didn't get a good customer service with my PC case) or Corsair.

You could get a cheap one just to test, but then you're already half way to buying a decent Gold Unit. That EVGA isn't the best. I wouldn't put it in with your hardware.

Maybe something like a Corsair CXM 650. But if you can snag a Corsair RMx/RMi/TXM, EVGA G2/G3, or Seasonic Focus Plus Gold, they would all be quality units, that will last from one build to another.

Edit: Something like this:
PCPartPicker Part List

Power Supply: Corsair TXM Gold 650 W 80+ Gold Certified Semi-modular ATX Power Supply (£79.98 @ Currys PC World Business)
Total: £79.98
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2020-12-17 18:26 GMT+0000

 

[Roland Of Gilead]

Ooops. Sorry. Just realised, I've recommended the same PSU you already have!! haha 🙈 DOH!!!

But, I think you get what I mean. At £47 for a relatively crappy unit, only an additional 40 or so will get you a proper unit. One you can have piece of mind with.

 

[Nickamus]

You could get a cheap one just to test, but then you're already half way to buying a decent Gold Unit. That EVGA isn't the best. I wouldn't put it in with your hardware.

Maybe something like a Corsair CXM 650. But if you can snag a Corsair RMx/RMi/TXM, EVGA G2/G3, or Seasonic Focus Plus Gold, they would all be quality units, that will last from one build to another.

Edit: Something like this:
PCPartPicker Part List

Power Supply: Corsair TXM Gold 650 W 80+ Gold Certified Semi-modular ATX Power Supply (£79.98 @ Currys PC World Business)
Total: £79.98
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2020-12-17 18:26 GMT+0000

Ooops. Sorry. Just realised, I've recommended the same PSU you already have!! haha 🙈 DOH!!!

But, I think you get what I mean. At £47 for a relatively crappy unit, only an additional 40 or so will get you a proper unit. One you can have piece of mind with.

Haha xD No worries 😛

I will see if Currys PC World will do anything about it first before I buy a second PSU. If they let me have a temporary one, that would be great

CyberPunk 2077 running for three hours using "Test 1" and no crashes yet. Probably let the game run for longer and see.
Currently:

• Temperatures:
  • GPU: Constant 77°C and 97°C for the junction
  • CPU: Constant 63°C with random peaks at 72°C (one part of the water that is circulating is hot - since it is a pattern when it happens)
  • RAM: Between 54 to 56°C
• Utilisation:
  • GPU: 98% to 100%
  • CPU: 35 to 40%
  • RAM: 9GB
• Power:
  • GPU: 185 to 195W - might have seen it tick to 200W
  • CPU:
    • +12V: Approximately 11.9V [under]
    • +5V: Approximately 4.9V [under]
    • +3.3V: Approximately 3.31V [over]
    • 4.5GHz @ less than 1.5V

 

[Nickamus]

After 5 hours of running, I get a crash to desktop but no forced restart! However, after reading AMD's support, PSU not getting enough power is a possibility.

I'll run Assassin's Creed Valhalla or Godfall overnight with the RAM @ 16-18-18-36 with 1.35V and see if the same error shows. If it does, I will roll back to a "Recommended" driver and not use the latest "Optional" version via DDU and test again.

 

[Nickamus]

Reinstalled the drivers and played AC Valhalla for a few hours and let it go idle for a few more. No issues, but the power drawn by the card did not reach 190 to 200 compared to Cyberpunk. Will be getting a 750W PSU tomorrow to try. I'll run Godfall for a few hours tomorrow/overnight to see if it is the same as AC Valhalla

 

[Nickamus]

Surprisingly Currys PC World gave me a refund for my 650W PSU, do decided to keep this 750W instead.

First day, no issues at all with Cyberpunk 2077.

Five days later, unfortunately even with a 750W PSU, computer restarts itself. However, I only see it happen with Cyberpunk 2077. Unless, I've received yet another faulty PSU.

Does anyone have any other ideas?

Edit: Finished a MemTest86 and zero errors in all tests.

 

[Nickamus]

No computer restarts since I reinstalled my GPU driver to the "Recommended" one.

However, CP2077 still crashes causing a new issue for the GPU [screenshot]. I can only fix this by restarting the computer.

 

[Nickamus]

Hope everyone had a good Christmas!

Unfortunately, 3 days later, the computer restarts have happened again.

Anyone got any ideas?

 

[Nickamus]

I was too focused on the "Critical" section on Event Viewer and didn't look at the "Error" section. This section reports "WHEA Logger Event ID 18" - which corresponds to all the time stamp when the PC restarts itself.

I've tried a new motherboard and generates the same result... So close in giving up, but I'm also so close in finding the issue.

My assumptions now are:

1. My CPU and GPU are not compatible. My reasoning for this is based off of this forum post [link]
2. Faulty CPU - the reason I say that is because my mate is running Cyberpunk 2077 for now 8 hours and no issues with my GPU -temperatures seen at 60°C to 70°C and 90°C for junction [GPU]
3. I've read recently that the power sockets house that could cause the issue. Is this true? Can't test it without my GPU.

 

[Nickamus]

I am not sure why I can't edit the main post but this is the latest update:

Tests I've done

• Test 1: Turned off fast boot
• Test 2:
  • RAM back at base 2133MHz and CPU controlled by ASUS ["Default" and not "Optimised"]
  • For the GPU, on the “Radeon Software”, I left it at “Automatic”
• Test 3: Same as Test 2 but for the GPU, I left it at “Undervolt”
• Test 4: Same as Test 2 but for the GPU, I left it at “Overclock”
• Test 5: Resat GPU and RAM, making sure all cables are plugged in and not loose, reseating and applying new thermal for the CPU
• Test 6: Tried the original PSU cables (I bought premium Corsair cables to rule them out)
• Test 7: Tried onboard GPU
• Test 8: Unplugged all unnecessary devices (only the keyboard & mouse connected)
• Test 9: Tried each RAM individually and different sockets individually
• Test 10: Reinstalled drivers and using "Recommended" and not "Optional"
• Test 11: Removed my Lego stand to stop the GPU drooping down
• Test 12: Used two dedicated 6+2 PCIe cables instead of one splitting into two 6+2 cable ends
• Test 13: Tried setting RAM at 3200MHz @ 16-18-18-36 1.2V
• Test 14: Tried MemTest86
• Test 15: Tried other games and see if the issue persists (when RAM is back at 3200MHz @ 16-18-18-36 1.35V)
• Test 16: Ran OCCT overnight
• Test 17: Removed an adapter for the GPU
• Test 18: Swapped GPU with my mate (he gave me his GeForce GTX 1660 Ventus XS 6G OC) to see if I get the same issue.
• Test 19: Added my GPU back to my system after I did a DDU with the nVidia card and a reinstall of Windows (again). Had to use diskpart because my third reinstall of Windows some how locked it from being reinstalled.

Results

• Tests 1 to 6, 8, 9, 11 to 14, 17: Still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue
• Test 7: Doesn't even output a signal - maybe it is a sign of a faulty motherboard?
• Test 10: Works fine, then after the next shutdown, still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue
• Test 15: Other games don't cause the same issue. Although, it does not go over 189W when Cyberpunk 2077 does
• Test 16: No errors running it. Running Cyberpunk 2077 Works still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue
• Test 17: Still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue.
• Test 18: No crashes for 3 days straight with 12 (day 1), 6 (day 2), and 4 (day 3) hours in each day with CP2077. Everything seems to work fine with a different card. My mate had zero issues with my card when he tested on their system (Intel CPU)

In progress

• Test 19: Yesterday, I left the CP2077 in a busy spot for maximum power usage of my 5700XT and no crashes (longest it has ran for - 8 hours!). Going to try it a few more days to see if the issue is gone.

Are there any other tests I should do?

 

[Flayed]

If it is only crashing in CP 2077 then perhaps the issue is with the game?

 

[ArtMorte]

Are you sure your GPU or CPU aren't overheating? That can sometimes trigger a shutdown / restart. Maybe monitor the temps when under load.

 

[Nickamus]

If it is only crashing in CP 2077 then perhaps the issue is with the game?

Possibly. It happens so randomly I can't recreate the problem. I'll try another game when I get my replacement computer case - two screws snapped off at the end of my side panel... this will be my third replacement... maybe this case is bad for me.

Are you sure your GPU or CPU aren't overheating? That can sometimes trigger a shutdown / restart. Maybe monitor the temps when under load.

CPU never reaches over 71/72°C (unless iCUE is faulty) and GPU max is 80°C - which is normal for the specific model I bought.

For visual using iCUE, I make 75°C red, 60°C yellow and 50°C green. My lights never turned red once, orange for a few seconds out of a few minutes, but never red.

OCCT tells the same story (when I exit iCUE and Radeon to prevent clashing). I've used in-game monitoring (from AC Valhalla i believe it was this game) and get the same results (no crashing unless I start CP2077.

RivaTuner tells the same information and so does CPUTemp.

 

[Nickamus]

I am not sure why I can't edit the main post but this is the latest update:

Tests I've done

• Test 1: Turned off fast boot
• Test 2:
  • RAM back at base 2133MHz and CPU controlled by ASUS ["Default" and not "Optimised"]
  • For the GPU, on the “Radeon Software”, I left it at “Automatic”
• Test 3: Same as Test 2 but for the GPU, I left it at “Undervolt”
• Test 4: Same as Test 2 but for the GPU, I left it at “Overclock”
• Test 5: Resat GPU and RAM, making sure all cables are plugged in and not loose, reseating and applying new thermal for the CPU
• Test 6: Tried the original PSU cables (I bought premium Corsair cables to rule them out)
• Test 7: Tried onboard GPU
• Test 8: Unplugged all unnecessary devices (only the keyboard & mouse connected)
• Test 9: Tried each RAM individually and different sockets individually
• Test 10: Reinstalled drivers and using "Recommended" and not "Optional"
• Test 11: Removed my Lego stand to stop the GPU drooping down
• Test 12: Used two dedicated 6+2 PCIe cables instead of one splitting into two 6+2 cable ends
• Test 13: Tried setting RAM at 3200MHz @ 16-18-18-36 1.2V
• Test 14: Tried MemTest86
• Test 15: Tried other games and see if the issue persists (when RAM is back at 3200MHz @ 16-18-18-36 1.35V)
• Test 16: Ran OCCT overnight
• Test 17: Removed an adapter for the GPU
• Test 18: Swapped GPU with my mate (he gave me his GeForce GTX 1660 Ventus XS 6G OC) to see if I get the same issue.
• Test 19: Added my GPU back to my system after I did a DDU with the nVidia card and a reinstall of Windows (again). Had to use diskpart because my third reinstall of Windows some how locked it from being reinstalled.

Results

• Tests 1 to 6, 8, 9, 11 to 14, 17: Still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue
• Test 7: Doesn't even output a signal - maybe it is a sign of a faulty motherboard?
• Test 10: Works fine, then after the next shutdown, still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue
• Test 15: Other games don't cause the same issue. Although, it does not go over 189W when Cyberpunk 2077 does
• Test 16: No errors running it. Running Cyberpunk 2077 Works still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue
• Test 17: Still crashes with Windows Event Viewer showing the Kernel-Power [event ID 41] issue.
• Test 18: No crashes for 3 days straight with 12 (day 1), 6 (day 2), and 4 (day 3) hours in each day with CP2077. Everything seems to work fine with a different card. My mate had zero issues with my card when he tested on their system (Intel CPU)

In progress

• Test 19: Yesterday, I left the CP2077 in a busy spot for maximum power usage of my 5700XT and no crashes (longest it has ran for - 8 hours!). Going to try it a few more days to see if the issue is gone.

Are there any other tests I should do?

Updates to the above:
Tests I've done

• Test 20: Recently, I've seen Cinebench being mentioned. Going to test both stress and stability test for both multicore and single core
• Test 21: Try Ubuntu to rule out Windows doing something in the background

Results

• Test 19: Crashes came back the next day. No issues with other games - so far
• Test 20: No issues running it
• Test 21: Couldn't run the high intensive games I wanted so uninstalled it (badly uninstalled so had to use diskpart to clean it out).

Current situation
After reinstalling from scratch and everything updated to the recent drivers, both CP2077 and Godfall has crashed to desktop and not random PC restart - using the new beta BIOS.

Conclusions so far (based on the above)

• Compatibility between the specific models of PowerColor's Red Dragon 5700 XT and the AMD Ryzen 7 3800X
• CP2077 is just badly optimised for my specific system since it has been the only one game that causes my computer to restart itself (no issues from September 2020 to the release of CP2077.

Think I've ruled out everything else to be honest.