Graphics card crashes after a couple minutes.

Tarman · 23 Jun 2023 at 10:19

Hello,

I've been having a problem running games on a new PC build..
I'll launch a game (e.g. Starship Troopers) and will be able to play for a couple of minutes; then the game will crash. The system will either return to desktop, or restart.

As starship troopers is running Unreal Engine, the error thrown from there is a "Device was removed" error.
In windows event viewer, there are one of two errors that are logged at the time of each crash.

NVIDIA OpenGL Driver:
DrvSetContext failed functionality indeterminant
(pid=14656 cncmd.exe 64bit)

And

NVIDIA OpenGL Driver:
Ran out of memory

I initially thought it could be a heat issue, with the card shutting itself off, but the GFX card doesn't get above 40c.

Rolled the dice on a B Grade graphics card; but perhaps this is the problem?

Any suggestions of things to try would be appreciated. I have time today to test anything suggested.

edit: Also tried turning the games graphics way down, but still experiencing the same problem.
After the game freezes, it will hang for a good minute, then all the colour will dissapear from the image, leaving it looking like so:

This will remain on screen for another 10-20 seconds, then the PC will either crash to desktop, or restart.

mickyflinn · 23 Jun 2023 at 10:58

Did you totally uninstall your old drivers before reinstalling the new ones ?

Unreal engine updated ?

System spec please ?

Quartz · 23 Jun 2023 at 11:10

This is your 4090, right? Which PSU do you have?

I see you have a 7950X3D: are you running the utility that shuts down the CCD that doesn't have the Vcache?

Tarman · 23 Jun 2023 at 11:16

Quartz said:
This is your 4090, right? Which PSU do you have?

I see you have a 7950X3D: are you running the utility that shuts down the CCD that doesn't have the Vcache?

PSU is Corsair RMx Series RM850x 80 PLUS Gold Fully Modular ATX Power Supply (CP-9020200-UK)

I was running a bunch of utilities after I installed Windows 11. But eventually uninstalled everything, in-case one of the utilities was causing the problem. but alas, no change.

Tarman · 23 Jun 2023 at 11:28

mickyflinn said:
Did you totally uninstall your old drivers before reinstalling the new ones ?

Unreal engine updated ?

System spec please ?

It was a brand new Windows install.. so never had any drivers on it bar the default ones Windows uses. Regardless, I hit clean install when installing the NVIDIA drivers.

System spec in the signature.

Quartz · 23 Jun 2023 at 11:43

Tarman said:
PSU is Corsair RMx Series RM850x 80 PLUS Gold Fully Modular ATX Power Supply (CP-9020200-UK)

That's a decent one and should have enough oomph for the 4090. I will note that I have a Zotac Trinity 4090 but have a 1000W PSU.

Tarman said:
I was running a bunch of utilities after I installed Windows 11.

Are you still running Ryzen Master? And check out this article:

AMD Ryzen 9 7950X3D Core Parking Problem & Solution - Page 3 of 6 - Hardware Busters

Hardware Busters - AMD Ryzen 9 7950X3D Core Parking Problem & Solution - CPU

hwbusters.com

mickyflinn · 23 Jun 2023 at 11:44

Tarman said:
It was a brand new Windows install.. so never had any drivers on it bar the default ones Windows uses. Regardless, I hit clean install when installing the NVIDIA drivers.

System spec in the signature.

From my understanding a few things can cause such errors, one is the game or open GL is requesting more vram than what's available so you need to check windows pagefiling so there's enough ram available. So if this is the case maybe lowering settings mite help and indicate this.

Another is to the disable pcie power management in the bios as the card is requesting more power than what's available. So could be the psu

Joxeon · 23 Jun 2023 at 11:46

Could well be down to unstable system memory.

Quartz · 23 Jun 2023 at 11:47

mickyflinn said:
so there's enough ram available.

Unlikely to be an issue on a 4090 with 24 GB VRAM

Quartz · 23 Jun 2023 at 11:49

mickyflinn said:
Another is to the disable pcie power management in the bios as the card is requesting more power than what's available. So could be the psu

OP, have you got all three 8 pin PCIe cables coming from separate ports on the PSU? Use of pigtails is not a good idea.

Tarman · 23 Jun 2023 at 12:00

Quartz said:
OP, have you got all three 8 pin PCIe cables coming from separate ports on the PSU? Use of pigtails is not a good idea.

Yeah I got 3 pcie power cables going into the card.

mickyflinn · 23 Jun 2023 at 12:02

Quartz said:
Unlikely to be an issue on a 4090 with 24 GB VRAM

It's not the 4090 its open GL requesting more ram than what's availablesort of a glitch it then request the data from system memory if not enough then reports and error shuting down the game.

That's my understanding of it but maybe not the best explanation.

Tarman · 23 Jun 2023 at 12:07

mickyflinn said:
It's not the 4090 its open GL requesting more ram than what's availablesort of a glitch it then request the data from system memory if not enough then reports and error shuting down the game.

That's my understanding of it but maybe not the best explanation.

This sums it up really. Certainly lines up with the errors being thrown. No idea what to do about it though.

When the game crashes, you can still hear people chatting away over their mics. So it's not a total system hang straight away. Very strange

mickyflinn · 23 Jun 2023 at 12:39

Tarman said:
This sums it up really. Certainly lines up with the errors being thrown. No idea what to do about it though.

When the game crashes, you can still hear people chatting away over their mics. So it's not a total system hang straight away. Very strange

Check windows page file hasnt allocated to much system memory, should be about 5 GB for 32 GB of memory but check guides.

There are also other possibilities such as pcie power management which when enabled windows wil limit power to the GPU we which can cause problems. Some have disabled this and it has solved it but I would then move to a larger psu.

Have you dropped the gane details and resolution to see what happens ?

Tarman · 23 Jun 2023 at 12:56

Page file currently managed by windows.. with 5gb in use.

Turned off pcie power management, restarted, and gave the game a test, but same problem.
Crashes seem to occur very frequently now. No longer getting the 2-3 minutes of gameplay anymore, just straight to crashing as soon as I try join the game server.

therealdippy · 23 Jun 2023 at 12:58

A Google search suggests that "cncmd.exe" is part of the AMD graphics driver.
But the NVIDIA OpenGL driver is also mentioned.
It is not good to have multiple graphics drivers installed on the same system.

mickyflinn · 23 Jun 2023 at 13:15

Tarman said:
Page file currently managed by windows.. with 5gb in use.

Turned off pcie power management, restarted, and gave the game a test, but same problem.
Crashes seem to occur very frequently now. No longer getting the 2-3 minutes of gameplay anymore, just straight to crashing as soon as I try join the game server.

Page file is fine and Renable pcie power management.

The fact that it's a b grade 4090 will always give concern until you can test it in another pc or just return it to be tested. If you got from Overclockers give them a call .

Install MSI afterburner and see what's going on with the GPU ,launch game until crash then check stats .

Another thing to try is load bios defaults and try the game again.

Keep us informed.

Tarman · 23 Jun 2023 at 13:29

Hmmm, something else weird going on with bios.
Making changes, pressing F10 or the save and exit button, the system hangs. No bios changes are applied.
Trying to flash the bios also hangs on a white screen. No changes are applied. System needs hard reboot.

What a nightmare this build has been!

Edit: also, reducing resolution amd gfx settings let's the game run no problem

Tetras · 23 Jun 2023 at 13:39

Tarman said:
Edit: also, reducing resolution amd gfx settings let's the game run no problem

Suggests a power problem then, or the card is just unstable/faulty.

How much do you need reduce them by?

Tarman · 23 Jun 2023 at 13:47

Tetras said:
Suggests a power problem then, or the card is just unstable/faulty.

How much do you need reduce them by?

Reduced resolution from 4k to 1080p.
Graphics from "Epic" to "Low".