Windows Nvidia Driver Issue - Solution
So I utilize
Windows 10 for gaming from time-to-time and for some software testing, but one main issue I've had (since resolved) has been Nvidia Drivers working correctly. In this I'll guide you through all the possible fixes and what ultimately worked for me. Not everyone will have the same issue so different resolutions are needed.
Please note that this guide is intended for Windows 10 users with an NVIDIA GPU.
The driver used for this fix is an RTX 3080 TI from Gigabyte (do not recommend).
The Main issue I would have that was consitant is that any newer game would crash after just 5-10 minutes of gameplay, every time. This only happened in modern games (2018+) with high resolution textures and effects (a la Battlefield 2042, CoD Warzone + MW2019, and Cyberpunk).
There are a variety of
nvlddmkm errors out there, so any of them could be related to the below errors.
Fatal Rrror: Direct3D driver returned error code (DXGI_ERROR_DEVICE_REMOVED) while Presenting. Reason ....
DirectX function FindClosestMatchingMode failed with DXGI_ERROR_DEVICE_REMOVED (The video card has been physically removed from the system, or a drive upgrade for the video card has occurred.)
GPU Hang Detected
Directx function "GPU HANG DETECTED" Failed with DXGI_ERROR_DEVICE_HUNG (the GPU will not respond to more commands) GPU 3080TI DRIVER 49709.
Some prerequisites before testing the other methods.
Updating Motherboard BIOS
Flashing a new BIOS could resolve issues regarding the GPU or other board parts not being recognized or not accessed properly.
Using System File Checker
sfc command in PowerShell/Cmd to check corrupted files and fix them. Windows Suppport Article.
DISM + SFC
Alternatively, you can use
DISM and then run
sfc to do a more thourough job of cleaning things up. Microsoft Docs for DISM.
Check if your Windows updates are up to date as I believe Microsoft distributes Intel driver updates within Windows.
Using Crystal Disk Info to check the SMART health of your drive(s). It may be that your drive is having read errors or it's near the end of it's lifespan.
Graphics Drivers Fix
First thing that would be the wrong with the install would be drivers if going through the Prerequisites didn't resolve anything. Driver issues account for most problems.
DDU and Uninstalling old drivers
Using the DDU tool, we can go in and uninstall the display drivers completely and start from scratch. If you've never used DDU before then here's a great tutorial from the website.
Now that the drivers are completly uninstalled let's install the latest NVIDIA GPU Drivers. I wouldn't recommend using the ones from the manufacturer's site, rather use the ones through NVIDIA's driver selection site ~~or GeForce Experience application~~.
After running DDU and reinstalling from Nvidia, make sure to check Windows updates to ensure that other graphics drivers are installed if needed.
ansel in your nvidia setting might solve the issue on some older cards and as a bonus can help with the performance of the card.
Disable Nvidia ShadowPlay and Overlay
If these are enabled they may cause issues with certain games. I initially was able to have the issue resolved but it came back shortly after. It seems it was just the performance that was affected rather than any driver fix.
If you're running the GPU is set to a silent mode you'll probably get some pretty high heat on your system, so I'd recommend using a custom fan curve in something like MSI Afterburner and adding more cooling to the system.
Voltage / Clock Speeds
Your card, despite it being marked as OC or not, may be factory overclocked. This is usually so the card, under ideal circumstances, can reach it's target performance metrics. With that said, it may be running unstable and causing your game to crash, and for this I'd recommend undervolting/underclocking it slightly to get a performance boost. A good brief guide on the underclocking and [how to do so in the MSI Afterburner Software](https://msiafterburnerguide.com/how-to-undervolt-gpu-with-msi-afterburner/.
If it's not something with your GPU itself and the above did nothing to help then perhaps checking if your memory is working would be a good move. At this step you should have already checked your BIOS version and ensured that you're running the latest version of whatever you need.
Testing Memory / Software
Using [MemTest86] should be able to run all the tests needed to know if your memory is working correctly. Please note, you don't need to pay for this software, just download the free version.
Testing Memory / Physical
(This partially resolved my issue, at least initially)
Even if you're not getting an error in the memory test it may be useful to move around the RAM to different slots and give the machine a reboot. Not sure why this would have an effect but I was testing with
Some thing you can check in the BIOS to resolve issues depending on hardware
Disabling XMP and any overclocks to your RAM might be effective. I have RAM that isn't compatible with XMP so I have it disabled. If you have facy RGB RAM with heatsinks then probably leave this setting to enabled or auto.
(This resolved my issue)
Disabling Fast Boot seemed to resolve my issue, which I assume is due to the RAM not initializing correctly on boot. I'm not 100% why, but this was the ticket for me
I've found that some people online that disabling this helped resolve their issue, perhaps on older machines or GPUs this could be the case.
TPM and Secure Boot
I don't believe these have anything to do with
CPU Clock Speeds
I didn't get any resolution with reducing or increasing the clock speeds, however I've seen from some that setting it to maximum performance or "gaming" mode will resolve other issues on 12th gen CPUs on newer games.
So I was able to fix my issue and get great results by; moving around the Memory Sticks, Disabling XMP, Disabling Fast Boot, using DDU, using DISM, fixing the fan curve though Afterburner, and slightly underclocking. It was disabling Fast Boot that did it though, and after that I am getting stable performance without any crashes. I think the underlying issue comes from me using DDR5 Memory and it not being fully supported or stable.
I will try and create a wiki article for this for noting down DirectX errors and such.