Driver issues causing repeated crashes on NVIDIA hardware
Driver issues causing repeated crashes on NVIDIA hardware
I'm struggling with this situation. I didn't post here casually, but I've encountered similar issues since switching back to NVIDIA for the RTX 3090 in February. According to Event Viewer, the Event ID is 14. Typically, two errors occur simultaneously, such as:
Device\Video3 CMDre 00000001 00000240 ff1fe209 00000007 00000000
And another like this:
Device\Video3 0000(0000) 00000000 00000000
I've tried several fixes:
- Swapping the PSU
- Replacing the motherboard
- Changing RAM speed and voltage
- Setting performance preferences in NVIDIA settings
- Disabling PCIe power-down
- Updating drivers (versions 472.12, 47.39, 472.47)
- Running DDU in safe mode
- Lowering GPU clock speeds
- Adjusting power limits and disabling certain optimizations
I've also experimented with:
- Using different driver versions
- Running the system in safe mode
- Reducing monitor count
- Changing RAM type and speed
- Modifying BIOS settings
- Testing with older monitors or a different GPU
The problem remains inconsistent—sometimes it works for days, other times it crashes repeatedly even when the system is off. A game won't need to run, and sometimes windows applications freeze until they're restarted. Occasionally, the display driver recovers, but usually it doesn't. I've seen the machine behave normally for a short time before issues return.
I'm considering further steps like:
- Lowering monitor count
- Replacing the CPU or GPU entirely
- Running a full system reinstall
The pattern isn't clear, and it's becoming increasingly frustrating. If you have more details or want to discuss specific changes, let me know.
It seems I’m questioning whether I should be doing this. It’s basically a criticism of stock Windows—preferring to invest time and money over doing it, right? I think my monitors might be contributing too, since varying refresh rates can mess with the drivers. Also, the VG27AQL1A tends to freeze for a second or two after logging in for the first time each day, which others have reported too.
Alright, update time... The PC was left on overnight. I haven’t reinstalled Windows yet. Upon waking, it wasn’t responding and the POST attempt failed. I’m seeing error codes 97 and sometimes B2 on the motherboard screen, which the manual mentions as "Console Output devices connect" and "Legacy option ROM initialization." Notably, the PC would do this a few times with the 3090 on the Gigabyte board. That’s what led me to suspect I’d damaged that card and needed to replace it. Usually, leaving the machine off briefly helps it boot again. I’ve tried wiping the CMOS, moving the card to another slot just to test, starting the PC without PCIe power (which changed the code from 97 to B2 but kept the same code afterward), and even trying to reinstall the 3090 without water cooling, which got me past the POST screen though it said EFI wasn’t supported. So for a functional PC this weekend, I think I’ll put the 3090 back in. This raises more questions about what else might be affecting the build.
The Ti 3080 appears to be quite inactive. The 3090 is functioning properly once again, which means it works intermittently at times.
I’ve stopped using features like 4G decoding and resizable barcode scanner. I also executed a script from the NVIDIA Reddit community, which tested all DSIM and SFC commands to perform a comprehensive fix on the Windows system. Recently, I played a game that typically causes GPU driver issues in Star Citizen, but it ran smoothly this time. I reviewed Event Viewer and noticed several errors during playtime, though they didn’t trigger a display driver crash as before. [Screenshot of event viewer] The logs indicated an issue with Event ID 14 from the nvlddmkm service—no component was found matching that ID. It might mean the service isn’t installed locally or its installation is damaged. You can reinstall or repair it on your machine. If the problem came from another device, the associated display data must have been saved with the event. The message resource exists but isn’t present in the table. Still, if crashes persist, I’ll consider turning 4G decoding back on and resizing the barcode scanner to determine if the fix was successful.