Update: 27/04/2024:
Turns out I didn't jiggle the right part of the cursed 4x8-pin to 16-pin adapter cable. So, the issue was connectivity problems on the 8-pin side, apparently one of the 8-pins wasn't fully inserted (ironic, isn't it?) and caused the problem. I jiggled the 8-pin side of the adapter, and poof, black screen. I can't confirm 100% this fixed the issue but I've resat all 4 connectors and cable-tied them to the 24-pin ATX power cable for stability so the adapter isn't having to act as a weight bearing support, too. I will say, though, this connector is absurd and extremely difficult to work with, given the power socket on the GPU faces away from the source of the cables, meaning you need to bend / stress the adapter cables regardless of anything.
I'll update this post if it happens again.
Original post:
And here we go again, with a persistent problem with NVIDIA's unstable, potato "Studio" drivers, while I'm trying to get shit done.
The description for Event ID 14 from source nvlddmkm cannot be found. Either the component that raises this event is not installed on your local computer or the installation is corrupted. You can install or repair the component on the local computer.
If the event originated on another computer, the display information had to be saved with the event.
The following information was included with the event:
\Device\000000a9
CMDre 00000001 00003ffc ffffffff 00000007 00ffffff
The message resource is present but the message was not found in the message table
Googling reveals it's a known issue with some of their drivers, going back from 2022 to 2016 (nice).
Also of note: the "CMDre 00000001 00003ffc ffffffff 00000007 00ffffff" part is repeated in the dozens of "Event ID 14" error entries, but the "00000001" goes up each time, to "00000002" and so on, but sometimes makes larger jumps, like from 00000008 to 00000021.
The issue is essentially, a black screen often with a system lockup and then a BSOD (Watchdog Timeout) or it will literally recover, minus the RTX 4090 in Task Manager (Windows Yeeted that shit out the window, apparently). The displays are still reporting no signal, but I use Remote Desktop from my phone to remote-in and shows the PC is still fully operational, minus the GPU. Nice. Rebooting fixes it entirely.
For the sake of records and googling:
I have tried reseating the GPU and cleaning the slot/finger (PCI-E connectivity issues)
Lowering all GPU overclocks to stock
closing random 3D software in the background
DDU and reinstall the latest drivers
Reseating the power cable
Jiggling the power cable while its under load (didn't cause it)
Jiggling the GPU while its under load (didn't cause it)
None of the above fixed it.
Things that are consistent with my machine:
I have multiple monitors, sometimes they are turned off, but always connected (2 displays usually).
I work with Blender (OpenGL preview and OptiX with Cycles) and premiere Pro every day.
I have multiple instances of blender open at any time.
I leave my PC on for >24 hours at a time, it can be under load for most of that.
My CPU and memory are not overclocked (JEDEC RAM)
860W Platinum PSU.
RTX 4090 with NVIDIA-provided 4x 8-pin to 1x16pin adapter, all populated with 3 cables to the PSU, 1 is shared.
Happens on Driver version 552.22 Studio, (And one before, can't remember version).
Notes on symptoms:
Doesn't seem to happen under load, even when the GPU is stressing at 450W in premiere pro. Happened at idle, usually after an intense load, or just sat idling on the desktop. Happened immediately when I opened another blender instance (the moment the interface loaded, black screen, GPU gone from Task Manager).
Some previous errors have indicated random applications have errored with "Out of memory" exceptions, and 3D-accelerated windows processes like dwm.exe (desktop window manager) have errored out and restarted multiple times before the TDR event in the Event Viewer.
Current Diagnosis:
driver issue.
Current prognosis:
gonna happen again -.-
current suggested remedial action:
FIX YOUR SHIT NVIDIA