F@H GTX 1650 BAD WORK UNIT on Fedora

[Alex Atkin UK]

Anyone else running a GTX 1650 on Linux and having issues?

 

Its been running fine for months but suddenly I'm getting a lot of BAD WORK UNIT jobs and it will basically get disabled after running for a while.

 

The strange thing is, if I reboot it will trigger a few then work fine for hours.  I had it running for 6 hours last night, paused it when I got up, then when I tried to resume it immediately triggered a BAD WORK UNIT.  Something seems very off with the GPU itself or the drivers.

 

What's worrying is at first it only happened after leaving the PC up for a few days, but now its happening the same day with the last reboot being the worst yet (the first time I saw BAD WUs immediately after a clean boot).

Is there a particular benchmark or something that might identify if the card is going bad?  Though I get no graphics corruption, freezes or anything obvious, its just F@H that is failing.  Though I DO get Firefox crashing randomly.  Desktop compositing is disabled as I've always had stability problems with that on KDE Plasma no matter what GPU is used.

[Gorgon]

5 hours ago, Alex Atkin UK said:

Anyone else running a GTX 1650 on Linux and having issues?

 

Its been running fine for months but suddenly I'm getting a lot of BAD WORK UNIT jobs and it will basically get disabled after running for a while.

 

The strange thing is, if I reboot it will trigger a few then work fine for hours.  I had it running for 6 hours last night, paused it when I got up, then when I tried to resume it immediately triggered a BAD WORK UNIT.  Something seems very off with the GPU itself or the drivers.

 

What's worrying is at first it only happened after leaving the PC up for a few days, but now its happening the same day with the last reboot being the worst yet (the first time I saw BAD WUs immediately after a clean boot).

Is there a particular benchmark or something that might identify if the card is going bad?  Though I get no graphics corruption, freezes or anything obvious, its just F@H that is failing.  Though I DO get Firefox crashing randomly.  Desktop compositing is disabled as I've always had stability problems with that on KDE Plasma no matter what GPU is used.

Alex,

 

I've had this happen on some of my Rigs (Ubuntu 18.04 LTS). I can't confirm, but I suspect what happened was the Nvidia support packages got updated and were out-of-sync with driver package (kernel module). I'm running driver 510.54 currently.

 

I eventually figured this out after attempting to run nvidia-smi and getting an error.

 

I have noticed that when I got the "Failed" slot message I often had to delete the slot, reboot, then add it again to clear it.

 

I'd suggest deleting the GPU slot; updating your drivers, rebooting & add the slot back.

[Alex Atkin UK]

So after the reboot:

Quote

23:09:22:WU00:FS01:0x22:  Configuring platform CUDA
23:09:22:WU00:FS01:0x22:Failed to create CUDA context:
23:09:22:WU00:FS01:0x22:The requested CUDA device could not be loaded
23:09:22:WU00:FS01:0x22:Attempting to create OpenCL context:
23:09:22:WU00:FS01:0x22:  Configuring platform OpenCL
23:09:24:WU00:FS01:0x22:Failed to create OpenCL context:
23:09:24:WU00:FS01:0x22:Error uploading array atomIndexDevice: clEnqueueWriteBuffer (-4)
23:09:24:WU00:FS01:0x22:ERROR:125: Failed to create a GPU-enabled OpenMM Context.

Quote

23:10:42:WU00:FS01:0x22:Attempting to create CUDA context:
23:10:42:WU00:FS01:0x22:  Configuring platform CUDA
23:10:43:WU00:FS01:0x22:Failed to create CUDA context:
23:10:43:WU00:FS01:0x22:The requested CUDA device could not be loaded
23:10:43:WU00:FS01:0x22:Attempting to create OpenCL context:
23:10:43:WU00:FS01:0x22:  Configuring platform OpenCL
23:10:43:WU00:FS01:0x22:Failed to create OpenCL context:
23:10:43:WU00:FS01:0x22:Error initializing context: clCreateContext (-6)
23:10:43:WU00:FS01:0x22:ERROR:125: Failed to create a GPU-enabled OpenMM Context.

Then suddenly, one of the WUs worked and it continued to work until I paused.

 

Resuming later:

Quote

11:50:20:WU01:FS01:0x22:Failed to create CUDA context:
11:50:20:WU01:FS01:0x22:Error creating array posq: CUDA_ERROR_OUT_OF_MEMORY (2)

Its a bit strange as the log says 5GB of system RAM was available and I'm fairly sure at least 2GB of VRAM would have been.

 

[Gorgon]

8 hours ago, Alex Atkin UK said:

23:09:22:WU00:FS01:0x22:The requested CUDA device could not be loaded

This suggests you might have a permissions issue with the device