Skip to content
Advertisement

Kernel update breaks CUDA

I have a NVIDIA Grid K2 GPU allocated to a virtual server running Ubuntu 14.04. To reinstall the proper drivers after an automatic kernel update I ran sudo apt-get update followed by sudo apt-get install nvidia-current.

Now I cannot get CUDA 7.5 to work any longer. If I run the deviceQuery sample I get the following message:

CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 35
-> CUDA driver version is insufficient for CUDA runtime version
Result = FAIL

This is the output from sudo lshw -c video

PCI (sysfs)

  *-display
       description: VGA compatible controller
       product: SVGA II Adapter
       vendor: VMware
       physical id: f
       bus info: pci@0000:00:0f.0
       version: 00
       width: 32 bits
       clock: 33MHz
       capabilities: vga_controller bus_master cap_list rom
       configuration: driver=vmwgfx latency=64
       resources: irq:16 ioport:1070(size=16) memory:ec000000-efffffff memory:fe000000-fe7fffff memory:c0300000-c0307fff
  *-display
       description: VGA compatible controller
       product: GK104GL [GRID K2]
       vendor: NVIDIA Corporation
       physical id: 0
       bus info: pci@0000:0b:00.0
       version: a1
       width: 64 bits
       clock: 33MHz
       capabilities: pm msi pciexpress vga_controller bus_master cap_list
       configuration: driver=nvidia latency=64
       resources: irq:19 memory:fc000000-fcffffff memory:e0000000-e7ffffff memory:e8000000-e9ffffff ioport:5000(size=128)

Advertisement

Answer

I solved this issue with the following steps using the hints from the installation guide:

  1. Uninstalled the packages that I had mistakenly installed by running sudo apt-get --purge remove nvidia-current.

  2. Uninstalled CUDA 7.5 with the command sudo /usr/local/cuda-7.5/bin/uninstall_cuda_7.5.pl

  3. Restarted the server using sudo reboot

  4. Installed CUDA 7.5 by running the downloadable .run file and following the instructions.

  5. Checking that everything works by running the deviceQuery CUDA sample.

User contributions licensed under: CC BY-SA
2 People found this is helpful
Advertisement