After installing cuda when I am trying to run deviceQuery demo I am getting following error.
CUDA Device Query (Runtime API) version (CUDART static linking)
cudaGetDeviceCount returned 35
CUDA driver version is insufficient for CUDA runtime version
Result = FAIL
After trying to verify the Driver version using the following command
cat /proc/driver/nvidia/version
cat: /proc/driver/nvidia/version: No such file or directory exists.
On the current Jetson system software, there’s no /proc/driver/nvidia/* like there is on x86…
To install the drivers, you login is user “ubuntu”, then:
cd /home/ubuntu/NVIDIA-INSTALLER/
cat README.txt
Follow the directions in that README file and you should be good to go after the reboot.
A note: the L4T “after the flash” version is Ubuntu 14.04 LTS. It does not have any such file or directory /home/ubuntu/NVIDIA-INSTALLER and no such name as NVIDIA-INSTALLER can be found by locate NVIDIA after 'updatedb.
One hopefully presumes that all of the drivers have been installed to the L4T 19.2 (Ubuntu 14.04 LTS) “sample file system” and “Tegra Drivers” downloads from the support page.
I have the same issue. I followed the Cuda getting started guide especially the section 3.7.
I get this error when I run the examples:
CUDA driver version is insufficient for CUDA runtime version
Could this be because I didn’t flash the tk1 but used the shipped OS? Is it worth burning the TK1 again?
Yes if you are having a CUDA driver vs runtime compatibility error then you are probably using one of the early releases. Flash with the latest version (currently L4T 19.2) and then follow the CUDA install guide again (a compact version is at [url]http://elinux.org/Jetson/Installing_CUDA[/url]).
So I got my Jetson Tk1 last week and ran into an issue where the default shipped OS and drivers would not work with the latest Cuda samples or Cuda toolkit, both 6.5 respectively.
So I flashed the Jetson TK1 to the latest available release from my x86 Ubuntu machine and proceeded with the install procedure as located here Jetson/Installing CUDA - eLinux.org. Installed the latest Cuda samples and Cuda toolkit, to be met with the same error.
1_Utilities/deviceQuery/deviceQuery Starting…
CUDA Device Query (Runtime API) version (CUDART static linking)
cudaGetDeviceCount returned 35
→ CUDA driver version is insufficient for CUDA runtime version
Result = FAIL