My device is jetson nano 4GB. When I ran my code after compiling opencv, I was prompted that the cuda runtime version and driver version did not match. The methods I used online including using the GUI in ubuntu and so on didn’t work. What should I do to upgrade the cuda driver.
Normally one would use JetPack/SDK Manager. You can uncheck the flash part, and simply check the optional CUDA components. This is done with a fully booted Jetson, not in recovery mode. This would be a CUDA 10 release. If your software wants CUDA 11, then you are out of luck, this would require Xavier or newer (Orin). What release was the software looking for?
The software requires CUDA11.4. However, after I installed version 11.4, other packages such as onnxruntime are called without the error of running version and driver version inconsistency
CUDA 11 is not compatible with a Nano (unless it is Xavier or Orin Nano). Keep in mind that Jetsons do not have a PCI based GPU (a discrete GPU or dGPU). Jetsons have the GPU integrated directly to the memory controller (an iGPU). Most drivers out in the wild use nvidia-smi to find a GPU over the PCI bus, but this does not work on an iGPU. The version supplied with JetPack/SDKM is the iGPU one. The Nano reached end of life for new features quite some time back. Only Xavier and Orin use JetPack 5.x/L4T R35.x, and this is required for the iGPU driver. Sorry, there is nothing available for CUDA 11 for that hardware.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.