Jetson TX2 install CUDA7.0 err

nvidia@tegra-ubuntu:~ ls buildOpenCVTX2 Face-Recognition-master Public cuda-repo-l4t-r23.1-7-0-local_7.0-71_armhf.deb gpuGraphTX tegra_multimedia_api cuda-repo-ubuntu1404-7-0-local_7.0-71_amd64.deb gstreamer tegrastats Desktop Templates Documents Music test Downloads NVIDIA_CUDA-9.0_Samples Videos examples.desktop opencv VisionWorks-SFM-0.90-Samples Face-Recognition Pictures weston.ini nvidia@tegra-ubuntu:~ sudo dpkg -i cuda-repo-l4t-r23.1-7-0-local_7.0-71_armhf.deb
[sudo] password for nvidia:
dpkg: error processing archive cuda-repo-l4t-r23.1-7-0-local_7.0-71_armhf.deb (–install):
package architecture (armhf) does not match system (arm64)
Errors were encountered while processing:

It looks like you have both the x86_64 and the 32-bit ARM CUDA packages in your home directory, did you copy them there?
The correct architecture for TX2 is arm64 (aarch64), which JetPack will install for you. The current supported CUDA version in JetPack is CUDA9.

The most interesting part of this:


None of the official releases were R23.1. Sounds like it might be from a git repo download.

YES,JetPack have install the CUDA9 for me;I want to install the CUDA7 because I want to use the NVIDIA System Profiler to look the information about GPU;but I found NVIDIA System Profiler donot support CUDA9.

I just want to look the GPU information in TX2, “sudo ~/tegrastats” This command Not enough detail and humanization;

if you have any advise,Please contact me;
my e-mail is thanks;

I don’t know if what you want is available. I can tell you though that the “armhf” of “cuda-repo-l4t-r23.1-7-0-local_7.0-71_armhf.deb” is for 32-bit user space and is not directly arm64 compatible. Finding a CUDA 7 with arm64 would be the only way and I don’t know if this even exists.

thanks,I try to find the current CUDA version;but cannot find,maybe it not exist
at present, I only use “sudo ~/tegrastats” to see GPU info;

Try using NVVP (NVIDIA Visual Profiler), nvprof, or Tegra System Profiler which comes with JetPack.

The only CUDA version supported is what ships with JetPack, it is tightly coupled with the L4T BSP version.