Reprint please specify the source:
Http://www.cnblogs.com/darkknightzh/p/5655957.html
Reference URL:
https://devtalk.nvidia.com/default/topic/862537/cuda-setup-and-installation/installing-cuda-toolkit-on-ubuntu-14-04/
Http://unix.stackexchange.com/questions/38560/gpu-usage-monitoring-cuda
Description: Because Nvidia did not give ubuntu16 above the Cuda Toolkit, this method is not necessarily feasible, my side of the installation success, feel completely blind cat touch dead Mouse ... However, no sample is installed, only other programs can use the video card.
1. First URL, using
sudo Install Nvidia-cuda-toolkit
Installs Cuda Toolkit, depends on the network speed, the download is very slow. Also, the URL says that restarting Ubuntu is problematic (I can ' t-log in to my computer and end-in-infinite login screen). After I installed this side, the normal landing, no problem occurred.
2. Post-Installation information:
It is 7.5.17, not the latest 7.5.18, but it can be used on the line.
3. In the second URL, QED gives a continuous display of the GPU's current usage in the terminal (NVIDIA-only graphics cards):
Nvidia-smi-l 1
Results:
Description: The above command looks like a graphics card support. You can also use the commands provided by Jonathan (not currently tested):
Watch-n0. 1 " nvidia-settings-q gpuutilization-q useddedicatedgpumemory "
4. After the installation of Cuda, install Cutorch, and then install the Cunn, are installed successfully. Programs that use the GPU also work correctly.
(formerly) Ubuntu16 in the installation of CUDA Toolkit