Learning CUDA: Step 01: Installation

This topic shows steps to install NVIDIA driver and Cuda toolkit on Ubuntu.

My card GT630 is old then I have to use Ubuntu 16.04 that have gcc 5 and linux kernel 4. However, the steps are same for newer card and kernel. You must check cuda version before selecting suitable linux version for your card.

1. Preparation

• Hardware: PC with NVidia Card, GT630 is used in this series (driver 384, cuda 8.0 GA2, gcc 5)
• OS: Ubuntu 16.04

2. Remove existing driver

sudo apt-get remove --purge nvidia*
sudo apt-get remove --purge "nvidia*"

3. Install driver

sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt update
ubuntu-drivers devices
sudo apt install -y nvidia-384 libxmu-dev freeglut3-dev libx11-dev libglu1-mesa libglu1-mesa-dev libfreeimage-dev
sudo reboot

4. Reboot and check

sudo reboot
nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.130                Driver Version: 384.130                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GT 630      Off  | 00000000:03:00.0 N/A |                  N/A |
| 40%   39C    P0    N/A /  N/A |      0MiB /  1985MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0                    Not Supported                                       |
+-----------------------------------------------------------------------------+

5. Download Cuda and patches

mkdir v8.0_ga2
cd v8.0_ga2
wget https://developer.nvidia.com/compute/cuda/8.0/Prod2/local_installers/cuda_8.0.61_375.26_linux-run
wget https://developer.nvidia.com/compute/cuda/8.0/Prod2/patches/2/cuda_8.0.61.2_linux-run

6. Extract packages

chmod +x cuda_8.0.61_375.26_linux-run
chmod +x cuda_8.0.61.2_linux-run
mkdir extract
./cuda_8.0.61_375.26_linux-run --extract=$(pwd)/extract

7. Install Cuda only

sudo ./extract/cuda-linux64-rel-8.0.61-21551265.run -noprompt
sudo ./cuda_8.0.61.2_linux-run --silent --accept-eula
sudo /bin/bash -c 'echo PATH=\"\$PATH:/usr/local/cuda/bin\" >> /etc/profile'
sudo /bin/bash -c 'echo "/usr/local/cuda/lib64" >> /etc/ld.so.conf'
sudo ldconfig

8. Reboot and check

sudo reboot
nvcc --version

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2016 NVIDIA Corporation
Built on Tue_Jan_10_13:22:03_CST_2017
Cuda compilation tools, release 8.0, V8.0.61

9. Install CuDNN

Go to this link to download CuDNN https://developer.nvidia.com/cudnn, version should >= 7.0

Install to system

sudo dpkg -i cudnn/libcudnn7_7.1.4.18-1+cuda8.0_amd64.deb
sudo dpkg -i cudnn/libcudnn7-dev_7.1.4.18-1+cuda8.0_amd64.deb
sudo dpkg -i cudnn/libcudnn7-doc_7.1.4.18-1+cuda8.0_amd64.deb

Check version

cat /usr/include/cudnn.h | grep CUDNN_MAJOR -A 2

#define CUDNN_MAJOR 7
#define CUDNN_MINOR 1
#define CUDNN_PATCHLEVEL 4
#define CUDNN_VERSION    (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
#include "driver_types.h"