- root or sudo permission
Please note that on the day of this writing I am using Red Hat version 8 and CentOS version 7 & 8
- Identify if the GPU is running on your system
lspci -vnn | grep VGA
or
sudo lshw -class display
- Install all development tools, kernel, EPEL repository, and other dependencies:
sudo yum update
sudo yum groupinstall "Development Tools"
sudo yum install epel-release
sudo yum install kernel-devel
sudo yum install dkms
However, you can install specific kernel-devel version on CentOS by using rpm package manager.
-
First, find and download rpm file of a specific version of library of interest from this website https://centos.pkgs.org, for example, kernel-devel-3.10.0-1062.1.2.el7.x86_64.
-
Then use the following three commands for installing
kernel-develpackage:cd /usr/local/src sudo wget http://mirror.centos.org/centos/7/updates/x86_64/Packages/kernel-devel-3.10.0-1062.1.2.el7.x86_64.rpm sudo rpm -i kernel-devel-3.10.0-1062.1.2.el7.x86_64.rpm
- Disable nouveau by modifying GRUB, like this
sudo grub2-editenv - set "$(grub2-editenv - list | grep kernelopts) nouveau.modeset=0"
- Reboot instance (and wait for 1 minute)
sudo reboot
- Switch off X server
sudo systemctl isolate multi-user.target
- Download a CUDA toolkit installer from this website https://developer.nvidia.com/cuda-downloads. This toolkit will install you both NVIDIA driver and CUDA.
cd /usr/local/src
sudo wget http://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda_10.1.243_418.87.00_linux.run
Nevertheless, you are able to install the NVIDIA driver separately by downloading an installer file from this website https://www.nvidia.com/Download/index.aspx, like this
cd /usr/local/src
sudo wget http://us.download.nvidia.com/tesla/418.87/NVIDIA-Linux-x86_64-418.87.01.run
sudo sh NVIDIA-Linux-x86_64-418.87.01.run
- Execute the installer using sh command
sudo sh cuda_10.1.243_418.87.00_linux.run
-
It will take you to the main menu of installation. Type accept to EULA agreement and select install. It can take 4-5 minutes.
-
When it is done, both NVIDIA driver and CUDA toolkits can be found at /usr/local/cuda/.
-
Finally, check if NVIDIA driver meets the graphical card
nvidia-smi
Example of output:
Tue Nov 19 02:26:13 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.39 Driver Version: 418.39 CUDA Version: 10.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla V100-SXM2... On | 00000000:04:00.0 Off | Off |
| N/A 37C P0 40W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 Tesla V100-SXM2... On | 00000000:06:00.0 Off | Off |
| N/A 34C P0 39W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 2 Tesla V100-SXM2... On | 00000000:07:00.0 Off | Off |
| N/A 33C P0 39W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 3 Tesla V100-SXM2... On | 00000000:08:00.0 Off | Off |
| N/A 34C P0 39W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
For Ubuntu, NVIDIA driver can be easily installed using apt or apt-get package manager.
- Update the system and install the drivers, for example, using the following commands
sudo apt-get update && sudo apt-get upgrade -y
sudo apt-get install --no-install-recommends nvidia-384 libcuda1-384 nvidia-opencl-icd-384
sudo reboot
The latest version of NVIDIA library on the day of this writing is 384. You can find the latest version before install using command
sudo apt search nvidia
-
The CUDA can be found at, e.g.,
/usr/local/cudaand/usr/local/cuda-10.1. -
Then see if NVIDIA driver found GPU or not by running command
nvidia-smi
Example of output:
Tue Nov 19 02:26:13 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.39 Driver Version: 418.39 CUDA Version: 10.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla V100-SXM2... On | 00000000:04:00.0 Off | Off |
| N/A 37C P0 40W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 Tesla V100-SXM2... On | 00000000:06:00.0 Off | Off |
| N/A 34C P0 39W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 2 Tesla V100-SXM2... On | 00000000:07:00.0 Off | Off |
| N/A 33C P0 39W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 3 Tesla V100-SXM2... On | 00000000:08:00.0 Off | Off |
| N/A 34C P0 39W / 300W | 0MiB / 16130MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
Rangsiman Ketkaew - rangsiman1993@gmail.com