You can install the Tesla driver by yourself according to the following steps. To install the CUDA toolkit, please see Installing CUDA - GPU Cloud Server - Best Practices - eSurfing Cloud (ctyun.cn).
To select the driver version, please see How to Select Driver and Related Libraries and Software Versions - GPU Cloud Server - Best Practices - eSurfing Cloud (ctyun.cn).
Prerequisites
· No driver is installed on the GPU cloud server.
· GPU cloud server is equipped with Elastic IP.
I. CentOS Driver Installation
1. Download the driver. Visit NVIDIA driver download website, select the corresponding GPU model, operating system and CUDA Toolkit version to download, this section takes A100 as an example, as shown in the figure below.
2. Click Search, select the driver version you want to download, and click Download.
3. Upload the downloaded driver installation package to the cloud server, execute the following command to add execute permissions to the installation package.
For example, add execute permission to the file named NVIDIA-Linux-x86_64-470.199.02.run.
chmod +x NVIDIA-Linux-x86_64-470.199.02.run
4. Install kernel-devel and gcc packages. Note that the kernel-devel version must be consistent with the kernel version.
sudo yum install -y gcc kernel-devel
5. Execute the following command to run the driver installer and follow the prompts for subsequent operations.
sudo sh NVIDIA-Linux-x86_64-418.126.02.run --disable-nouveau --kernel-source-path=/usr/src/kernels/$(uname -r)
6. After the installation, execute the following command to verify.
nvidia-smi
If the returned information is similar to the GPU information in the figure below, then the driver is successfully installed.
II. Ubuntu Driver Installation
1. Download the driver. Visit NVIDIA driver download website, select the corresponding GPU model, operating system and CUDA Toolkit version to download, this section takes A100 as an example, as shown in the figure below.
2. Click Search, select the driver version you want to download, and click Download.
3. Upload the downloaded driver installation package to the cloud server, execute the following command to add execute permissions to the installation package. For example, add execute permission to the file named NVIDIA-Linux-x86_64-470.199.02.run.
chmod +x NVIDIA-Linux-x86_64-470.199.02.run
4. Install gcc and linux-kernel-headers.
sudo apt-get install gcc linux-kernel-headers
5. Execute the following command to run the driver installer and follow the prompts for subsequent operations.
sudo sh NVIDIA-Linux-x86_64-418.126.02.run --disable-nouveau
6. After the installation, execute the following command to verify.
nvidia-smi
If the returned information is similar to the GPU information in the figure below, then the driver is successfully installed.
III. Windows Driver Installation
1. Download the driver. Visit NVIDIA driver download website in the GPU Cloud Server, select the corresponding GPU model, operating system and CUDA Toolkit version to download, this section takes A100 as an example, as shown in the figure below.
2. Click Search, select the driver version you want to download, and click Download.
3. Open the folder where the downloaded driver is located, double-click the installation file to start the installation, follow the prompts on the interface to install the driver, and restart the GPU cloud server as needed.
4. After installation, check the Device Manager. If the following is displayed, then the driver installation is successful.