Install NVIDIA GPU drivers on N-series VMs running Linux

Applies to: ✔️ Linux VMs

Important

To align with inclusive language practices, this documentation replaces the term "blacklist" with "blocklist". This change reflects a commitment to avoiding terminology that might carry unintended negative connotations or perceived racial bias. However, in code snippets and technical references where "blacklist" is part of established syntax or tooling (for example, configuration files, command-line parameters), the original term is retained to preserve functional accuracy. This usage is strictly technical and doesn't imply any discriminatory intent.

To install NVIDIA GPU drivers manually, follow the steps outlined in this article. Manual driver setup information is also available for Windows VMs.

For technical specifications, see GPU Linux VM sizes.

Warning

Installing NVIDIA drivers by using methods other than those outlined in this article might result in failure of the intended driver installation. Microsoft and NVIDIA don't support this installation. To ensure proper functionality and support, follow only the installation steps and use the driver versions specified in this article.

CUDA drivers

NVIDIA distributes CUDA drivers for NCasT4_v3 VMs.

Supported CUDA drivers

For the latest CUDA drivers and supported OS, visit the NVIDIA website. Ensure that you install or upgrade to the latest supported CUDA driver for your distribution.

CUDA driver installation

For CUDA driver installation, visit the NVIDIA website.

Verify CUDA driver installation

Run nvidia-smi. If the driver is installed, NVIDIA SMI lists the GPU-Util as N/A until you run a GPU workload on the VM.

GRID drivers

Microsoft redistributes NVIDIA GRID driver installers for NCasT4_v3, and NVadsA10_v5 VMs used as virtual workstations or for virtual applications.

Supported GRID drivers

Install these GRID drivers only on these VMs and only on the operating systems listed in the following table. These drivers include licensing for GRID Virtual GPU (vGPU) software in Azure. You don't need to set up an NVIDIA vGPU software license server.

The NCasT4_v3 series are the only non-NV GPU VM series that support GRID drivers.

VM Series GPU GRID Driver Supported Linux OS Versions
NCasT4_v3 NVIDIA Tesla T4 GPU (16GB) vGPU20 • Ubuntu 24.04 LTS
• Ubuntu 22.04 LTS
• Ubuntu 20.04 LTS
• Red Hat Enterprise Linux (RHEL) 9.4, 9.6, 9.7
• Red Hat Enterprise Linux (RHEL) 8.10
NVadsA10_v5 NVIDIA A10 GPU (24GB) vGPU18.8 • Ubuntu 24.04 LTS
• Ubuntu 22.04 LTS
• Ubuntu 20.04 LTS
• Red Hat Enterprise Linux (RHEL) 9.4, 9.6, 9.7
• Red Hat Enterprise Linux (RHEL) 8.10

For more information on the specific vGPU and driver branch versions, visit the NVIDIA website.

GRID driver installation

  1. Ensure that Secure Boot and vTPM are disabled.

  2. Install prerequisites.

    Ubuntu:

    sudo apt update 
sudo apt install -y build-essential

    RHEL (8.10, 9.4, 9.6, 9.7):

    sudo yum check-update 
sudo yum install -y make automake gcc gcc-c++ kernel-devel-$(uname -r) kernel-headers-$(uname -r)

  3. Download the Linux driver.

    a. For NVadsA10_v5:

    wget -O ./NVIDIA-Linux-x86_64-570.211.01-grid-azure.run https://download.microsoft.com/download/2a04ca6a-9eec-40d9-9564-9cdea1ab795f/NVIDIA-Linux-x86_64-570.211.01-grid-azure.run

  4. Install the driver.

    a. For NCasT4_v3:

    sudo chmod +x ./NVIDIA-Linux-x86_64-595.58.03-grid-azure.run
sudo ./NVIDIA-Linux-x86_64-595.58.03-grid-azure.run

    b. For NVadsA10_v5:

    sudo chmod +x ./NVIDIA-Linux-x86_64-570.211.01-grid-azure.run
sudo ./NVIDIA-Linux-x86_64-570.211.01-grid-azure.run

Verify GRID driver installation

Run nvidia-smi. If the driver is installed, NVIDIA SMI lists the GPU-Util as N/A until you run a GPU workload on the VM.

Troubleshooting

  • You can set persistence mode using nvidia-smi so the output of the command is faster when you need to query cards. To set persistence mode, run nvidia-smi -pm 1. If the VM is restarted, the mode setting goes away. You can always script the mode setting to run upon startup.

  • If jobs are interrupted by ECC errors on the GPU (either correctable or uncorrectable), first check to see if the GPU meets any of Nvidia's RMA criteria for ECC errors. If the GPU is eligible for RMA, contact support about getting it serviced; otherwise, reboot your VM to reattach the GPU as described here. Less invasive methods, such as nvidia-smi -r, don't work with the virtualization solution deployed in Azure.

Next steps

  • Last updated on