Running instances with GPU accelerators

This page describes how to use NVIDIA graphics processing unit (GPU) hardware accelerators on Container-Optimized OS virtual machine (VM) instances.


By using Compute Engine, you can create VM instances running Container-Optimized OS that have GPUs attached. You can only use two machine families when running GPUs on Compute Engine: accelerator-optimized and N1 general-purpose.

  • For accelerator-optimized machine types, each machine type has a specific model of NVIDIA GPUs attached.

    • For A3 accelerator-optimized machine types, NVIDIA H100 80GB GPUs are attached.
    • For A2 accelerator-optimized machine types, NVIDIA A100 GPUs are attached. These are available in both A100 40GB and A100 80GB options.
    • For G2 accelerator-optimized machine types, NVIDIA L4 GPUs are attached.
  • For N1 general-purpose machine types, you can attach the following GPUs:

GPUs provide compute power to drive deep-learning tasks such as image recognition and natural language processing, as well as other compute-intensive tasks such as video transcoding and image processing.

Google Cloud provides a seamless experience for you to run your GPU workloads within containers on Container-Optimized OS VM instances so that you can benefit from other Container-Optimized OS features such as security and reliability.

To learn more about the use cases for GPUs, see Cloud GPUs.

To learn about using GPUs on Google Kubernetes Engine (GKE), see Running GPUs on GKE.


Running GPUs on Container-Optimized OS VM instances has the following requirements:

  • Container-Optimized OS x86 images: only x86-based Container-Optimized OS images support running GPUs. Arm-based Container-Optimized OS images don't support the feature.

  • Container-Optimized OS version: To run GPUs on Container-Optimized OS VM instances, the Container-Optimized OS release milestone must be a LTS milestone and the milestone number must be 85 or higher.

  • GPU quota: You must have Compute Engine GPU quota in your chosen zone before you can create Container-Optimized OS VM instances with GPUs. To ensure that you have enough GPU quota in your project, see Quotas in the Google Cloud console.

    If you require additional GPU quota, you must request GPU quota in the Google Cloud console. If you have an established billing account, your project automatically receives GPU quota after you submit the quota request.

  • NVIDIA GPU drivers: You must install NVIDIA GPU drivers by yourself on your Container-Optimized OS VM instances. This section explains how to install the drivers on Container-Optimized OS VM instances.

Create a VM

The following sections explain how to run GPUs on Container-Optimized OS VMs.

First, you need a Container-Optimized OS VM instance with GPUs. The method used to create a VM depends on the GPU model selected.

You can also add GPUs to existing Container-Optimized OS VM instances.

When you create VMs, remember to choose images or image families from the cos-cloud image project.

To check all GPUs attached to your current Container-Optimized OS VM instances, run the following command:

gcloud compute instances describe INSTANCE_NAME \
    --project=PROJECT_ID \
    --zone ZONE \

Replace the following:

Install NVIDIA GPU device drivers

After you create an instance with one or more GPUs, your system requires device drivers so that your applications can access the device. This guide shows the ways to install NVIDIA proprietary drivers on Container-Optimized OS VM instances.

Container-Optimized OS provides a built-in utility cos-extensions to simplify the NVIDIA driver installation process. By running the utility, users agree to accept the NVIDIA license agreement.

Identify GPU driver versions

Each version of Container-Optimized OS image has a default supported NVIDIA GPU driver version. See the release notes of the major Container-Optimized OS LTS milestones for the default supported version.

You may also check all the supported GPU driver versions by running the following command on your Container-Optimized OS VM instance:

sudo cos-extensions list

Identify the required CUDA toolkit version

If your applications use CUDA, install NVIDIA's CUDA toolkit in your containers. Each version of CUDA requires a minimum GPU driver version or a later version. To check the minimum GPU driver version required for your version of CUDA, see CUDA Toolkit and Compatible Driver Versions. Ensure that the Container-Optimized OS version you are using has the correct GPU driver version for the version of CUDA you are using.

Install the driver

You can install GPUs by using either shell commands, startup scripts, or cloud-init. All three methods use the sudo cos-extensions install gpu command to install the default GPU driver for your Container-Optimized OS LTS version.


After you connect to your Container-Optimized OS VM instances, you can run the following command manually to install drivers:

sudo cos-extensions install gpu

Startup scripts

You can also install GPU drivers through startup scripts. You can provide the startup script when you create VM instances or apply the script to running VM instances and then reboot the VMs. This lets you install drivers without connecting to the VMs. It also makes sure the GPU drivers are configured on every VM reboot.

The following is an example startup script to install drivers:

#! /bin/bash

sudo cos-extensions install gpu


Cloud-init is similar to startup scripts but more powerful. The following example shows how to install GPU driver through cloud-init:


  - cos-extensions install gpu

Using cloud-init lets you specify the dependencies so that your GPU applications will only run after the driver has been installed. See the End-to-end: Running a GPU application on Container-Optimized OS section for more details.

For more information about how to use cloud-init on Container-Optimized OS VM instances, see the creating and configuring instances page.

In some cases the default driver included with Container-Optimized OS doesn't meet the minimum driver requirements of your CUDA toolkit or your GPU model. See the Required NVIDIA driver versions for the version requirements for specific types of GPUs.

To install a specific GPU driver version, run the following command:

sudo cos-extensions install gpu -- -version=DRIVER_VERSION

Replace DRIVER_VERSION with one of the following options:

  • default: Installs the default driver designated by the Container-Optimized OS release. This version receives bug fixes and security updates.
  • latest: Installs the latest driver available in the Container-Optimized OS release. Be aware that this might introduce compatibility changes due to potential major version updates across COS releases.
  • The full version: Use this to pin to a specific version for workloads sensitive to driver changes. For example, specify version 535.183.01.
  • NVIDIA driver branch: Installs the latest stable driver within a specific NVIDIA branch to stay current with security updates and bug fixes within that branch. For example, specify branch R535. This option is available starting from cos-gpu-installer:v2.2.1.

To see the available versions for each of those options, run the command to Identify GPU driver versions.

Verify the installation

You can run the following commands on your Container-Optimized OS VM instances to manually verify the installation of the GPU drivers. The output from the command shows the GPU devices information, such as devices state and driver version.

# Make the driver installation path executable by re-mounting it.
sudo mount --bind /var/lib/nvidia /var/lib/nvidia
sudo mount -o remount,exec /var/lib/nvidia

Configure containers to consume GPUs

After the GPU drivers are installed, you can configure containers to consume GPUs. The following example shows you how to run a CUDA application in a Docker container that consumes /dev/nvidia0:

docker run \
  --volume /var/lib/nvidia/lib64:/usr/local/nvidia/lib64 \
  --volume /var/lib/nvidia/bin:/usr/local/nvidia/bin \
  --device /dev/nvidia0:/dev/nvidia0 \
  --device /dev/nvidia-uvm:/dev/nvidia-uvm \
  --device /dev/nvidiactl:/dev/nvidiactl \

You can run your containers through cloud-init to specify the dependency between driver installation and your containers. see the End-to-end: Running a GPU application on Container-Optimized OS section for more details.

End-to-end: Running a GPU application on Container-Optimized OS

The following end-to-end example shows you how to use cloud-init to configure Container-Optimized OS VM instances that provision a GPU application container myapp:latest after the GPU driver has been installed:


- name: myuser
  uid: 2000

  - path: /etc/systemd/system/install-gpu.service
    permissions: 0644
    owner: root
    content: |
      Description=Install GPU drivers docker.socket docker.socket

      ExecStart=cos-extensions install gpu
  - path: /etc/systemd/system/myapp.service
    permissions: 0644
    owner: root
    content: |
      Description=Run a myapp GPU application container

      ExecStart=/usr/bin/docker run --rm -u 2000 --name=myapp --device /dev/nvidia0:/dev/nvidia0 myapp:latest

  - systemctl daemon-reload
  - systemctl start install-gpu.service
  - systemctl start myapp.service

About the NVIDIA CUDA-X libraries

CUDA® is NVIDIA's parallel computing platform and programming model for GPUs. To use CUDA applications, the libraries must be present in the image you are using. You can do any of the following to add the NVIDIA CUDA-X libraries:

  • Use an image with the NVIDIA CUDA-X libraries pre-installed. For example, you can use Google's Deep Learning Containers. These containers pre-install the key data science frameworks, the NVIDIA CUDA-X libraries, and tools. Alternatively, NVIDIA's CUDA image contains the NVIDIA CUDA-X libraries only.

  • Build and use your own image. In this case, include /usr/local/cuda-XX.X/lib64, which contains the NVIDIA CUDA-X libraries, and /usr/local/nvidia/lib64, which contains the NVIDIA device drivers, in the LD_LIBRARY_PATH environment variable. For /usr/local/cuda-XX.X/lib64, the name of the directory depends on the version of the image you used. For example, the NVIDIA CUDA-X libraries and debug utilities in Docker containers can be at /usr/local/cuda-11.0/lib64 and /usr/local/nvidia/bin, respectively.


Just like other kernel modules on Container-Optimized OS, GPU drivers are cryptographically signed and verified by keys that are built into the Container-Optimized OS kernel. Unlike some other distros, Container-Optimized OS does not allow users to enroll their Machine Owner Key (MOK) and use the keys to sign custom kernel modules. This is to ensure the integrity of the Container-Optimized OS kernel and reduce the attack surface.


Container-Optimized OS version restrictions

Only Container-Optimized OS LTS release milestone 85 and later support the cos-extensions utility mentioned in the Installing NVIDIA GPU device drivers section. For earlier Container-Optimized OS release milestones, use the cos-gpu-installer open source tool to manually install GPU drivers.

VM instances restrictions

VM instances with GPUs have specific restrictions that make them behave differently than other instance types. For more information, see the Compute Engine GPU restrictions page.

Quota and availability

GPUs are available in specific regions and zones. When you request GPU quota, consider the regions in which you intend to run your Container-Optimized OS VM instances.

For a complete list of applicable regions and zones, see GPUs on Compute Engine. You can also see GPUs available in your zone using the Google Cloud CLI.

gcloud compute accelerator-types list


For GPU pricing information, see the Compute Engine pricing page.


Each Container-Optimized OS release version has at least one supported NVIDIA GPU driver version. The Container-Optimized OS team qualifies the supported GPU drivers against the Container-Optimized OS version before release to make sure they are compatible. New versions of the NVIDIA GPU drivers may be made available from time-to-time. Some GPU driver versions won't qualify for Container-Optimized OS, and the qualification timeline is not guaranteed.

When the Container-Optimized OS team releases a new version on a release milestone we try to support the latest GPU driver version on the corresponding driver branch. This is to fix CVEs discovered in GPU drivers as soon as possible.

If a Container-Optimized OS customer identifies an issue that's related to the NVIDIA GPU drivers, the customer must work directly with NVIDIA for support. If the issue is not driver specific, then users can open a request with Cloud Customer Care.

What's next