Ubuntu with NVIDIA GPU driver
Deploy GPU-accelerated AI workloads on Ubuntu 22.04/24.04 with pre-installed NVIDIA drivers, CUDA, Docker, and NVIDIA Container Toolkit.
Pre-installed software
The following table lists the drivers and software included in each image.
Software | Ubuntu 22.04 with NVIDIA GPU driver | Ubuntu 24.04 with NVIDIA GPU driver | Ubuntu 22.04 (NVIDIA open-source driver) | Ubuntu 24.04 (NVIDIA open-source driver) |
Kernel version | 5.15.0-173-generic | 6.8.0-106-generic | 5.15.0-173-generic | 6.8.0-106-generic |
NVIDIA GPU driver | 580.126.09 | 580.126.09 | 580.126.09 (open-source kernel module type) | 580.126.09 (open-source kernel module type) |
NVIDIA Container Toolkit | 1.17.8 | 1.17.8 | 1.17.8 | 1.17.8 |
CUDA | 12.8 (default), 13.0 | 12.8 (default), 13.0 | 13.0 (default), 12.8 | 13.0 (default), 12.8 |
cuDNN | 9.19.1.2 | 9.19.1.2 | 9.19.1.2 | 9.19.1.2 |
NCCL | v2.27.7-1 | v2.27.7-1 | v2.29.3-1 | v2.29.3-1 |
nccl-test | v2.17.9 | v2.17.9 | v2.17.9 | v2.17.9 |
OpenMPI | 4.1.3 | 4.1.3 | 4.1.3 | 4.1.3 |
Python 3 | 3.10.12 | 3.12.3 | 3.10.12 | 3.12.3 |
Docker | 29.1.3 | 29.1.3 | 29.3.0 | 29.3.0 |
Keentune Disabled by default. | Support | Support | Supported | Supported |
OFED and elastic Remote Direct Memory Access (eRDMA) | Support | Supported | Supported | Supported |
For previous image versions, see Ubuntu Public Image Release Notes.
Ubuntu 22.04 with NVIDIA GPU driver
Instance families
ecs.gn7e,ecs.gn7s,ecs.gn7i,ecs.gn7t,ecs.gn6v,ecs.gn6i,ecs.gn6e,ecs.gn5,ecs.gn5i,ecs.gn8is,ecs.gn8v,ecs.gn8t,ecs.gn8te,ecs.gn8ep
ecs.ebmgn7e,ecs.ebmgn7i,ecs.ebmgn7t,ecs.ebmgn6v,ecs.ebmgn6i,ecs.ebmgn6e,ecs.ebmgn7ix,ecs.ebmgn7ex,ecs.ebmgn8is,ecs.ebmgn8v,ecs.ebmgn8t,ecs.ebmgn8te,ecs.ebmgn8ts
Environment variables
/etc/profile.d/nccl.sh
export NCCL_HOME=/usr/local/nccl export LD_LIBRARY_PATH=${NCCL_HOME}/lib:$LD_LIBRARY_PATH/etc/profile.d/openmpi.sh
export MPI_HOME=/usr/local/openmpi export LD_LIBRARY_PATH=${MPI_HOME}/lib:$LD_LIBRARY_PATH export PATH=${MPI_HOME}/bin:$PATH/etc/profile.d/cuda.sh
export PATH=/usr/local/cuda/bin:$PATH export CUDA_HOME=/usr/local/cuda export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
Ubuntu 24.04 with NVIDIA GPU driver
Instance families
ecs.gn7e,ecs.gn7s,ecs.gn7i,ecs.gn7t,ecs.gn6v,ecs.gn6i,ecs.gn6e,ecs.gn5,ecs.gn5i,ecs.gn8is,ecs.gn8v,ecs.gn8t,ecs.gn8te,ecs.gn8ep
ecs.ebmgn7e,ecs.ebmgn7i,ecs.ebmgn7t,ecs.ebmgn6v,ecs.ebmgn6i,ecs.ebmgn6e,ecs.ebmgn7ix,ecs.ebmgn7ex,ecs.ebmgn8is,ecs.ebmgn8v,ecs.ebmgn8t,ecs.ebmgn8te,ecs.ebmgn8ts
Environment variables
/etc/profile.d/nccl.sh
export NCCL_HOME=/usr/local/nccl export LD_LIBRARY_PATH=${NCCL_HOME}/lib:$LD_LIBRARY_PATH/etc/profile.d/openmpi.sh
export MPI_HOME=/usr/local/openmpi export LD_LIBRARY_PATH=${MPI_HOME}/lib:$LD_LIBRARY_PATH export PATH=${MPI_HOME}/bin:$PATH/etc/profile.d/cuda.sh
export PATH=/usr/local/cuda/bin:$PATH export CUDA_HOME=/usr/local/cuda export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
Ubuntu 22.04 with NVIDIA open-source driver
Instance families
ecs.gn9t,ecs.gn9gc,ecs.gn9i
ecs.ebmgn9t,ecs.ebmgn9i
Ubuntu 24.04 with NVIDIA open-source driver
Instance families
ecs.gn9g,ecs.gn9t,ecs.gn9ge,ecs.gn9gc,ecs.gn9i
ecs.ebmgn9g,ecs.ebmgn9gc,ecs.ebmgn9ge,ecs.ebmgn9t,ecs.ebmgn9i,ecs.ebmgn9ga
FAQ
How do I enable the Keentune tuning tool for an Ubuntu image with a pre-installed NVIDIA GPU driver?
Enable the tool by running the following commands. The changes take effect after a restart.
systemctl stop tuned
systemctl disable tuned
systemctl start keentune-target
systemctl enable keentune-target
systemctl enable keentuned
systemctl start keentuned
keentune profile set cpu_ubuntu_common.profileTo disable Keentune, run keentune profile rollback and restart the operating system.
What should I note when using an Ubuntu image with a pre-installed NVIDIA GPU driver in an ACK cluster?
See Create a custom image from an ECS instance and use the image to create a node and Usage notes and high-risk operations.
Switch CUDA versions
Check the current CUDA version:
nvcc --versionThese images use update-alternatives to manage multiple CUDA versions. To switch from CUDA 13.0 to 12.8, use one of the following methods:
Run
update-alternatives --config cudato interactively switch the CUDA version.update-alternatives --config cuda There are 2 choices for the alternative cuda (providing /usr/local/cuda). Selection Path Priority Status ------------------------------------------------------------ 0 /usr/local/cuda-13.0 20 auto mode 1 /usr/local/cuda-12.8 10 manual mode * 2 /usr/local/cuda-13.0 20 manual mode Press <enter> to keep the current choice[*], or type selection number: 1 --> Select the number for your desired version update-alternatives: using /usr/local/cuda-12.8 to provide /usr/local/cuda (cuda) in manual modeRun
--setto non-interactively set the CUDA version.update-alternatives --set cuda /usr/local/cuda-12.8