DGX Spark Setup

This guide covers the required steps to set up anNVIDIA DGX Spark. The DGX Spark runs DGX OS, an Ubuntu-based distribution with drivers and utilities optimized for the DGX Spark hardware.

The installation guide aims to be self-contained. However, theDGX Spark User Guide is a good reference for further details.

Post-Installation Setup

Note

The following steps can also be executed via:

./scripts/configure-system.dgx-spark.sh

Update packages and install dependencies:

sudoaptupdatesudoaptdist-upgrade-ysudoaptinstall-yapt-utilscoreutilsgit-coregitcmakebuild-essential\bclibssl-devpython3python3-pipninja-buildca-certificatescurlpandoc

Download the Sionna Research Kit:

cd~# We assume sionna-rk is cloned in the home directorygitclone--recurse-submoduleshttps://github.com/NVlabs/sionna-rk.git

Docker Installation

Install Docker from the official Docker repository:

# Add Docker's official GPG keysudoinstall-m0755-d/etc/apt/keyringssudocurl-fsSLhttps://download.docker.com/linux/ubuntu/gpg-o/etc/apt/keyrings/docker.ascsudochmoda+r/etc/apt/keyrings/docker.asc# Add the repository to Apt sourcesecho\"deb [arch=$(dpkg--print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \$(./etc/os-release&&echo"${UBUNTU_CODENAME:-$VERSION_CODENAME}") stable"|\sudotee/etc/apt/sources.list.d/docker.list>/dev/null# Install Docker, plugins, and NVIDIA container toolkitsudoaptupdatesudoaptinstall-ydocker-cedocker-ce-clicontainerd.iodocker-buildx-plugindocker-compose-pluginnvidia-container-toolkit# Add user to docker groupsudousermod-aGdocker$USER# Log out and log in again for changes to take effect

Configure Docker runtime for GPU support:

sudonvidia-ctkruntimeconfigure--runtime=docker# Create Docker service overridesudomkdir-p/etc/systemd/system/docker.service.dsudotee/etc/systemd/system/docker.service.d/override.conf<<EOF[Service]Environment="DOCKER_INSECURE_NO_IPTABLES_RAW=1"EOF# Restart Dockersudosystemctldaemon-reloadsudosystemctlrestartdocker

Set the following environment variables:

exportSRK_PLATFORM="DGX Spark"exportSRK_THREAD_POOL="15,16,17,18,19"exportSRK_UE_THREAD_POOL="4,5"

TensorRT Installation

Install TensorRT and monitoring tools:

sudoaptinstall-ycuda-toolkittensorrtnvtop# Add trtexec alias for convenienceecho'alias trtexec=/usr/src/tensorrt/bin/trtexec'>>~/.bash_aliases

Version Information

Check OS version:

cat/etc/lsb-releaseDISTRIB_ID=UbuntuDISTRIB_RELEASE=24.04DISTRIB_CODENAME=nobleDISTRIB_DESCRIPTION="Ubuntu 24.04.3 LTS"

Check DGX OS version:

cat/etc/dgx-releaseDGX_NAME="DGX Spark"DGX_PRETTY_NAME="NVIDIA DGX Spark"DGX_SWBUILD_DATE="2025-09-10-13-50-03"DGX_SWBUILD_VERSION="7.2.3"DGX_COMMIT_ID="833b4a7"DGX_PLATFORM="DGX Server for KVM"DGX_SERIAL_NUMBER="XXXXXXXXXXXX"DGX_OTA_VERSION="7.3.1"DGX_OTA_DATE="Wed Nov 19 05:05:30 PM CET 2025"