Operating Systems

NVIDIA DGX systems are purpose-designed machines for the AI infrastructure, from analytics to training to inference. They are built around performant interconnected GPUs providing unprecedented acceleration. DGX servers include dedicated network interface cards for direct memory transfers between the GPUs of different nodes reducing the bandwidth bottleneck.

DGX systems come preinstalled with DGX™ OS, a customized installation of Ubuntu with additional system-specific software from NVIDIA building the DGX Software Stack. It provides platform-specific configurations, diagnostic and monitoring tools, and includes all drivers that are required for a stable, tested, and supported OS for these AI workloads.

Customers also have the option to install Red Hat Enterprise Linux or vanilla Ubuntu and the NVIDIA DGX Software Stack on a DGX while still benefiting from the advanced DGX features. This enables cluster installations that allows users to share the resources while also monitoring and improving system utilization. The DGX software supports traditional HPC installations using workload managers such as SLURM or PBS, and Kubernetes, the choice of many modern deployments.

Additional information

Additional information and installation instructions for various system operating systems can be found in the software section of the DGX Systems Documentation.