Supported Features

Different types of workloads in NVIDIA Run:ai, including both NVIDIA Run:ai native workloads and workloads enabled via the Resource Interface (RI) offer varying levels of feature support. When selecting a workload, it is important to consider which platform capabilities are required for your use case.

The availability of specific features and capabilities can evolve across different NVIDIA Run:ai versions. Refer to the table and documentation for the most current support details.

Note

Other Kubernetes workloads can also be submitted via kubectl, receiving only minimal scheduling and platform capabilities.

Workload Submission

	Workspace	Standard Training	Distributed Training	Inference	Distributed Inference	Workloads via Resource Interface
UI
API
CLI

Scheduling and Resource Management

Functionality	Workspace	Standard Training	Distributed Training	Inference	Distributed Inference	Workloads via Resource Interface
Fairness
Priority and preemption
Over quota
Node pools
Bin packing / Spread
Multi-GPU fractions
Multi-GPU dynamic fractions
Node level scheduler
Multi-GPU memory swap
Elastic scaling
Gang scheduling
Network topology-aware scheduling
GB200 NVL72 and Multi-Node NVLink domains (MNNVL)
Scheduling rules

Operational and Platform Features

Functionality	Workspace	Standard Training	Distributed Training	Inference	Distributed Inference	Workloads via Resource Interface
Monitoring
Workload awareness
RBAC
Workload actions (stop/run)
Rolling updates
Workload templates
Workload assets
Workload Policies

Workload awareness

Specific workload-aware visibility, so that different pods are identified and treated as a single workload (for example GPU utilization, workload view, dashboards).

Other Kubernetes Workloads

Other Kubernetes workloads can also be submitted via kubectl, receiving full monitoring and minimal scheduling capabilities.

• Supported scheduling capabilities include:

  • Fairness

  • Priority and preemption

  • Over quota

  • Node pools

  • Bin packing / Spread

  • Multi-GPU fractions

  • Multi-GPU dynamic fractions

  • Node level scheduler

  • Multi-GPU memory swap

  • Gang scheduling

• All monitoring capabilities are supported including event history, metrics and logs.