PhysicsNeMo Datacenter CFD is a computational fluid dynamics dataset focused on data center thermal management, released by NVIDIA in 2026, designed to provide high-quality simulation data support for optimizing data center cooling systems.

The dataset contains normalized OpenFOAM simulation data for typical data center hot aisles, covering variations in the number of IT racks and hot aisle geometric parameters (such as length, height, and width). These configurations were simulated under steady-state conditions at maximum flow rate and rack outlet temperature (maximum load conditions), with the generated OpenFOAM data exported in VTK format.

Dataset Composition

• Simulation Data: Normalized OpenFOAM steady-state simulation results for various data center hot aisle configurations, in VTK format.
• Variable Parameters: Geometric parameters including the number of IT racks, and the length, height, and width of the hot aisle.
• Operating Conditions: Set to maximum flow rate and rack outlet temperature under maximum load conditions.
• Reference Configuration: A reference OpenFOAM configuration file is provided to reproduce the workflow, but users must modify the geometry and boundary conditions themselves.