Developer Blog

AI / Deep Learning | HPC |

NVIDIA SimNet — AI-Accelerated Multi-Physics Simulation Toolkit

Today NVIDIA announced the availability of NVIDIA SimNet, a simulation toolkit intended to address the challenges of using AI and physics.

Simulations are pervasive in every domain of science and engineering, but they’re constrained by long computational times, limited compute resources, tedious manual setup efforts, and the need for technical expertise.  

SimNet not only accelerates simulations compared to traditional solvers but it also extends the scope of simulations into new frontiers not accessible using existing simulation tools. 

NVIDIA SimNet v0.1 is intended for academics and researchers who are either looking to get started with AI driven physics simulations or are looking to leverage a powerful, existing framework to implement their domain knowledge to solve complex real-world problems. 

Compared to traditional solvers, SimNet offers:

  • Fast turnaround time  Parameterized system representation that solves for multiple scenarios simultaneously. Once the model is trained, SimNet can do the inference interactively. Traditional simulations need to be evaluated one at a time and each run is computationally expensive. In addition, most solvers are written using CPUs not GPUs. 
  • Broad Applicability.  Model PDEs with physical constraints while maintaining accuracy and convergence. Driven by the laws of physics, SimNet is generalizable. This enables users to address a wide range of use cases including real time simulations, data assimilation, and inverse problems with easy point cloud preprocessing step. 
  • Scalable Performance.  Solve larger problems faster.  Optimized performance with Multi-GPU and Multi-Node implementation of accelerated linear algebra routines (XLA).
  • Customization and Ease-of-Adoption:  APIs for implementing new physics, domains and geometry, and detailed user guide examples.  Suitable for those just starting with AI driven physics simulations as well as experienced AI researchers needing a high-performance, modern toolkit with APIs and detailed user guide examples. 

Researchers and students interested in participating are encouraged to apply to the SimNet Early Access program today.

Learn more > >