Enabling GPU for PhysicsAI 2025.1 version on a Linux machine
Hello PhysicsAI users,
When using PhysicsAI 2025.1 version on a Linux machine, cuDNN 8.9.7 version needs to be installed on the machine.
- CUDA Toolkit 11.8 version remains the same.
If cuDNN 8.7 version is used along with CUDA Toolkit 11.8 on a Linux machine for PhysicsAI model training, one can either get the below error when using TNS architecture or model training will be significantly slower when using GCNS architecture.
- Error when using TNS architecture : Loaded runtime cuDNN library: 8.7.0 but source was compiled with : 8.9.7
Appendix:
CUDA Toolkit 11.8 :
https://developer.nvidia.com/cuda-toolkit-archive
CUDNN 8.9.7
https://developer.nvidia.com/rdp/cudnn-archive
For others who are interested in configuring GPU on a Windows machine or on a Linux machine (having 2025.0 HyperWorks version), can refer the below article.
