Does the headless mode of Omniverse Audio2Face need NVIDIA GEFORCE RTX GPUs. I need to run Audio2Face on a virtual machine, which doesn’t have an RTX GPU. Would this work.
Here’s what our engineers have to say about this:
- A2F SDK CUDA code supports Compute Capability 5.2 and higher (Maxwell arch)
- TensortRT inference currently does not use fp16, so should also work on non-TensorCore GPU
- This is the compute capability list: CUDA GPUs - Compute Capability | NVIDIA Developer
non RTX GPUs in particular will you be using?
I’ll be using any of those GPUs, NVIDIA T4, NVIDIA L4, NVIDIA V100, NVIDIA Tesla P4, or NVIDIA Tesla P100.
From our engineering team:
Yes, these cards should be fine for A2F SDK, but no 100% guarantee without a test