Looking for guidance for benchmark for level design for virtual studio

We are in the process of designing environments for a virtual studio shoot and wondered if anyone has any guidance with what FPS and poly counts we should be looking for to get our levels to work with a tracked real-time camera, using an Nvidia P6000 machines.

Or if anyone has an example project to give us an idea of how much we need to optimise the levels to work smoothly in Ndisplay across the network. I know the latency is likely to be between 7-9 frames through the whole system but can we get away with using a set dressing of megascans assets for the foreground or not???

Any advice or documentation would be much appreciated.
Thanks. Rob