I am currently trying to make a server physics simulation then use snapshots to send data back to client for interpolation. Currently, everything seems work fine, I do run into problems with server able to run how many rigidbodies concurrently, it seems like when rigidbodies reaches around 100 units, the CPU start have a dramatic increase.
I am thinking about setting up a server with GPU to enhance PhysX calculations:
If I use -batchmode -nographics, would PhysX still do calculations over GPU as long as this computer has GPU?