I am using UE 5.3 and I am now looking at the animation system with its multi threaded Thread Safe Animation Update function. You can’t do as much with it as the original Animation Update, but it offloads the work from the game thread. This is still CPU based, but it does improve overall frame rate by reducing the game thread work load.
I also found that the spline curve calculations can be expensive (e.g., finding the 3D position on the curve at a given distance into the curve). I prototyped it in BP, but I knew from the start I would have to replace it with a highly optimized C++ solution if I chose to go that route. What I mean by optimize, quite possibly use my own highly optimized methods of calculating the position, not even relying on UE5’s framework.
Isn’t there a way to push large scale parallel calculations onto the GPU (not talking about graphics/mesh/materials/textures) for even faster results? I know there is a solution by NVIDIA, but I don’t know if it is available in UE5.