I have this photo-realistic model of a car… and it has alot of polygons/vertices (1.2 millions)… I have managed to import this as a combined mesh, but the performance went quite poor once you tried to render the scene in an Oculus Rift.
Now… I have read on different forums and such that if you break up your mesh into multiple parts and then assemble the parts once they are imported this would give a performance advantage… is this true?.. so basically… is it more “heavy work” for the engine to render a combined mesh than rendering the same meshes but not as combined?
Currently there is not a way to optimize importing a realistic mesh with 1.2 million polys/verts. This is far to expensive for any modern engine to render in real time smoothly. The best thing to do would be to go into your modeling program and optimize your model and reduce the amount of polys/verts significantly. Try and find the middle ground between performance and realism for your mesh. It might take some tweaking and a bit of time, but in the end it will decrease import time, increase performance, and you will not see a huge loss of detail when applying materials to your model.
I hope this helped provide you with some valuable information on the subject!