I second Udgnim.
I moved from Ultra to High w/ 16x capped tesselation. I could not notice any visual difference here in actual gameplay. The bigger improvement in FPS I got from moving shadow distance from max to medium, like Ugdnim is saying it is definitely noticeable but the improved FPS is more noticeable than the distant shadows. Shadow distance gave me mostly reduced CPU load from my testing. The shadow distance thing ends up stopping the majority of the FPS drops as opposed to bringing up my averages. With max shadow distance I would be at 60 fps, then drop to 40, then back to 60 with some regularity. Now I drop down to the 50s only in the heaviest areas and pretty infrequently.
I haven't had to do anything else but these two things to get nearly constant 60 fps (except for seemingly random drops) with Win10, 2500k @ 4.5ghz, 16GB DDR3 1600 CL9, r9 290 @ 1100 core w/ stock memory on Cat 15.11. Gpu load looks like a very sharp sine wave going up and down only occasionally hitting 100% utilization but it doesn't go below around 40%. I also have FPS capped at 60 in my drivers because that is my screen's max Hz for full disclosure.
I'm very surprised by how much memory speed seems to affect the FPS. Big thanks to Head1985 for the testing and screenshots there, thanks man. Very unusual. Something is going on there...