This is forced disabled. Since the clip thread is probably the bottleneck on this hardware, it would be nice to enable it when we can, so fewer threads need to be spawned.
To investigate:
Any known errata in the guardband clipping
Guardband coordinate limits (a bit more restrictive than SNB, IIRC)
Does the clip kernel have to do extra work?
Viewport must cover entire render target surface to avoid spurious fragments
This is forced disabled. Since the clip thread is probably the bottleneck on this hardware, it would be nice to enable it when we can, so fewer threads need to be spawned.
To investigate: