In that case, what are the 8 segments that you split a disk into?
The vertex buffer is "all terrain"
The disk is "potentially visible terrain"
The segments are "visible terrain"
By intersecting the view frustum with the segments you can render only the segments that are visible.
You can also use a draw call per segment and using the startindex and primcount to render only the visible depth band
This can handle degenerate case of camera pointing straight down (requiring small depth range of all 8 segments)
Also consider using a "plug" for the inner disk to simplify things.
