Fragment bound means that your GPU is maxed out rendering pixels. So with DX 9 you have basically two units at work: Vertex and fragment. You can be unbound by either, meaning that your GPU has no problem dealing with the amount of vertex and fragment operations thrown at it. If you are vertex bound it means that your GPU has too many vertex operations going for it and it can't rasterize all your polygons in time. If you are fragment bound, well, then it just can't draw enough.

There are multiple reasons you can run into either. For example you can easily be fragment bound if you are drawing things in a very stupid way. For example rendering a lot of geometry back to front. You will end up drawing all the geometry and a huge amount of it will be overdrawn by geometry that is in front of it, so you are wasting a lot time drawing pixels that are never seen. Not saying that that is your issue, but that's a very easy example of such a problem.

The point is, if you are fragment bound, rendering to a larger render target by switching to fullscreen isn't doing you much good. There was already a problem pushing pixels, and now you are pushing more of them.

Of course, you could also be vertex bound and now you need to rasterize more geometry. But in reality it is way easier to make a modern GPU fragment bound than vertex bound.


Shitlord by trade and passion. Graphics programmer at Laminar Research.
I write blog posts at feresignum.com