floating buffer and fbo perfomance issue


I’m trying to learn how to render to float buffers using FBO. I made a simple program that renders a model to a 16-bit floating buffer then displays it on a fullscreen quad. Everything works fine, but I noticed that my framerate is slower with smaller texture sizes.

Rendering to a 1024, 512, 256, and 128 size texture works reasonably fine, but rendering to a 64 or 32 size buffer takes roughly twice as long to render. I tried setting the filter mode of the texture to both NEAREST and LINEAR, but it made no difference. Is there a minimum size for floating buffers? When I use a non-floating buffer the performance is fine across all sizes.

Is this expected behaviour or might I be making a simple mistake?

I’m using a GeForce 6800 GT card with 256 MB and the latest nVidia drivers.


can you give test app for others to see?