pbuffer performance question

[jwatte]

This is mostly a question for nVidia and ATI driver guys, I guess, although hearing anyone with experience in the area would be helpful.

What is the overhead of repeatedly switching between rendering into a pbuffer and rendering into the framebuffer while drawing a frame? Is it "free", "on the order of texture switch" or "fairly expensive" ? I'm specifically asking about the switch cost, as I'm assuming the render cost will be comparable between frame buffer and pbuffer.

(I realize I should just code it up and try it, but before I even get that far I'd like a rough estimate).

[mcraighead]

MakeCurrent/LoseCurrent is generally going to be fairly expensive.

- Matt

[Nutty]

For example, dynamic cube maps. Does this mean it is more efficient to render dynamic cube maps into the backbuffer, or use a PBuffer?

Nutty

[davepermen]

using a pbuffer does the mostly the same.. but most of all you dont have a problem if you move another window over your one ( move the doswindow over yours at the leftbottom and you see the buggy reflection )..

try it, and we will see..

[jwatte]

Too bad, as depth_texture is nv2x only :-(

Well, thanks anyway, it saved me a lot of typing followed by disappointment.