I used to have a rendering preformence test using 3 different drawing method: normal opengl draw, displaylist and VBO. It works well on my old ATI 9800Pro, where VBO and displaylist is much fast then normal drawing (when rendering a static scene).
Today I tested on my new Nvdia 6800. The frame rate for 3 different drawing methods are almost the same!
Maybe your tests aren’t stressfull enough. The 6800 is leaps and bounds faster (and just more efficient) than your 9800. Could you provide details about your tests exactly? I’ve had many demos where using VBO and the like not make a bit of difference b/c I wasn’t geometry (vertex) limited.
Agreed. The you have to be careful that you’re not CPU limited on the latest high end graphics hardware. If you’re not keeping the graphics pipe fed constantly, it won’t matter so much how efficiently it can render the stuff that you’re feeding it.