lots of excellent points. size of the window will definitely stress
a hardware accelerated machine's fillrate performance. unless your
scene consists of thousands of enormous polygons and lots of overdraw,
a 640x480 scene is much more likely to be polygon-bound. newer chips
such as the GeForce from nvidia support geometry acceleration on-chip
(which, for some reason, they call hardware transform & lighting,
probably to appeal to the directX set), which offload most of those
calculations to the graphics hardware.
there are also bus issues: although vertex data is fairly light,
texture data is not. for hw accelerated machines, if your scene
contains more textures than will fit into the card's texture ram,
you will be hitting the AGP bus pretty hard with texture downloads,
so the speed of the bus might come into play here.
stephen
--
stephen f. white
[EMAIL PROTECTED]
http://www.csclub.uwaterloo.ca/u/sfwhite/
i'm not a complete freak; some parts are missing.
