You could be correct mate, but when I was writing computer games, the tech support guys at nVidia were always pushing the split between CPU and their GPU and the resources that each can poll independantly.
It makes sense for it to operate as you describe; with the limit in place, but that doesnt seem to fit with what we were doing back then.
Interesting - I'll have to dig around and see what I can find on that one just for my own curiosity.
Cheers
