Originally posted by Edge
I simply think we need hardware acceleration for light reflections. It would make things a LOT easier, and if a 5 year old soundcard can trace SOUND reflections, why can't a modern day videocard be made to trace LIGHT reflections?
The simple answer is that light has to be traced at a much
higher resolution. Anyway, raytraced lighting really isn't that good, and there are tricks available on modern video cards that get you most of the way to raytraced lighting, without the huge performance impact (provided you're not worried about multiple reflections and stuff...).
I think faking radiosity would be easy if someone would actually try. For example, you could have a negative light radius around an object, so the closer it gets to something the darker the surface gets.
I'm not sure that has much at all to do with radiosity.
Anyway, lighting is indeed the big problem, and it's not a small one. Doing proper lighting just requires algorithms that are undoable in realtime. People are working on approximations that get you 99% of the way there, but it's just taking time to get the processing power and flexibility up high enough.
One possible bit of excitement is that if quantum computing does start to become viable in the next 5-10 years, it won't just increase processing power, but it will open the way to whole new algorithms for computing that just weren't available before (basically, since a quantum computer works differently, it likes different kinds of algorithms altogether...it would be comparitively poor for running today's programs, but would be excellent for other types of programs...).