Algorithms, Voxels, and Octrees - Oh my indeed!
Subject: Editorial, General Tech, Graphics Cards | December 2, 2011 - 03:45 AM | Scott Michaud
Even back around 2005 when the rush was still getting higher and higher resolutions to fill the finally high-resolution TVs, I always assumed that the next trail blazed in the graphics war would be lighting. Lighting is a complicated process which we are all very accustomed to it being done perfectly due to our living in the real world. Technologies such as Unreal Lightmass, PureLight, and Autodesk Beast have created more realistic lighting profiles that account for multiple bounces but cannot change in games like Mirror’s Edge. Battlefield 3, thanks to Geomerics, is one of the first games to take this problem on in semi real time such that if you alter a light the indirect lighting changes with it. The advancement does not stop there according to a recent NVIDIA blog which details research into better real time lighting.
That hand has got to be illegal in all 50 states.
P.S. -- For a 3d Technology company, just 480p Youtube -- really?
While the blog is quite vague in how the technology actually is producing its results, those results appear to be quite spectacular in quality. Unfortunately, while the quality looks amazing for being rendered at 25-70 FPS, there is no mention of what system is required to achieve those 25-70 FPS. Back to the vagueness: the demonstration is apparently not being performed upon triangular meshes relying on voxels instead. According to their explanation, their second lighting bounce is approximated to a single cone rather than multiple rays. If I understand their cone method enough, this approximation is incapable by design of expanding to third bounces and beyond; it appears to be a simplification that falls out of restricting yourself to just two lighting bounces in a voxel environment.
Regardless of when and how it will influence our technology; does the demonstration excite you for technologies to come? Place your predictions in the comments.