At last week's WinHEC show, Nvidia Chief Scientist David Kirk pontificated on where 3D graphics rendering is going over the next decade. What can you expect? Photorealistic rendering and cinematic realism, of course. It's not about pushing clock speeds and pixel rates, but rather making pixels much better looking! Kirk elaborated that we've moved from texture-mapped rendering to programmable pixel shading in pursuit of cinematic realism, but we aren't quite there yet. In particular, real-time 3D graphics rendering is not yet capable of global illumination. In texture-mapped rendering, each pixel in each triangle is drawn separately, and the color is determined by the information attached to the triangle's vertices. A similar situation exists with programmable shaders, where shadows, reflections, and resultant pixel lighting is calculated in a serial manner. Pixels are actually painted with texture information, and then shadows, reflections, and other information are blended in to make the final pixel color. In contrast, the process used to generate images like those below is fundamentally different.