As a researcher in the field of computer graphics, I can safely say that we haven't seen the end of it. Yes, a lot of hope has to be put on faster hardware. And lets face it - light is a tricky business...you do not think about it, but the complexity of the processes involved that enable you your eyesight is enormous.
Yet, every now and then a new algorithm is developed, which introduces more accuracy and less approximations, which runs faster and more optimal. It is not yet sure which idea will prevale (ray tracing or raster or something else), but it is sure that visible pixelization, coarse models, unrealistic lighting etc are going to be pretty much the past. When? Nobody can honestly answer this question...lets wait and see.