Chess had steady algorithmic improvements on the same order as the gains from hardware: Deep Fritz and Rybka both got to better performance per FLOP than Deep Blue, etc. More generally, I think that looking at quantitative metrics (as opposed to whole new capabilities) like game performance, face recognition, image processing, etc, will often give you independent hardware and software components to growth.
Chess had steady algorithmic improvements on the same order as the gains from hardware: Deep Fritz and Rybka both got to better performance per FLOP than Deep Blue, etc. More generally, I think that looking at quantitative metrics (as opposed to whole new capabilities) like game performance, face recognition, image processing, etc, will often give you independent hardware and software components to growth.