I understand that every P-N junction emits radiation as a byproduct. I always thought that if we could observe the right part of the spectrum, it should be possible to "see" the transistors in a processor die light up as it computes. The fact that nobody's done this (that I'm aware of) probably means it's probably not practical.