I don't quite understand what you're trying to ask. Are you wondering how hardware manufacturers implement their shader units?

Well, the hardware is totally different. Modern GPU's (aka shader driven ones) have a series of ALU's that are used to run the shader instructions (over simplified - but you get the idea), and hence are programmable. Earlier GPU's - aka fixed function ones - don't have this at all, and are essentially "hardwired" in what they can do, all you have control over is what path is used. The Geforce 3/4 series of cards - even though they could support PS1.3 were still hardwired, in essentially a switches and dials type configuration.