movie frames per second are generally around 30 and it doesn’t take 12 hours like you said - those high end graphic cards (that cost thousands themselves) can output images fairly quickly! I have heard that with ten GPUs networked together it can take as little as 1 hour to complete a rather complicated 2000x2000 plus resolution frame multiply that times 30 and then multiply that times 60 and you have about 2 weeks to one minute - but with the big budget films like nemo they can afford to have around 10 networks rendering scenes at the same time cutting the time in fractions!
Also they use other common media editing tricks like green screens where they can render a high detailed scenery on one network then render a high detail model on another and in the fraction of the time it would take to put both together they can superimpose the one over the other to make it faster - it is like some create huge images then scale them down to create anti-aliasing without havig to take the extra time required to render using built in osa!
BTW those high end GPUs are designed just for rendering so they have huge tex per pix and they can render billions of pixels per second! Those computing powers also alow the computer to computate all of the image aspects before rendering which decreases rendering time incredibly! In high end rendering Crays are leased but as desktops and small supercomputers come up in performance Crays aren’t being used as much.