<<Using two separate buses may allow for a faster flow of data (?). >>
No, that is not correct.
Once the data has been placed in the display buffer, the RE-draw of display data is such a minuscule processor load, it is not even measurable. And since the Video RAM is right on the display card, fetching that data does not compete with instruction and operand fetching from main RAM memory.
The data rates for displays are the Fastest and most unrelenting data stream on an integrated-display computer like these Macs. But these data rates are already WAY too fast for the processors to handle every byte. Instead the processor's job is the set up a Hardware Automaton, the display-generator, whose job it is is to get the display data out of the display memory, and send it out to the display, then ask for more, about sixty times a second or more.
Your MacPro can handle only a limited number of displays. That number is limited by the number of display-generator automata inside your specific MacPro -- Six. And they are ALL on ONE of the cards, and there are no others on the other card.