correct, the currents add up. The voltages remain the same in parallel.
Like running batteries in parallel. You increase the overall current sourcing capability of the original single batteries, but retain the same voltage.
Power sourcing increases likewise.
Pt = vt*(i1 + i2 + i3 + ...) per rail.
If I recall correctly, my nvidia GeForce graphics card required two separate 12v molex rails connected for that reason.
The PSU manufactures are trying to accommodate a wide variety of loads as efficiently as possible. While the video card manufactures are aware of this they also make provisions; sometimes when
unexpectedly hungry loads like GeForce come along, that is why they provide for two sources connected in parallel to get the wattage capability required from the supply rail.
Separating the rails this way helps in both current limiting under unknown loads as well as maximizes efficiency.
P.S. Just make sure your video card has more than one molex power connector to add more than one rail. Or else, the card itself will limit you. I.e. in the case where you required a current much larger than your supply outputted per each +12V molex connector, but only had one female molex connector available, you would have to either buy a supply with a much larger power available per single rail connector output (like some of those shown earlier on Landis' post), or you would have insufficient drive from the less equipped multiple rail supply.
