You have your game - stuff runs on the cpu - ai, physics, math, some collisions etc. it calculates a scene and passed it to the card(s).
If you have 2 cards, then you double up on bandwidth and pixel processor power, with the limitation that the power is combined at a higher level than the most optimal - i.e. in the chip. If you're pushing extremely hogh resolutions, then the doubling of fillrate and bandwidth can nearly double the frame rate.
Lets say your game wasnt desiged for dual core (in other words, it's not multithreaded - think of a thread of a series of instructions to be computed for the game). If a game is multithreaded - say the AI and physics run in threads seperate from the rest, then the AI thread can occupy the attention of one cpu core, the physics another, and the rest of the game on another core, distributing the processong power accross multiple cores.
For games that run one thread, the game puts it all on one cpu core. in a dual core or multi processor machine the other cores have nothing to do, and arn't beneficial. In a game like quake 4 which does support multiple processors, you can get a performance boost using dual core.
The thing is that most processors arn't saturated enough to as yet necessitate dual cores for gaming. The videocard and ram should be top priority. Even an awesomely powerful card like teh x1900xt appears to skip frames with 1GB of ram playing battle field 2. The reason is because all the data can't be held in ram, so it constantly swaps data to and from the harddrive - the slowest component in your system. With 2GB, it was smooth sailing.