But making the graphics better is something that doesn't seem to help your score much. More realistic fog may even make the game a little harder. So it's probably no coincidence that GPUs are leading the way in parallelism.
While videogames have been one of the main economic forces pushing processor technology, I've heard that few games can actually use more than 1 CPU core. Is this because of a difference in the way games are programmed? Is it because single-core programming is (I've heard) so much easier, and games are more often written by aspiring novices than professional drones?
It's difficult to come up with a good multicore game engine design and even more difficult to do it on the first try. What I've observed over the years is that these AAA games are steadily becoming more and more parallelizable as they are iterated on.
An architectural pattern that seems to be working well these days is a job/task-based one where the program is expressed as a graph with the tasks (well defined units of work) are connected via a network of dependencies. This model works well because it doesn't force you into requiring certain hardware configurations, it scales well from single core phones up to 32+ core workstations. Now, whether or not your content can scale as well is another question :)
The gameplay logic and networking code for most games could probably be optimized for many cores than it usually is. But again, the primary consideration is that it plays well for everybody.