It works in parallel with the GPU, which is more important - in effect, the rendering is threaded on the graphics card, which is a big timesaver.
As for CPU cores, it's actually fairly difficult to utilise them because most of the runtime is sequentially executed. However, some candidates for possible multicoring is the Physics behavior, the RTS movement pathfinding, and maybe some other processing heavy plugins, but that's the best that can be managed I think.