Graphene FETs are analog only. They can't be completely turned off like Si FETs. The drain current can only be modulated, so you can't make traditional digital circuits from them. Researchers are primarily investigating them for use in really high frequency communication circuits, among other niche applications. Don't expect a carbon Pentium any time soon.
The article lacks depth. GaAs transistors can already switch into the 100+ GHz range and have the notable advantage of being commercially producible. I'm sure there's some good science here, and maybe some practical advantages. I just don't see it in the linked text. Maybe someone who's read the paper in Science can comment?
At what cost? If it's not commercially viable, it doesn't matter. If someone could buy four processors clocked at half the speed for the same price, I doubt the chips will be useful in most circumstances.
Dammit; how are we supposed to get people to learn FP techniques now without the "we're going to be running on a hundred medium-speed CPUs in five years" problem?
