See, now THIS is the future. Not more cores, because we'll eventually hit a physical limit there, not smaller manufacturing processes, because we'll hit a limit around 11 nanometers with the whole quantum tunneling nonsense...
Plus, we'll have IMMENSE amounts of heat to deal with.
The future is materials. Screw silicon. Screw copper. Processors made of a room-temperature superconductor and paths on logic boards printed in (well, since we're talking about room-temperature superconductors, now) that same superconductor.
Forget this gigatransfer business. We'll be in the yottatransfers for decades before the rest of the system technologies could catch up.
Materials has
always been the key. It's actually how I began. BS in Materials, then Computer Eng.
Silicon, and the current methodologies have their limits. This has been known for some time. A couple of alternative technologies are exist, or are in development. One, is taking the signal optical. A nice way to eliminate heat as an issue. I studied with Dr. Hummel at the University of Florida, and he perfected an optical design of IBM's 486 equivalent back in ('90?). It's advanced since then, but the only systems I'm truly aware of, are way beyond what is affordable to the masses. DOD territory, so the details are classified.
Another attempt, was to make the gates incredibly small. One molecule to be precise acting like a transistor (axial rotation with a wire at each end). It always reminded me of a compass with a wire at E and W. They were successful at developing the materials, and the process to do it, but IIRC, it was very difficult. Not sure if the existing form could be mass produced cheaply. It was called the Quantum Computer. Lucent and later, HP worked on it. I'm not sure what happened, as I haven't seen anything on it recently.
Silicon is already "dead". Hafnium is the current medium used in the current Intel parts (45nm). As for copper, not yet. IIRC, superconductor development is still looking hard at copper and nickel as it's base.
Personally, I think we'll continue to see multi core systems grow. Paralleling processors is a way to get around Moore's Law in terms of a single processor solution. Materials not yet available? Pair 'em.
It'll get exponential, and eventually Moore's will catch up here as well. But it buys time to develop and refine other solutions.
Now if software would catch up...
