In 1965, Intel co-founder Gordon Moore predicted that the number of transistors on a given area of chip would double every two years. For like ten years or something. So far, Moore's Law (as it's now known) has been true for close to fifty years, and Intel has its sights set on keeping it rolling for another decade.
Right now, you can go out and buy computer processors with 22-nanometer features. The smaller the features that can be manufactured, the denser you can pack them together. That means more transistors can be crammed onto a chip, making for a faster (and more efficient) processor. A 22 nanometer transistor is tiny, about the same size as arhinovirus. That's much smaller than a 32 nm transistor (which is the size of the previous generation), but Intel has started talking about its 10-year roadmap that gets transistor size all the way down to just five nanometers.
The next step down from 22 nm is going to be 14 nm, which should show up next year, with 10 nm transistors making an appearance in 2015 or so. After that, we're looking at 7 nm followed by 5 nm, at which point you should be paying just a tenth of the cost per transistor as you are right now. Faster, cheaper, more efficient — all good things.
Intel isn't saying just exactly how it's going to make it down to transistors that small, just that it's pretty sure it can make it work. According to Intel senior fellow Mark Bohr, "Intel's technology pipeline is full with research extending out 10 years and down to the 10 nm, 7 nm, and 5 nm nodes. It looks like we have a a solution for 10 nm and I'm confident we will have solutions for 7 nm and 5 nm." Those solutions may very well involve a departure from silicon into something crazy like graphene. That'll involve some breakthroughs and likely a new manufacturing process, but Intel will make it happen, because it's the law, man. Moore's Law. And you don't break Moore's Law.
Posts
