Can We Ever Agree On Moore’s Law?

Nearly five decades after Moore’s Law was first introduced—the famous observation turns 50 next year—we’re still debating it. There are disagreements about whether it was an economic or technology statement. There are questions about whether it was really every 24 months or every 18 months—it was actually both. And there are questions about which companies it really applied to and who it applies to now.

As any good historian will tell you, history has to be judged in its time. You’d think 50 years wasn’t much in the general scheme of things, but in semiconductors that’s the equivalent of more than 25 process nodes. And in 25 steps of doubling the number of transistors on ICs, there have been a lot of changes—economically, technologically, and even geographically. Interpreting Moore’s Law in today’s terms is a lot different than in 1965, when a single company could generate all of the necessary components and produce its own chips.

Most analog and mixed signal companies have always been distant observers of this debate. FPGA companies are newcomers. And leading-edge makers of SoCs are of two minds when it comes to Moore’s Law, its future and whether it will apply to all or only parts of chips after 14/16nm. That would make the debate even more confusing. If only the processor platform in a 3D stack is shrinking, does that mean Moore’s Law is still applicable?

You can see where this is heading, and it isn’t a neat little formula anymore. In fact, judged from the context of an SoC, or even worse, from a 2.5D package or a fan-out, it’s impossible to say whether it’s Moore’s Law or Metcalfe’s Law or some other law that isn’t a law.

What’s clear is that just shrinking features isn’t working anymore. We now have to potentially replace silicon with other materials, we have to go vertical with the transistors to fit them on the silicon without burning everything up, and we have to deal with all sorts of effects such as electromigration and ESD from 50 years of cramming things closer and closer together.

So what will become of Moore’s Law. Like everything else on an advanced semiconductor these days, it will be broken down into components, fused together with other components, and become part of a bigger structure. In the future, Moore’s Law will be just another piece. It will always be an important one, of course, and we all owe a lot of our own history to it. But it will become so small a part of the whole that you may not even recognize it. And what an interesting fade-out that is.