Moore ponders on the end of his law

Two hundred Intel employees gathered on the front lawn of the company’s Californian campus last week to mark its 35th anniversary. There they buried a time capsule containing, amongst other things, an Itanium 2 processor, chopsticks donated by Intel’s Malaysian subsidiary, and a copy of Time Magazine featuring Intel co-founder and Chairman, Andy Grove, on the cover.

In attendance at the ceremony was another of Intel’s founders, Gordon Moore, who served as the company’s CEO from 1975 to 1987 and is best known as the creator of Moore’s Law, which predicts that the number of transistors on a computer chip will double every two years. Though Moore says it took years before he could even refer to his idea as Moore’s Law, the maxim has proved to be a remarkably accurate predictor of the computer industry’s growth since it was coined in 1965. But would it remain accurate into the future?

How long will Moore’s law remain in effect?

Gordon Moore (GM): The thing that’s been driving it (since it was formulated in 1965) is the ability to make things smaller and smaller, and eventually the fact that the materials are made of atoms is a real limit. We’ve started seeing quantum mechanical effects even in the devices we’re making now. I think we’ve got two or three more generations, moving in the same path we’ve been on. Then we’ll have to change.

We can make bigger chips. It may not be at quite the breathtaking pace it’s been so far — something that doubles every four years instead of every two years is still almost unprecedented. It will slow down. It depends on a lot of factors that are hard to predict. It depends a lot on investment. A lot of investment is a lot easier in a growing market than it is in a flat market. I expect the market will continue to grow, but these things are all tied together.

So, even assuming that we slow down to a doubling of transistors every four years, where will that growth come from?

GM: I think we could just make bigger chips. We could put more stuff onto bigger chips. There are a lot of continuing opportunities there. Once you give the engineers a billion-transistor budget and tell them to design something useful, they’ve got a lot of flexibility.

The x86 architecture is now about 25 years old. Did you expect it to last this long?

GM: It’s really evolved considerably over that 25 years. We’ve added a lot of stuff. We took advantage of essentially all the new inventions of the computer architects and academic communities. It carries some baggage, but the baggage is not that bad. That baggage lets us run all the historical applications, which is really an important thing.

I think it’s going to be around for a long, long time. It’s hard for me to envision circumstances that would be appropriate to abandon it right now. If somebody made completely hardware-independent software, maybe. But then you’d still have to go back and take the tens of thousands of programs, the legacy stuff, and convert them. I don’t think that’s going to happen. I think the compatibility is such a powerful asset for the typical user, it’s going to keep the Intel architecture around for the foreseeable future.

Its success seems to be, in part, holding back Intel’s own Itanium processor.

GM: They are really aimed at different markets. Itanium really doesn’t depend much on legacy software, and it really is a big-machine-oriented architecture. It may find its way onto the desktop too, If people really want to go that way, but I’m a little skeptical. We’ll have to wait and see.

How much longer will x86 be around? Do you think it will last another 25 years?

GM: It will still be around when we dig up the time capsule.