The end of Moore's Law may ultimately be as much about economics as physics, says a DARPA director.
"My thesis here is that it's time to start planning for the end of Moore's Law, and that it's worth pondering how it will end, not just when," Robert Colwell, director of the Microsystems Technology Office at the Defense Advanced Research Projects Agency, told CNET.
Colwell gave a presentation at the Hot Chips Conference at Stanford University on Monday titled "The Chip Design Game at the End of Moore's Law." On Tuesday, he provided further comment to CNET.
"The silicon business is incredibly expensive for folks like Intel, who have to pay huge amounts of money to develop the next-generation silicon technology," he said.
How expensive? The new Globalfoundries chip plant in upstate New York was a $6 billion undertaking, and the capital expense will rise to $8 billion when a new R&D facility is added to it, with that work expected to finish up in late 2014.
Moore's Law, named after Intel co-founder Gordon Moore, states that the number of transistors that can be placed on an integrated circuit doubles roughly every two years. For decades, chipmakers have succeeded in shrinking chip geometries, allowing Moore's Law to remain on track and consumers to get their hands on ever more powerful laptops, tablets, and smartphones.
Below are the comments he provided to CNET based on his speech at the Hot Chips conference:
[It takes] huge amounts to build the fab plants, and yet more...to pay for the design teams to design new chips. Intel makes these investments, which are in the [billions of dollars], because they expect to reap way more [billions of dollars] in profits in the following years.
But if there is doubt that those profits will arrive, and possibly if they just doubt they can come up with the necessary silicon improvements, they may not want to make the investment at all. Should a major player like Intel make such a call, that would effectively end Moore's Law all by itself, because then the various companies that make the super expensive tools for chip production will themselves not make the investments needed to keep Moore's Law alive.
Expanding on the comments above, Colwell said companies need to focus on the economics of getting to 7 nanometers and beyond:
I think most technologists will, at least privately, rationally consider the prospects for the end of Moore's Law. But they focus on the physics, which change substantially from one silicon process technology to the next. These technologists know how monumental the challenges are to even get to [7 nanometer] silicon technology, so the natural assumption is that eventually some problem with the next silicon process technology will turn out to be physically insurmountable, and that's what will end Moore's Law.
They are right, there are very serious challenges looming, but then, the silicon industry has beaten back such challenges in the past. This gives a lot of people false confidence that we can do it again. My attitude is, maybe we can, maybe we can't, but physics isn't the only challenge. That was my point in focusing on economics, to remind the field that we have to succeed at the physics AND the economics or we will have failed, and Moore's Law will have ended.
Colwell said that for the Defense Department, he uses the year 2020 and 7 nanometers as the "last process technology node."
But he adds, "In reality, I expect the industry to do whatever heavy lifting is needed to push to 5nm, even if 5nm doesn't offer much advantage over 7, and that moves the earliest end to 2022. I think the end comes right around those nodes."
Intel has already begun discussing commercial 10-nanometer technology, which is expected to ramp up in 2015.
But Intel component scientists have also said that it's not clear what technology will be used beyond 7 nanometers.
Currently, Intel processors, such as its Ivy Bridge and Haswell processors, are made on a 22-nanometer process.