As most defense technologies are shifting to an open systems architecture, the director of the Defense Advanced Research Projects Agency (DARPA) said computers and microprocessors may only become more advanced if they switch to specialized architectures.
Arati Prabhakar, who launched DARPA’s Microelectronics Technology Office and spent years in Silicon Valley before taking over as DARPA’s director in 2012, said that for decades computing power has been driven by Moore’s Law: the idea that the number of transistors in a dense integrated circuit doubles approximately every two years.
But, she said Nov. 19 at the Defense One Summit, she believes Moore’s Law has already begun coming to an end. Defense systems in particular can find ways to be more efficient with the transistors they do have, so the progress in computing power and performance will not hit a brick wall, she said.
Still, DARPA is researching “what are the things we need to now be investing in in research that will give us ways to continue to advance when all of that [efficiency] has been squeezed out and we no longer have Moore’s law as a driver,” Prabhakar said.
Specialized architectures may be a key to continuing improvement.
“A lot of the magic of Moore’s law was that general-purpose computers and microprocessors just continued to get better and better and better,” Prabhakar said. “The magic of that was that that was a computational unit you could use to do the broadest possible class of problems. So now it turns out that if you’re willing to work on special classes of problems, you can actually get a lot more from specialized architectures, and there happen to be classes of problems that we care a lot about in DoD.”
She mentioned image processing as an example, saying that specialized architectures coupled with new algorithms will help the military and defense industry continue to improve its intelligence technologies.
“If you look at what Moore’s law has given us in the last decades–if you look at the power of microelectronics, the [radio frequency] systems that have also come out of that same technology base, the micromechanical systems, instead of doing precision navigation and timing with these big complex very exquisite systems many of those are collapsed into smaller systems–if you add to all those physical technologies what’s gone on in software computing and algorithms, those are the ingredients that I think are going to allow this new generation of technology,” she said.