As transistors reach the limits of miniaturization, it is only a matter of time until Moore’s Law runs out of steam. The latest expert to weigh-in says Moore’s Law will expire in 2020 at the 7nm node.
The prediction was made by Robert Colwell, director of the microsystems group at Defense Advanced Research Projects Agency (DARPA), during the Hot Chips conference.
“For planning horizons, I pick 2020 as the earliest date we could call it dead. You could talk me into 2022, but whether it will come at 7 or 5nm, it’s a big deal,” Colwell observed, as quoted in EETimes piece.
The doubling of transistor density every 18-24 months was a boon to the entire computing industry, one that Colwell is already mourning. These exponentials have “unsustainable heady growth” and “such rides are rare” in Colwell’s opinion.
The observation made by Gordon Moore described a growth factor that boosted speeds from 1 MHz to 5Ghz over a 30 year span – a 3,500-fold improvement. Architectural innovations during the same timeframe delivered only about a 50x increase by comparison.
As Gordon E. Moore himself said in 2005: “It can’t continue forever. The nature of exponentials is that you push them out and eventually disaster happens.”
DARPA has its eye on lots of alternative technologies, but Colwell is dismissive of the blind faith in an as-good-or-better replacement, saying that “you can’t fix the loss of an exponential.”
The list of contenders to maintain the pace of progress post-CMOS includes 3D stacking, and new architectures and switching technologies, as well as better human interfaces and even creative marketing, according to Colwell. DARPA is also investing in approximate computing and the use of spin-torque oscillators.
While it’s been proposed that single-atom transistors could extend Moore’s Law, Colwell believes that economics not physics will signal the final death knell. “So keep your eye on the money,” he counsels.