If you have a passing familiarity with computers, you probably know about Moore’s Law. First promulgated during the 1960s by Gordon E. Moore (co-founder of Intel), Moore’s Law was a prediction that the density of transistors on a chip would double once every twelve months. Initially, this was simply an observation and extrapolation of industry trends. In fact, Gordon Moore only predicted that the doubling phenomenon would last for ten years. In 1975, he actually changed up his original formula, stating that transistor density would only double once every 24 months—again, with just a ten-year predictive horizon. However, this prediction stuck, and transistor density doubled about as regularly as clockwork. Now, according to a new article from Nature, those days are at an end.
The important thing to realize is that at some point in the 80s, the semiconductor industry stopped thinking of Moore’s Law as some kind of industry yardstick, and started thinking of it like some kind of dare. Instead of conforming with Moore’s Law more-or-less by accident, as they were wont to do, they began setting research and production targets with the goal of hitting that two-year target. There’s actually an organization, the International Technology Roadmap for Semiconductors (ITRS), which was initially founded with the sole goal of keeping Moore’s Law on track. They’ve suspended their adherence to the Law as of 2014, and now more and more chipmakers have followed suit.
Why’s everything stopping all of a sudden? For an answer, first take a look at this giant chunky thing in the picture above.
What you’re looking at is a replica of the very first transistor ever, made in 1947. It takes up a few cubic inches of space.
Below, take a look at an Intel 8080 chip from 1974. It’s about an inch long, and contains 6,000 transistors.
So, the size of each transistor in the chip above is 6 micrometers, or six thousand nanometers. However, the size of each transistor in a chip from 2014 is 14 nanometers. If a 1974 transistor is a bowling ball, a 2014 transistor is a ball bearing. Better yet: transistors are built on a silicon substrate. A 14 nm chip is only the size of 90 silicon atoms. Quite simply, we are running up against the physical constraints of our material universe.
What does the end of Moore’s Law mean for the future of computing? Nothing particularly bleak. While chip density has of course been improving, processing speeds haven’t quite been keeping up. Manufacturers will most likely work on improving the efficiency of existing designs, allowing them to achieve higher clock speeds. There’s also the question of functionality. As the Internet of Things (ugh) gains popularity, we’ll likely see a plethora of designs that include most of the functionality of a cell phone (antenna, WiFi, RAM, etc) on a single chip.
Lastly, major chip manufacturers are also experimenting with different chip substrates that lack silicon’s limitations. We’ll probably see several more doublings of chip density as time goes on–just, not within the next two years.