Posted by Joe Mariani
For more than 50 years, computers have been getting faster and smaller at a regular, predictable rate. This process is known as Moore’s law. In 1965 Gordon Moore noted that every year the number of components on an integrated circuit doubled, largely as a function of the shrinking size of the transistors that made up the majority of the components in those circuits.i While recent chip launches have held true to the advances predicted by Moore’s law, the cost to produce those chips is beginning to increase to a point that could threaten further advances. Perhaps ironically, the next generation of super-fast chips may actually be thanks to a growing market for chips that do not need to be particularly fast at all.
Almost since it was first coined, detractors (including Moore himself, we might add) have been predicting the demise of Moore’s law.ii Chips simply cannot go on getting faster, and transistors smaller, forever. One problem is physics. As transistors get smaller and smaller, electrons can begin quantum tunneling through the gate of a transistor, losing power like gas leaking out of the tank in your car.
So far, clever engineering solutions have been found for these physics problems. With the release last year of chips with transistors measured at 14 nanometers by multiple manufacturers, Moore’s law seems to have survived yet another hurdle.iii However, the complex engineering needed to achieve these smaller and smaller sizes does increase cost and slow development. If that trend continues, it threatens Moore’s law itself, as the chips may become so costly and take so long to reach the market that there may be very little demand for them. Already, we have seen tech OEMs begin to hedge against these very challenges with strategies such as using either of two chips from two different manufacturers in the same device.iv
Aside from increasing costs of the highest performing chips, demand for those cutting-edge chips is also slowing.v With the growth of cloud computing, consumers can easily tap into the capabilities of large data centers, where racks of servers can share tasks previously done by one stand-alone machine. If that computing power no longer needs to fit into a single device, but rather into a many acre data center, you no longer need expensive, tiny chips but can get by with many, larger, cheaper alternatives.vi With consumer demand for small, high speed processors potentially slowing, how will semi-conductor manufacturers find the resources necessary to break the next set of barriers the laws of physics throw in their way?
This is where the Internet of Things (IoT) may come into play. IoT is fundamentally about capturing, processing, and acting on data from sensors set in the world. Given the often rugged and austere environments the endpoints of these IoT systems find themselves in, changing batteries and transmitting data can become quite expensive. This means that builders of IoT devices often value low power consumption and simple on-chip processing more than the small size or processing speed typically valued in microprocessors. Chips produced to meet these IoT needs are often larger and less dense than the cutting edge, high-speed chips that push the boundaries of physics and advance Moore’s Law. However, with more than 26 billion IoT devices expected by 2020, creating a high volume of these relatively low speed chips could provide a stable revenue stream for semiconductor manufacturers. That revenue could in turn support the R&D to continue Moore’s law and push the boundaries into 7 nanometers, 5 nanometers, and even smaller chips.vii
So, it could be IoT and its consistent appetite for relatively big, slow chips that help keep Moore’s law in effect for another few years.
|i Moore, Gordon E. (1965). “Cramming more components onto integrated circuits” (PDF).Electronics Magazine. p. 4. Retrieved 2006-11-11. Moore later revised his law to state that the doubling would occur every two years.|
|ii Ibid. Initial estimates were that this rate would continue only through the 1970s.|
|iii Transistor size is measured as the distance between the source and the drain. Both Intel and Samsung shipped devices containing 14nm transistors in 2014.|
|iv Barrett, Brian. “It Doesn’t Matter Which A9 Chip Your iPhone Has. Get Over It.” Wired Magazine. October 10, 2015. http://www.wired.com/2015/10/iphone-6s-a9-battery-life/|
|v Aruna, P. “Chip Sector Slowing Down.” The Star. October 24, 2015. http://www.thestar.com.my/business/business-news/2015/10/24/chips-sector-slowing-down/|
|vi The End of Moore’s Law. The Economist Explains series. The Economist. April 19, 2015.|
|vii Gartner -The Internet of Things Will Demand New Application Architectures, Skills and Tools, 4/1/2014|