Moore's law has ended. What comes next?

The speed of our technology doubles every year, right? Not anymore.
Credit: Carnegie Mellon University, Department of Civil and Environmental Engineering

The speed of our technology doubles every year, right? Not anymore.

We've come to take for granted that as the years go on, computing gets faster, cheaper and more energy-efficient.

In their recent paper, "Science and research policy at the end of Moore's law" published in Nature Electronics, however, Carnegie Mellon University researchers Hassan Khan, David Hounshell, and Erica Fuchs argue that future advancement in microprocessors faces new and unprecedented challenges.

In 1965 R&D Director at Fairchild (and later Intel co-founder) Gordon Moore predicted continued systemic declines in cost and increase in performance of integrated circuits in his paper "Cramming more components onto ." This trend, later coined Moore's Law, held for over 40 years, making possible a "dazzling array" of new products from intercontinental ballistic missiles to global environmental monitoring systems and from smart phones to medical implants.

"Half of economic growth in the U.S. and world-wide has also been attributed to this trend and the innovations it enabled throughout the economy," says Engineering and Public Policy Professor Erica Fuchs.

In the seven decades following the invention of the transistor at Bell Labs, warnings about impending limits to miniaturization and the corresponding slow down of Moore's Law have come regularly from industry observers and academic researchers. Despite these warnings, continually progressed along the Moore's Law trajectory. Khan, Hounshell, and Fuchs' archival work and oral histories, however, make clear that times are changing.

"The current technological and structural challenges facing the industry are unprecedented and undermine the incentives for continued collective action in research and development," the authors state in the paper, "which has underpinned the last 50 years of transformational worldwide and social advance."

As the authors explain in their paper, progress in semiconductor technology is undergoing a seismic shift driven by changes in the underlying technology and product-end markets. Achieving continued performance improvements through transistor miniaturization has grown increasingly expensive and the emergence of new end-markets has driven innovation in more specialized domains. As such, in recent years there has been a splintering of technology trajectories, such that the entire industry moving in lock-step to Moore's Law is no longer of economic benefit to all firms. Examples in the paper include search companies (such as Microsoft Bing) using field-programmable gate-arrays in data centers as accelerators in conjunction with CPUs, and Google's announcement of proprietary 'tensor-processing unit' chips developed in-house for its deep-learning activities.

"While these innovations will drive many domain-specific advances, to continue advancing general purpose computing capabilities at reduced cost with economy-wide benefits will likely require entirely new semiconductor process and device technology." explains Engineering and Public Policy graduate Hassan Khan. "The underlying science for this technology is as of yet unknown, and will require significant research funds – an order of magnitude more than is being invested today."

The authors conclude by arguing that the lack of private incentives creates a case for greatly increased public funding and the need for leadership beyond traditional stakeholders. They suggest that funding is needed of $600 million dollars per year with 90% of those funds from public research dollars, and the rest most likely from defense agencies.

In terms of allocating those funds, they argue for pursuing two avenues in parallel: A research institute with academics and government program managers as key players, that also engages industry players across the computing technology stack; coupled with a semi-coordinated government funding effort focused on the next generation of transistor technology across all government agencies, such as was undertaken in the case of the National Nanotechnology Initiative, in which key program managers met weekly to discuss initiatives and share insights.


Explore further

Intel's Breakthrough in High-K Gate Dielectric Drives Moore's Law Well into the Future

More information: Hassan N. Khan et al. Science and research policy at the end of Moore's law, Nature Electronics (2017). DOI: 10.1038/s41928-017-0005-9
Provided by Carnegie Mellon University, Department of Civil and Environmental Engineering
Citation: Moore's law has ended. What comes next? (2018, February 13) retrieved 20 November 2018 from https://techxplore.com/news/2018-02-law.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
3641 shares

Feedback to editors

User comments

Feb 13, 2018
Randall's Law - the total amount of energy used for computing doubles every year.

Feb 13, 2018
Quantum patent for densities of Yottawords cubic inch (in3) and Exawords read bandwidths upward being realized.

Feb 13, 2018
And for us mere mortals, multi-core processors, under-clocked for economy...

Feb 13, 2018
I think people underestimate the amount of implementation there hasn't been time for, and more that hasn't even been thought of, at any given stage of development.

Feb 14, 2018
Quantum patent for densities of Yottawords cubic inch (in3) and Exawords read bandwidths upward being realized.

Does this mean anything?
I cannot imagine what.

Feb 14, 2018
Computing power will continue to increase exponentially for many more years in the form of parallel processing and neural networks.

Feb 14, 2018
"In 1965 R&D Director at Fairchild (and later Intel co-founder) Gordon Moore predicted continued systemic declines in cost and increase in performance of integrated circuits"


BZZZT - Wrong. This is yet another one in a long line of articles that gets the Moore's Law completely wrong. Moore never said anything about performance: only about the number of transistors on a chip at the most cost-effective point for manufacturing.

Moore's law ended 20 years ago. It just took this long for the marketing departments to run out of excuses. Intel added the performance metric to Moore's law after they couldn't re-double the number of trasistors on their chips around mid-90's.

But that's no longer Moore's Law.

Feb 14, 2018
"warnings about impending limits to miniaturization and the corresponding slow down of Moore's Law have come regularly from industry observers and academic researchers. Despite these warnings, semiconductor technology continually progressed along the Moore's Law trajectory."


That's because they kept shifting the goalposts. The original Moore's Law can be stated as number of transistors per area per dollar. When that started slowing down, they changed it to "number of transistors per area per dollar per megahertz" or some other metric like "instructions per second"... basically since they couldn't cram any more transistors on the chip without making the chip more expensive and/or running too hot. The economics of making bigger denser chips with more transistors started lagging behind the curve.

So Intel re-defined "Moore's" law just in time to keep up with their own promises of keeping up with the Moore's law.

Feb 19, 2018
so far and no further

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more