By 2040, it has been projected that computer chips will need more electricity than global energy production can deliver. The Semiconductor Industry Association (SIA) confirmed that earlier projection with the release of the new 2015 International Technology Roadmap for Semiconductors (ITRS), a collaborative report that surveys the technological challenges and opportunities for the semiconductor industry through 2030. The ITRS seeks to identify future technical obstacles and shortfalls, so the industry and research community can collaborate effectively to overcome them and build the next generation of semiconductors – the enabling technology of modern electronics. The current report marks the final installment of the ITRS.
“For a quarter-century, the Roadmap has been an important guidepost for evaluating and advancing semiconductor innovation,†said John Neuffer, president and CEO, Semiconductor Industry Association. “The latest and final installment provides key findings about the future of semiconductor technology and serves as a useful bridge to the next wave of semiconductor research initiatives.â€
Faced with ever-evolving research needs and technology challenges, industry leaders have decided to conclude the ITRS and transition to new ways to advance semiconductor research and bring about the next generation of semiconductor innovations. While the final ITRS report charts a path for existing technology research, additional research is needed as we transition to an even more connected world, enabled by innovations like the Internet of Things. Some of these technology challenges were outlined in a recent SIA-Semiconductor Research Corporation (SRC) report, “Rebooting the IT Revolution,†but work continues to define research gaps and implement new research programs.
“SIA appreciates the hard work, dedication, and expertise of those involved with the ITRS over the years and looks forward to continuing the industry’s work to strengthen semiconductor research and maintain the pipeline of semiconductor innovations that fuel the digital economy,†Neuffer said.
The projection could mean that our ability to keep pace with Moore's Law – the idea that the number of transistors in an integrated circuit doubles approximately every two years – is about to slide out of our grasp.
In current mainstream systems, the lower-edge system-level energy per one bit transition is about 10^-14 J, which is referred as the “benchmark†in the above figure. For this benchmark energy per bit, computing will not be sustainable by 2040, when the energy required for computing will exceed the estimated world’s energy production. Thus, radical improvement in the energy efficiency of computing is needed. The physics-based device-level theoretical lower limit is 3*10^-21 J/bit (known as the Landauer limit for binary switching), and a practical lower limit for a system-level energy consumption can be estimated, based on the analysis in43, to be about 10^-17 J/bit, which is referred to as the “target†in the figure.
Energy-Efficient Sensing and Computing Greater energy efficiency is mandatory in order to fully realize future IT capabilities. Without progress, power demands at the transistor, chip, and system levels will become prohibitive. Energy efficiency is vital at all levels—from the smallest sensor to ultrahigh performance processors and systems. Industry’s ability to follow Moore’s Law has led to smaller transistors but greater power density and associated thermal management issues. More transistors per chip mean more interconnects— leading-edge microprocessors can have several kilometers of total interconnect length.
But as interconnects shrink they become more inefficient. Today more than half of the energy used by a processor is to move data via interconnects, e.g., between memory and logic. Conventional approaches are running into physical limits. Reducing the “energy cost†of managing data on-chip requires coordinated research in new materials, devices, and architectures. Whereas in the past these topics have been addressed separately, meaningful progress will require a multidisciplinary and coordinated approach.
Energy efficient sensors are essential to the IoT. Distributed sensors with significant on-site “intelligence†and communications ability that allow decisions and actuation to be handled locally will reduce the energy needed to manage the anticipated volume of data. Remotely located sensors will need to be securely accessible for calibration and validation checks, and if they are stand-alone for long durations, they must have mechanisms for efficient energy collection and utilization. Moreover, intelligent sensor devices could be networked to perform collective tasks, creating distributed information computing and storage capacity on demand and further reducing energy consumption.
Fundamentally, there must be a new conceptual model and new supporting architecture to enable processing the vast Internet data flows and to enable extracting and delivering timely insights. This new technology and architecture needs to be several orders of magnitude more energy efficient than best current estimates for mainstream digital semiconductor technology if energy consumption is to be prevented from following an explosive growth curve.
Additionally, there is a critical need for miniature energy sources (both batteries and harvesting) with intelligent power management. In order to minimize the volume of transmitted data, sufficiently high local nonvolatile memory density will be needed at the sensor node.15
The core technology that is in high performance computers and large data centers also must be made more energy efficient if more capable systems are to be built, e.g., beyond exascale computing. The most cost-effective solution would be a single technology that could be implemented in high performance, mobile, and embedded sensor systems. However, optimal solutions in such diverse applications likely will take different approaches. Therefore, multiple technologies should be further explored.
Focus areas for research should include extremely low power devices for sensors and advanced processors, novel energy-harvesting devices, novel circuit architectures that reduce data movement, explorations to overcome physical limits due to noise fluctuations without significant power trade off, and nanophotonics and integration of photonics on-chip. Next-generation electronic design automation (EDA) tools are needed, including improved models for sensors, processors, and systems, and validation techniques that measure the effectiveness of energy reduction efforts. At the system level, research is needed to build intelligence and energy efficiency into the IoT at all levels.
