Fort Worth, TX. The semiconductor industry finds itself in uncharted waters, said Ken Hansen, CEO, Semiconductor Research Corp., in the concluding keynote address of the International Test Conference. There are more unknowns and uncertainties than ever, but that presents engineers with the opportunity to be their most creative. Consequently, it’s “…the most exciting time ever in semiconductor industry,” he said.
The current situation, however, require investment in research. He advised doubling down on research through the SRC, which takes a consortial approach to innovation. He noted all keynoters at this year’s ITC have been involved with SRC. For example, Walden C. Rhines, chairman and CEO, Mentor Graphics, delivered a Tuesday keynote address titled “The Business of Test: Test and Semiconductor Economics,” and on Wednesday, Rob A. Rutenbar, Bliss Professor and Head Department of Computer Science University of Illinois at Urbana-Champaign, delivered a talk titled “Hardware Inference Accelerators for Machine Learning.”
He cited some figures suggesting challenges for the industry. CAGR has been 2.2% over the last five years compared with 9.5% over the last 20 years. 2016 is predicted to see a small decline in sales, making it only the second time sales have declined two years in a row; the first was 2008-2009.
Nevertheless, he said, “I have a lot of optimism for the future.” He predicts an insatiable demand for data collected at the edge, with the IoT being a clear market driver. Applications include efficient generation, distribution, and use of power; independent living for the aging; secure financial services; autonomous transportation; and affordable medical care. He cited estimates that in 2020, the HPC market will be worth $36.62 billion, the mHealth market, $59.15 billion, and the IoT market, $1.7 trillion.
But success in these areas, he added, will require fundamentally new technologies as Moore’s Law comes to an end. He noted that since the diameter of a silicon atom is on the order of 0.25 nm, traditional scaling won’t work, and semiconductor devices are increasingly exhibiting stochastic, statistical behavior.
He cited several areas for research: the interface to real world with enhanced HCI; sensor proliferation; security; package integration (with fanout and 3D approaches); architecture exploration (with non-von Neumann compute approaches, accelerators, and new algorithms that exploit approximations and randomness); and EDA tool optimization for power, die size and performance.
Throughout, he said, we must invent and build technologies that enable us to handle simultaneous explosion in collecting data at the edge as well as communication throughout gateways and up to the cloud where the data can be stored, analyzed, and converted to actionable information.
He said he has heard people refer to data as the new oil, but he prefers to consider it the new gold—you have to sift through the nuggets to find what’s valuable.
He cited a report titled “Rebooting the IT Revolution: A Call To Action” released in September 2015 in conjunction with the Semiconductor Industry Association. The report calls for fundamental research in the following areas in order to fully realize IoT breakthroughs and sustain America’s technology leadership: energy-efficient sensing and computing, data storage, real-time communication ecosystem, multi-level and scalable security, a new fabrication paradigm, and insight computing.
In his ITC address, Hansen called for a new research approach based on co-optimization with inter-science research. Interesting things, he said, happen at the boundaries of chemistry, physics computer science, mechanical engineering, and electrical engineering.
He described the work of the multi-university initiative SONIC, for Systems On Nanoscale Information fabriCs. The organization takes a systems-driven approach including bio-inspired and Shannon-inspired investigations. Bio-inspired research aims to mimic the brain, which affords a noisy, unreliable environment for computing but achieves amazing performance for the amount of power consumed. Shannon-inspired approaches employ statistical information processing using unreliable components. To address platform energy challenge, he said, SONIC is investigating in-sensor computing and in-memory computing on energy-constrained stochastic fabrics.
With regard to cybersecurity, he said anything attached to the network can be hacked, adding attacks at the hardware level are most dangerous, but most investment in at the software levels. SRC, he said, is focusing research on hardware security, exploring concepts like the chip odometer.
Hansen observed that a new device comes along every half century or so, citing the vacuum tube in the late 19 and early 20th century and the transistor and MOSFET in the 1947 to 1960 timeframe.
What’s next? He posited a steep-slope transistor-like device or a hysteretic memory-like device. New device comments may be investigated under the forthcoming Joint University Microelectronics Program (JUMP), which will commence in January 2018 and is expected to be the catalyst to enable the technology roadmap for 2025 and beyond.
He noted that the test problem has been considered NP hard, and it will continue to be. Data mining of customer returns will be necessary to improve yield, he said, and test engineers, he said, will need to contend with stochastic devices, new architectures, new package-integration technologies, security issues, new devices, and Shannon- and brain-inspired information processing. He concluded, “There is more and more excitement for test than there has ever been.”
