IRDS Picks Up Roadmap Baton
By David Lammers
The ITRS semiconductor roadmap has morphed into the International Roadmap for Devices and Systems (IRDS), with a different, more system - and application-focused purpose than the scaling-oriented ITRS.
In most competitive environments, there is a balance between the desire to strengthen the overall industry by sharing, and the need to keep things confidential. That tug-of-war is doubly true for the semiconductor industry and has presented challenges for the pioneering International Technology Roadmap for Semiconductors (ITRS) since its inception in 1998.
Steve Moffatt has had a front row seat since the establishment of the first National Technology Roadmap for semiconductors in 1986, predating the founding of the global ITRS. As a young engineer, Moffatt was assigned by Applied Materials CEO Jim Morgan and then-COO Jim Bagley to interface with the fledgling roadmap group. Moffatt, now the CTO for Front End Products at Applied, is still playing an active role in the latest evolution of the ITRS, the IEEE-affilliated International Roadmap for Devices and Systems (IRDS).
By closely coupling its activities to the approximately two-year technology transitions that form the basis of Moore’s Law, the ITRS “really was the engine, the boiler room, of the industry,” Moffatt said. More than 2,000 people, representing chip makers and their key supply chain partners, organized into 15 to 20 technology working groups (TWGs) that met regularly to discuss what technologies would be needed to advance to the next node.
The ITRS, founded in 1993, closely coupled its activities to the approximately two-year technology transitions that form the basis of Moore’s Law.
“From the people involved with building fabs to those supplying deionized water, basically all of them used the roadmap as a guide to thinking and planning,” Moffatt said. “However,” he added, “after roughly two decades of effectively describing transistor scaling challenges—including the infamous ‘red brick wall’ problems with ‘no known solutions’—the roadmap hit its own brick wall.”
Dan Hutcheson, CEO of market research firm VLSIresearch (Santa Clara, California), said that during the final years of the ITRS the largest chip companies became less willing to sit down in the committee meetings and talk openly about what they were doing. Largely a product of consolidation, this reticence gradually worsened, as only four or five memory companies and a similarly small number of logic companies came to dominate leading-edge processes.
“It all became very insular in terms of technology, where people certainly didn’t want to share at an open committee meeting. The number of people participating became fewer, both in terms of their level within their organizations and the overall headcount, and the roadmap became overwhelmed by academics,” Hutcheson said. He recalled one top executive asking, "Why are we spending all this money on a roadmap that we don’t use?’ And because there were so many consortia doing duplicative research, a lot of companies dropped out of Sematech. That was kind of the end of things for both.”
Hutcheson said while the semiconductor industry executives he has spoken with remain skeptical of the role of the IRDS, he pointed out that the new roadmap can play a key role in coordinating between the myriad standards bodies in the industry, including the standards activities within the IEEE.
Led by the IEEE
The IEEE Computer Society, the sponsoring committee of the new IRDS, had its initial meeting earlier this year in Leuven, Belgium. Paolo Gargini, who headed up the ITRS, continues in a similar role at the IRDS, while Thomas Conte, who served as the 2015 president of the IEEE Computer Society, is now serving as head of the Applications Benchmarking International TWG.
The IEEE describes the IRDS as “a comprehensive, end-to-end view of the computing ecosystem, including devices, components, systems, architecture, and software” and explains that “the IRDS will identify key trends related to devices, systems, and all the related technologies by generating a roadmap with a 15 year horizon.”
The IEEE Computer Society is the sponsoring committee for the new IRDS initiative.
Moffatt said that while the ITRS was largely driven by focused “bottom-up” activities defined by Moore’s Law transistor doubling, the new roadmap is characterized by a more balanced “top-driven” paradigm, with companies such as Apple, Facebook, Qualcomm and others playing major roles in setting the industry’s direction.
“It feels like the industry is fragmented, but it is fragmented because technology is everywhere. Big data in the cloud, just to take one example, is impacting so many more things. How we deal with memory in the cloud, or put photonic interconnections right up close to the chip, those kinds of things will determine how things go, and we don’t have collaborative efforts to look after them,” Moffatt said.
He also said that the emergence of augmented reality, virtual reality and artificial intelligence are examples of applications that will drive process, design and integration activities, adding that “the IRDS has been formed to address the business from the application end rather than the wafer end.”
While broadening the view to a top-down examination of applications and systems, the new roadmap will continue to coordinate bottom-up activities. More than 50 people, for example, are participating in the Factory Integration TWG of the new IRDS.
And Moffatt is careful when saying that scaling in various forms will continue, with areal scaling continuing at least until the 5nm node, and performance and power improvements continuing as new materials and device-types are brought in. The success of vertical NAND, for example, indicates that logic devices also may scale in the vertical dimension, with layer upon layer of transistors in 3D monolithic integration.
“These advancements are a fascinating story of evolution in our industry requiring the best in materials engineering. We are shifting from strong reliance upon traditional litho-based scaling to materials-enabled scaling and new architectures,” said Moffatt. “For example, industry leaders have already transitioned from 2D to 3D scaling with the adoption of FinFET and 3D NAND architectures, both of which require significant expertise in materials engineering.”
But much work remains to be done, Moffatt said, including attracting more participants from outside the US. “We need to stress that we intend to make the IRDS an international roadmap, even though in its initiation it is primarily a US-driven idea. Certainly it means more when it is the global industrial map.”
For additional information, contact firstname.lastname@example.org
Steve Moffatt, PhD, is Chief Technical Officer for Front End Products at Applied Materials. A 34-year veteran of the company, he holds numerous patents for semiconductor technologies and has served with the ITRS roadmap organization since 1986.