Europe and Singapore are two sources of ideas for GF following the decision to pivot away from 7nm technology development. One company that sees strategic value in the 22FDX® solution is Synaptics.
While Tom Caulfield was trying to figure out how to re-position GLOBALFOUNDRIES as a differentiated foundry, Rick Bergman, the CEO of Synaptics, had concluded that GF’s 22FDX process could do exactly that: differentiate Synaptics in the market for artificial intelligence-enabled Internet of Things applications.
“For the very large customers that we have, we are banking on 22FDX for unique solutions specifically for the IoT market,” Bergman said in a keynote address at the foundry’s annual technology conference, GTC 2018, in Santa Clara.
Synaptics, with revenues of roughly $2 billion this year, focuses on the human-machine interface (HMI), a market that is rapidly moving to voice-enabled interfaces. On-chip neural network processing in IoT edge devices is a key enabler, requiring the right mix of performance, power consumption, and cost.
Bergman said one Synaptics chip has taped out in the fully depleted silicon-on-insulator (FD-SOI) 22FDX process, another is “right behind it, aimed at voice and video,” while a third chip will support augmented reality and virtual reality (AR/VR) capabilities.
The company evaluated 28nm bulk transistors and found that “they didn’t have the horsepower required,” he said, while advanced FinFET-based leading-edge processes “require a certain amount of (design) investment, and in many cases the volumes don’t justify it in a fragmented market like IoT.” GF’s non-volatile memory solutions were another factor, he added.
Edge IoT devices increasingly are being used to process AI workloads, rather than sending them to the cloud. For IoT edge solutions, Bergman said Synaptics requires “extremely low power,” using FDX’s forward and back-biasing capabilities to switch to high performance, when needed, for products such as smart speakers.
“With biasing, we can gain performance when we need it, and when we don’t need it we have very low power. Since IoT tends to be a very competitive market, cost is an important factor as well,” he said.
After being named CEO of GF in early March, Caulfield began visiting customers, including the major European semiconductor companies. These companies had looked at the rising costs of leading-edge designs, the slowing improvements to Moore’s Law scaling, and made their own pivots.
“They learned that leading edge is not the only game in town, that innovation is shifting toward the creation of differentiated features. They have prospered as a result. I came away from there knowing what we had to do to differentiate ourselves. And I knew we had to get out of our mind the idea that we had to have leading-edge technology to be strategic to these customers,” Caulfield said.
Caulfield said he wants the larger GF to learn from the foundry’s Singapore operation, which has mastered the skills needed to ensure zero defects, provide improvements to its technology platforms, and hone manufacturing discipline: achieving high yields on a wide variety of products while keeping the fabs full.
“Much of our revenue in differentiated silicon comes from Singapore,” he said, adding that “the revenue model for GF is Singapore.”
He made the point that many people are fixated on leading-edge technologies, currently 7nm and beyond, with relatively little attention given to the many advances in power, MEMs, RF, and other technologies. The heart of the day-long GTC 2018 was spent describing the innovations which differentiate GF in those areas. GF’s technologists are coming up with ways to boost the performance of the 12nm FinFET process, adding options such as high-voltage and NVM offerings. By halting 7nm logic development, executives promised that more of the foundry’s resources will be available for power, analog, IoT, automotive, and other markets which have their own technology differentiations.
During a coffee break, I asked Subramani Kengeri, vice president of client solutions, if the pivot away from 7nm might leave GF fighting for business with foundries such as Taiwan’s UMC or China’s SMIC.
Kengeri said GF differs from its competitors in three ways: first, the breadth of offering and level of technology differentiation is much greater. In automotive, eNVM, HV, mixed-signal, FinFET, Silicon Photonics and several other key areas, GF simply offers a much richer and more complete technology portfolio than other foundries. With its FDX technology, it offers a unique solution. In RF it dominates. “Thinking RF? Think GF,” was a clever refrain heard from Bami Bastani, senior vice president of GF’s business units.
Secondly, Kengeri said GF has superior packaging technologies, and going forward will have more resources to invest in packaging to support growing needs for heterogeneous integration. And thirdly, GF — having acquired Chartered Semiconductor and IBM Microelectronics and grown to $6 billion in revenues — is a more global company than its competitors, with fabs in United States, Germany, and Singapore, and a strong global client solutions team.
Caulfield said with the pivot decision behind it, GF is focused on improving manufacturing efficiencies and boosting fab utilization rates, while adding capacity where needed in RF, power, automotive, and other fast-growing markets. “We want to fill the assets that we have,” he said.