Decades after the concept of embedding programmability into ASIC devices was first imagined, a number of industry trends have converged to finally make it economically practical to implement. On the supply side, the most fundamental of those trends is that mask costs have gone up dramatically with each new generation of semiconductor process technology, while the cost of transistors (and thus, gates) has continued to fall. On the demand side, the emergence of the IoT market — with its associated fragmented application space and low cost and power requirements — has engendered a need for highly flexible devices that have a low per-unit cost and consume little power. This need has been partially met through the use of embedded processors, but their power consumption versus performance trade-off makes them a less-than-ideal solution for IoT edge applications. Let's look at each of these trends in a little more detail.
Programmability, in the form of FPGA gates, was prohibitively expensive for SoC applications a few decades ago. Credible industry estimates are that, early on, programmable gates cost as much as 50 times the amount of an equivalent level of “hard-wired” ASIC gates. At the same time, ASIC mask costs were multiple orders of magnitude lower than they are today (tens of thousands of dollars instead of tens of millions of dollars), making the cost of mask spins not too terrible. Add to the equation the fact that early programmable technologies often had unique (and therefore expensive) processing requirements, and the total cost/benefit analysis always showed a significant disadvantage for adding programmable logic to ASICs.
The consumer semiconductor market over the last ten years or so has largely been driven by highly standardized mobile devices such as tablets and smartphones. These devices often share common platforms with common architectures, and so don't require a high degree of flexibility from the SoCs that service them. The emergence of the IoT as the next wave of growth for consumer electronics comes with a demand for higher device flexibility, as these applications are much more fragmented and difficult to predict. They also have a requirement for devices with a lower per-unit cost and level of power consumption. Nobody wants to pay too much or be constantly changing batteries for the constellation of IoT devices that's sure to inhabit our collective future.
It could be argued that the “flexibility” problem for SoCs has largely been solved using embedded processors. It's certainly true that this approach has created extremely flexible devices with the highly desirable attribute of being software programmable. Unfortunately, however, processors can consume as much as five to ten times the power of hardware-based solutions when these are possible. High power consumption is a big problem for most IoT applications — especially those at the very edge of the network.
The logical solution is embedded FPGA (eFPGA) technology. The cost of programmable logic transistors and gates has fallen to the point whereby they can be added to an SoC without a noticeable increase in die size. In fact, pad-limited devices can effectively get them for free. Having hardware programmability allows SoC developers to address a broad range of IoT application needs without having to create separate devices for each one. It also allows designs to tackle problems that previously required a processor (or a larger processor) while reducing power and potentially even increasing performance. Emerging and/or rapidly evolving market requirements can easily be addressed, thereby eliminating the need to predict the functionality requirements for a device that can only be economically iterated every few years.
It has been a long time coming, but eFPGA technology has finally come of age.
Timothy Saxe (Ph.D) has served as QuickLogic's Senior Vice President and Chief Technology Officer since November 2008. Dr. Saxe has been with QuickLogic since May 2001. Prior to QuickLogic, he was Vice President of FLASH Engineering at Actel Corporation, a semiconductor manufacturing company. Dr. Saxe joined GateField Corporation, formerly known as Zycad, in June 1983, and was a founder of their semiconductor manufacturing division. He became GateField's CEO in February 1999, and served in that capacity until GateField was acquired by Actel. Dr. Saxe holds a B.S.E.E. degree from North Carolina State University, and an M.S.E.E. degree and a Ph.D. in electrical engineering from Stanford University.
Originally posted on Embedded's sister site, EE Times: “Embedded programmability (finally) comes of age.”