When a friend told Mark Twain that some press reports said that he was dead, Twain replied, "The reports of my death are highly exaggerated." At a recent Semico Summit held in Arizona, a panel of SOC industry executives insisted that "SOC was not dead," and speculated on the product's future. I wondered what made them discuss whether SOC was dead or alive. I, for one, feel that SOC has not yet been born; it will come into being with the maturing of nano-technology, which will then enable integrating full functionalities into a single piece of silicon.
What passes for SOCs today - silicon chips with limited functionalities - represented US$13 billion of the US$141 billion semiconductor market in 2002. This shows that reports of SOC's death, whatever their source, are "highly exaggerated," much like the reports of Twain's "death."
The conceptual birth of SOCs is not a low-key affair either. Nanotechnology SOC was born conceptually in 1959 when Dr Richard Feynman predicted that products would be built one molecule or one atom at a time by 2000. Some called this concept hype; others said it was a hoax. The hope surrounding nanotechnology got the better of hype and hoax in 1990 when IBM engaged in an atom-by-atom assembly by placing 35 argon atoms in nickel to write IBM. SOC moved ahead of the conceptual plane.
For the last decade it has been a front-runner towards technological frontiers. What has suddenly changed that even as some designers believe that nanotechnology SOC is near birth, others wonder whether SOC is alive or dead? The industry is now coming to terms with the techno-commercial aspects of SOC: Commercial benefits through a further extension of Moore's Law is possible only if design-to-production processes are refined to an extent that was never possible throughout the entire industrial history of mankind.
The price for moving to nanotechnology SOC is phenomenal-billions of dollars. With this kind of money at stake, there is hardly any room for error. Ian Getreu, general chairperson of 40th Design Automation Conference, says that nanometer designs must be right the first time to ensure that the benefits of moving to this new technology outweighs commercial costs. "Success is not just meeting spec the first time," he enthuses, "it also requires having spec right the first time." Are we capable of achieving such first-time successes? With today's latest technologies, over 50 percent of tapeouts are wrong the first time, according to Aki Fujimura, general manager of design for manufacturing at Cadence. He adds that this figure is getting worse as we move deeper into the submicron range. So one can well imagine the immense challenges ahead for "first-time successes" for nanotechnology SOC.
Interestingly, the move towards 0.13 nm has so many stumbling blocks that there are reports, unconfirmed by foundries but confirmed by EDA vendors, that 40 to 50 percent of the produce goes waste for many makers. This is cited as a reason that IDMs and foundries are clinging to the 0.18 mm process longer than expected. Both design and production communities are trying to come out with development methodologies and tools that will make nanotechnology SOC commercially feasible. Yet, the gap between process capabilities and design capabilities are widening.
The quirks of manufacturing are not understood by designers, and vice versa. Some predict that designing and manufacturing can no longer be separate professions. Designers will have to be manufacturers, and vice versa, calling forth a new discipline in learning. Australia seems to be at the leading edge of this new science, with universities quickly responding to the nanotechnology change. South Australia's Flinders University offered the world's first undergraduate degree in nanotechnology in 2001. Reports of SOC's death, whatever their source, are highly exaggerated.
You can reach Kirtimaya Varma at
kirti.varma@rbi-asia.com
