Prediction: By 2020 there will be 50 billion mobile devices generating 1,000 times today's traffic. How can the semiconductor supply chain make this possible? By providing complete solutions that go beyond silicon alone, according to speakers at last week's Global Semiconductor Alliance (GSA) conference.
The one-day conference focused on "global mobile data." I attended a keynote and a panel discussion on infrastructure innovation. The 50 billion device prediction (a 10X increase over today) came from the keynote given by Arpit Joshipura, vice president for strategy and communications at Ericsson.
Joshipura implored infrastructure providers to support standards and to think "end to end" as opposed to just supplying chips for a box. "The ASIC industry is no longer isolated to just an SoC or a device or a board or a box," he said. "It's a new world. Screen to server is the new end-to-end and you have to think of the big picture."
Explosion of Data
This literal "think outside the box" message was also reflected in the panel, entitled "From MIDs to Base Stations - Where Mobility Infrastructure Meets Innovation." Panelists came from across the semiconductor supply chain, including Cadence.
Moderator Art Reidel, CEO of Scintera Networks, started the discussion by noting predictions of a "literally exponential" increase in packet data over mobile networks in the 2007-2014 time frame. By 2015, he noted, there could be a 90X gap between an advertised, theoretical peak data rate and a fully loaded cellular network. How will the semiconductor industry help solve these problems, he asked?
A New Approach to Design
If the semiconductor industry is going to enable tomorrow's mobile communications networks, a new approach to design is needed. Such an approach is part of the EDA360 vision, which "acknowledges that applications are king," Bruggeman said. He noted that software is far outpacing hardware in terms of resources and importance, yet the EDA industry has never addressed the software problem. That's changing with EDA360, which will offer new solutions "to attack the challenges my friends on the panel are facing every day."
"Our products have become a lot more software intensive," responded Abdi. "The more transistors you put in there, the more software has to do." Building a software interface to hardware is a new and difficult challenge, he said. Abdi noted that tools like the Cadence Palladium emulator help, but said that "a lot more needs to be done."
During a discussion of low power design, Bruggeman noted that designers don't get much power reduction by looking at hardware alone. "If you start with software, you can have much more impact on power," he said. What needs rethinking, he said, is how to give software visibility into the hardware.
Given that hardware at the transistor level is "almost free," there's an ongoing shift in value towards software, Harding said. "We don't do business with a chip company that doesn't have a software solution to go with its silicon. It's dead on arrival," he said.
You Still Need a Good Hardware Team
Abdi ended the panel discussion on what should be an encouraging note for hardware designers. "What happens when silicon and MIPS become free? Your business becomes software," he said. But that doesn't mean companies can ignore hardware design. "You have to have a stronger hardware team to make it all work together," he said. "It actually requires better hardware."
This is an important point. Hardware designers should not regard an application-driven, software-centric approach to electronic design as a threat. Rather, such an approach will require better hardware and place a real premium on those whose skills span the hardware and software worlds. It's a challenge but it's also a great opportunity for those who step forward to embrace it.