News and New Products

Panelists predict EDA and semiconductor challenges of 2017

DAC: EDA executives make conservative 10-year predictions, but touch on Moore's Law, multicore trends, the need for system-level design, and more.

By Michael Santarini, Senior Editor -- EDN, 6/6/2007

In 2017, Moore's Law will continue to hold true, chip design will have mixed with adjacent fields such as mechanical engineering, and design complexity and the need for efficiency will continue to rise. So predicted participants in a panel entitled "Megatrends and EDA 2017" Tuesday at the 2007 Design Automation Conference.

Although the panelists kept their prognostications fairly modest and conservative (no one predicted flying cars or anything of that nature), they did touch on the size of the semiconductor and EDA industries in the year 2017 and made some predictions regarding whether programmable platforms will become dominant and, if so, how that will impact the EDA design flow and product mix.

Moderated by Greg Spirakis, who retired a couple of years ago from Intel, where he had been vice president of the Design Technology Group, the panel included Fu-Chieh Hsu, vice president of design and technology platforms at TSMC, Juan-Antonio Carballo, general partner with Argon Venture Partners, University of California, Berkeley, Proffessor Kurt Keutzer, Kazu Yamada, vice president and general Manager of Custom SOC Solutions at NEC, and Synopsys CEO and chairman Aart de Geus.

Before the panelists kicked off their presentations, Spirakis first illustrated how far the electronics industry has come since 1997. Some of the more notable datapoints: The number of Internet users rose from 205 million in 1997 to 2.6 billion in 2006. The number of cell-phone suscribers grew from 57 million to 1.1 billion in the same time frame. And digital cameras and MP3 players, which did not exist in 1997, accounted for $18 billion and $6 billion, respectively, in 2006.

"What will be the megatrends 10 years from now?" Sprikas asked.

Panelists predicted that the driving factors influencing the industry over the next 10 years include globalization, specifically the emergence of markets and new avenues of chip production and design in mainland China and India; the mix of EDA and electronics with adjacent fields such as mechanical design; and the continuation of the trend toward programmable, multicore platforms.

Keutzer, who a few years ago predicted the rise of multiprocessor programmable platforms, stated that the trend will culminate by 2017 in handheld devices that essentially do everything advanced handheld devices do today (camera, phone, GPS, email, Internet) plus all the functions found on laptop computers today, and likely more. He also predicted that multicore SOCs would proceed not in a neat two-core, four-core, eight-core, 16-core trajectory generation over generation, but rather in huge jumps to 32 cores, 64 cores, and 128 cores over the next 10 years.

Over that same time, TSMC's Hsu predicted, design starts and the number of hardware platforms will continue to decrease while design complexity and the need for productivity will continue to increase. "To solve complexity problems, the EDA industry needs economy of scale as well as a mechanism to reach greater efficiency," Hsu said.

Panelists largely agreed that Moore's law will continue to drive the industry. DeGeus noted that EDA and semiconductor predictions have traditionally been very easy to make because the industry could simply follow the trajectory of Moore's Law, but that could change for "technomic" reasons.

"Complexity is manifesting itself in both technology as well as the economic side of the market and the economic driver is becoming much more relevant than in the past," de Geus said.

In the case of video-related electronics, de Geus noted, the EDA industry, semiconductor industry, and ultimately the electronics industry are today driven by the needs of operators such as Verizon, AT&T, and Sprint, which are in turn driven by the needs of application providers such as YouTube, Microsoft, Google, Yahoo, and eBay, which in turn are driven by the needs of content providers such as TimeWarner, Sony, NBC, and the BBC.To deal with this complexity, de Geus holds that the semiconductor and EDA industries must provide a technomic model that simultaneously addresses capital and cost complexity, ecosystem complexity, scale complexity, and systemic complexity.

NEC's Yamada called for an "EDA 2.0," a tool flow and methodology that would help systems companies with true system definition (system meaning a cell phone or a refrigerator, rather than just an SOC) where hardware and software functions are determined based on performance and functionality requirements. After system definition, EDA 2.0 would also have to have a seamless link from true system definition to hardware definition, where designers would pick what goes onto a board and what functions should be implemented in SOCs versus SIPs (systems in package). This step would also be tied in with cost estimation. The flow would then proceed, again seamlessly, to the traditional hardware/SOC implementation flow of synthesis to GDSII.

TSMC's Hsu added that tomorrow's EDA tools will have to become more "deeply rooted in the physics" of the manufacturing platforms and more closely tied with the platforms themselves. The tight requirements to tie tools more closely to flows will likely mean fewer EDA flows and perhaps fewer big EDA players to offer them, Hsu added.

Panelists agreed that by 2017, the semiconductor industry would grow into a $1 trillion dollar industry from the $280 billion industry it is today, but they varied in predicting whether the EDA industry would follow in lockstep. By 2017, panelists said, the EDA industry would grow to between $8 billion and $12 billion, up from $5 billion in 2007 (Editor's note: It's uncertain whether the panelists were including IP revenues, a matter of some controversy).