A Roadmap to Evolving SOC Capability

The winners in the system-on-a-chip race will have to provide proven IP libraries, an effective Internet-based infrastructure, and methodologies developed in close collaboration with EDA vendors.

by Badar Baqai

To predict how the semiconductor industry will transform itself in the future, we must look briefly at how it has evolved in recent years. Following the Sematech initiative of the late 1980s, the industry routinely released a new generation of scaled technology every 18 months or so. Yet while silicon technology continued its rapid growth, we struggled with deficiencies in tools and methodologies. And although the EDA industry has seen unprecedented growth in this past decade, we still see the gap widening between our manufacturing capability and our ability to utilize it.

Defining new problems at a rate faster than we can solve them, coupled with the accelerated pace of geometric scaling at the manufacturing level, may offer the best explanation for the current design predicament. It seems evident that change is needed--and in fact is already underway on several fronts--to make true system-on-a-chip (SOC) solutions a reality. Such programs shorten development time and reduce implementation risks for SOC designers. Combining world-class experience in proven intellectual property (IP) and mastery of various semiconductor technologies, these programs satisfy customers' demands for speedy time-to-market turnaround and true system differentiation.

First comes the need for libraries of reusable IP, culled from in-house designs and licensed from third parties, allowing customers to evaluate functions in real-world applications. These libraries range from functional blocks, such as USB, PCI, IrDA, and JPEG, to sophisticated standards such as MPEG and ATM. This category would also include processor and DSP cores such as Sparclite, Oak DSP, ARC, and ARM. To build up these libraries, we must standardize the quality of the IP. So the ongoing project to ensure the compatibility of internal and external IP uses the design and coding standards promulgated by the industry's Virtual Socket Interface Alliance (VSIA).

The development of IP libraries parallels the need for what we call an IP highway, an Internet-based infrastructure that allows efficient global transfer of IP information and design data. We liken this delivery system to a depository, where multiple sources can deposit knowledge in an "IP bank" and users can access this knowledge from multiple locations. With a mechanism like this, which overcomes the limitations of time and distance, chip designers can quickly retrieve necessary IP unique to their requirements.

In addition to the above, SOC vendors must also provide both the methodologies and environments necessary to implement their designs efficiently and quickly. Since SOC vendors typically lack EDA experience, they must work closely with those vendors to establish a standard methodology. In addition, when SOC vendors have accumulated expertise in ASIC technology and system LSI development, customers receive wide support in the selection, use, and integration of IP. In such a system, customers can access proprietary cores--along with third-party cores, design service organizations, and core-information providers--by way of the Internet.

All of these practices contribute to product differentiation, shorten time to market, and reduce development risk for customers. Silicon technology will reestablish itself as the most important differentiating factor for those utilizing SOC solutions. Those vertically integrated companies with a track record, willing to invest to establish the required infrastructure, a commitment to meet the challenges of the silicon technology roadmap, and an established true end-user systems heritage, will enjoy the clear upper hand and provide the best technology and service to their customers.

Although market data remains somewhat sketchy, I predict a paradigm shift toward SOC, with the crossover point coming shortly after the turn of the century. The outline of the new era and the benefits that it will bring are already evident. In particular, SOCs will provide a boon to medium-sized companies, enabling them to differentiate themselves from bigger players without having to make a phenomenal investment in time and resources.

Badar Baqai is vice president of the Systems Solutions group of Fujitsu Microelectronics, Inc. in San Jose. He has more than 20 years experience in microprocessors, graphics, and computing systems. Previously he served as the chief technical officer for Fujitsu's System LSI group and vice president and general manager of the Processor Products business unit.

To voice an opinion on this or any other article in Integrated System Design, please e-mail your comments to jeff@isdmag.com.