A New 'Chipless' Semiconductor Industry for India Date:   Monday , November 17, 2008 Although numerous VLSI chip design initiatives are sprouting up at universities throughout India, the country has no semiconductor capability of any economic consequence. And one could argue that it is too late. The collective requirement for vast capital expenditures, decades of accumulated design and manufacturing expertise, a well-established distribution system and an entrenched customer base all create a formidable barrier to entry. But the integrated circuit industry has seen several paradigm shifts in its 40 years of existence. With each, a new constituency has become empowered to participate in the industry. And now it may be time for an entirely new group to take the spotlight. We may be witnessing the dawn of “chip-less” chip design, and the era of the programmer. And world-class programmers — talented, sophisticated, creative software engineers — India has in abundance. Could this be an industry waiting to happen? Cirrus Logic spin-off Cradle Technologies is hoping to make it a reality. Embedded Intellectual Property Fundamentally, chip design is the embedding of complex algorithms in silicon, just as programming is the embedding of complex algorithms in computer code. Whether to use a chip or software to accomplish some sophisticated (or simple but lengthy) processing task depends upon the availability of a “platform” to run the software, and the performance of the combined platform and software in the intended application. Word processing was performed, in the “dark ages,” with expensive, dedicated hardware. It eventually migrated to software. The result was Microsoft Word and WordPerfect, $250 applications running on general-purpose platforms such as the Intel Pentium or the PowerPC. By contrast, functions like Internet packet processing, or video graphics processing, migrated over the years from software into specialized, application-specific chips. In each case, different solutions to a cost-performance equation dictated the endpoint. In recent years, in particular with the explosion of high bandwidth communications and multi-media-type applications (sometimes called “stream processing applications” for the simultaneous processing of multiple streams of data) the three predominant platforms for software — micro-controllers, digital signal processors, and microprocessors — have increasingly proven unable to provide the required performance in leading-edge applications. As a consequence, thousands of companies have been forced to embed their intellectual property in silicon, a process requiring substantial sacrifices in time-to-market, money, and flexibility. And one that takes chip designers — lots of them. This, in some sense, is unfortunate. Cradle CEO Satish Gupta argues that software is a far better medium for the design and capture of complex algorithms. It can be written anywhere, modified and extended easily, debugged modularly, shared, understood by almost anyone, fixed in the field, and reused. Says Gupta, “Archimedes said: ‘Give me a lever long enough, and a prop strong enough, and I can single-handedly move the world.’ Had he been a programmer, he would have said, ‘Give me a platform broad enough and strong enough and I can program the world.’” Archimedes Redux Archimedes could be getting his wish. In response to the explosion in stream processing applications, and the growing inability of both existing platforms and the most sophisticated IC design strategies to cope with cost-performance demands, Cradle is looking to rewrite the rules of chip design. The company’s new platform — called “universal microsystem” or “UMS” — seeks to empower software engineers the world over to join its unique movement. The question is whether the company’s bold approach will actually take off. The universal microsystem gets its name from the fact that it is designed to replace — particularly in stream-processing applications — micro-controllers, microprocessors, digital signal processors, ASICs, application-specific processors (ASSPs), input-output processors, and even “glue logic,” with a single, universal, off the shelf, fully pre-characterized standard product that only needs software to give it its functionality in any application. Additionally, the UMS is an “open source” platform. Cradle has entered into a partnership with Linux firm Red Hat. According to Gupta, “The universal availability of a wide selection of sophisticated, fully-characterized algorithms contributed and enhanced by software engineers around the world means engineers can concentrate on developing value-added functionality, and not repeating the past — as chip designers are routinely forced to do.” Profound Implications for Software Engineers The characteristics of the UMS platform mean that OEMs can design specific functions and interfaces for their products using chips without ever having to do any chip design at all. In addition, because of the modularity of software and the low barriers of entry to develop code, a wide variety of constituencies — from individual consultants in Bangalore, to entrepreneurs in Delhi, to software development companies in Bangalore and Pune — could potentially participate in entire industries, particularly communications and personal electronics, that in the last two decades have been the hegemony of enterprises and nations with strong, entrenched chip design capabilities. In the United States, the Federal Communications Commission has mandated that by 2006, all television transmissions will be digital (DTV). Of these, a significant percentage will be high-definition TV (HDTV). This mandate fundamentally means that every household in America that wishes to continue watching television (statistically equal to every household in existence in the country) will have to purchase a television or set top box containing HDTV/DTV electronics. Manufacturers in this space in recent years have committed hundreds of millions of dollars to the design of chips that encode and decode these signals. Until now, they have had no choice, as the requirement for extremely high performance coupled with very low cost could until now only be satisfied with a custom IC. According to Cradle, the required algorithms, such as MPEG encode and decode, are readily available in software developed for the UMS platform by, surprisingly enough, a team of software engineers in Pune. The same team has developed or assisted in the development of UMS-ready functions in a variety of areas, such as Ethernet and ATM, DSL, TCP/IP, IP Sec, Voice Synthesis, MPEG-2 decode/encode, 3-D engine (rendering, geometry and lighting,) and miscellaneous image processing/ enhancement. The World’s Source of “Chipless” Chip Designs India has gained worldwide recognition for the quality and number of its software engineers, particularly during the recent dot-com ramp-up. But since that bubble burst, it is becoming painfully apparent that this new national treasure is in danger of being viewed as simply “programmers for hire.” As Gupta suggests, “There is a big difference between picking the tea and owning the plantation.” According to Cradle, their UMS technology could bring India’s software engineering community an opportunity to “own the plantation.” Gupta recalls that twenty years ago, Taiwan set out to become the semiconductor manufacturer to the world, and succeeded. This great paradigm shift in the IC industry opened the way for fabless semiconductor companies — widely-known, publicly-traded, billion-dollar-plus giants that surpassed (and survived) most of the famous semiconductor houses of the 1970s. And they never had to build a chip. For Gupta, the emergence of the “fourth platform” — a fully programmable, stream-processing microsystem platform — will bring with it the next paradigm shift, the “chipless” chip house. For an entire, major class of 21st century applications, perhaps the majority of applications in the coming decades, the relevant intellectual property could be embedded entirely in software, using the off-the-shelf UMS. This is certainly an ambitious goal for Cradle, which has recently taped out its first chip and is still moving from a development phase into the deployment of its technology. The real question is whether the market will embrace Cradle’s technology in a sector crowded with potential competitors developing more traditional products. Suhas Patil, Cirrus Logic founder and chairman of Cradle’s board explains, “The old way of doing chips must give way to newer ways of realizing chips.” But Patil, an industry veteran who was one of the principle pioneers of the fabless semiconductor industry, does admit, “It’s still the early stages.” What he is confident about is that the kind of shift that Cradle is trying to implement will eventually come into effect. “As we progress over the next fifteen years,” says Patil, “this change will be increasingly required and I believe it’s only a matter of time. The bigger question is whether Cradle is too early, or whether the company is here just at the right time.” As it is, other companies are beginning to make the shift, at least incrementally, that Cradle is boldly seeking to implement, though nobody is going as far as Cradle in the shift to software. Inevitably, it becomes a cost and performance issue. Gupta is confident that the flexibility and relatively low cost of his solution will be attractive to OEMs. If the UMS does gain significant traction, then an unprecedented opportunity for India and its software community could emerge. Gupta explains that his company has already gained some software development momentum for UMS in India, as well as the ready support of university groups around the country. Much remains to be proven for the UMS, but if the paradigm shift that Gupta speaks of occurs, India could potentially become a powerhouse in stream-processing chip applications. Without ever designing a single chip.