Golden Years or Golden Age?
12/01/2005
A familiar refrain from market analysts these days is that the semiconductor industry has “matured.” To support this notion, they point to a number of recent developments, including the consumerization of the market, the commoditization of components, the consolidation of companies, and the impending end of further device shrinks. To hear them, one would conclude that the best years of the industry have come and gone.
But there’s little doubt that instead of entering the Golden Years, we are on the threshold of the Golden Age of semiconductors. In fact, one of the leading proponents of this view, John E. Kelly III, senior VP of technology and intellectual property at IBM, recently spoke at the fifth annual Albany Symposium on the Global Business of Semiconductors and Nanotechnology about how the nanotech era will usher in the new Golden Age.
To characterize how we would recognize a Golden Age of semiconductors, Kelly noted, we need only turn to other, relatively recent Golden Ages, such as in radio and television. The hallmarks of these ages were impressive breakthroughs in the development of new technologies followed by explosive creativity in the application of those technologies that significantly changed our lives individually and as a society.
One could argue that these events have already occurred in semiconductor electronics. After all, there’s no question that we’ve seen staggering technological advances over the past few decades in all the familiar measures of size, cost, and performance. Likewise, we’ve seen tremendous innovation in the uses of the technology across virtually every professional and personal endeavor. But have we reached the point of diminishing returns in achieving additional technological improvements, and have we run out of ideas for new killer applications?
Regarding further technological breakthroughs, there are three main barriers to reaching the Golden Age, according to Kelly. The first is a “power wall,” whereby dielectric insulators, which are now only a few atomic layers thick, leak so much current that passive power is approaching active power, resulting in serious heat problems and switches that do not fully shut off. The second is a “frequency wall,” which is a clocking problem in large, high-speed chips that imposes a 30% performance penalty when progressively more transistors are squeezed onto a chip. And the third is a “memory wall,” or a bus bandwidth limitation that results in serious latency problems when moving data to and from the system’s processing units.
Granted, to a large extent, these technological barriers can be circumvented through chip design innovation and industry collaboration, Kelly claims, citing as one example an international partnership between IBM, Sony, and Toshiba that led to the development of the parallel-processing “Cell” system-on-a-chip, scheduled to appear in the new Sony PlayStation 3 game console this spring. One key feature of this 2-teraflop chip is a 300GB/sec main interconnect bus that enables high-speed distribution of workload and data between the system’s nine processors.
Moreover, the Cell system, which is capable of running real-time 3D applications, is already enabling some of the most interactive simulations ever developed, not only in popular console-based games, but also in fields beyond gaming. Consider, for example, the Cell’s ability to create a realistic image of a person’s beating heart from PET and MRI scans or to generate detailed images of the earth’s surface from satellite data.
But these applications provide only a glimpse of what is yet to come. Indeed, it will take no great leap to get from real-time imaging of the heart to creating interactive, photoreal models of all other organs, tissues, cells, structures, and processes in the body to aid in medical research and treatment. Likewise, it’s a small step from simply imaging the earth’s surface to creating live, physics-based simulations of real locations, say, of New Orleans during a hurricane, to aid in disaster planning and rescue operations. Such are the explosions in creativity common to a true Golden Age.
Of course, to realize such applications will require new capabilities. These will include design tools that can perform power, timing, and performance analysis in parallel to create more advanced Cell-like architectures designed to circumvent technology walls, and will entail further development of nanomaterials and techniques to push straight through such walls. And perhaps most needed to take us to the Golden Age and beyond will be a more technically advanced, next-generation workforce. As Kelly urges, we need to get our universities, our companies, our government, and our kids excited about where semiconductor electronics is going. If we can, they will create technologies and applications that have greater influence on more aspects of our lives than anything we’ve seen before.
 |
Phil LoPiccolo
Editor-in-Chief