Limits on Moore’s Law
In the April 1965, Dr. Gordon Moore, one of the early pioneers in semi-conductors, wrote in Electronics Magazine his observations about the rate of growth of the then emerging Semi-conductor technology. He postulated that the density of transistors would double every twenty months, with a proportionate decrease in cost. Popular media noticed his writing dubbing it “Moore’s Law.”
Since then, the Semi-conductor industry’s has pursued a Holy Grail quest to prove true Moore’s Law by continually improving performance through increased transistor density and lower prices. The pursuit of this quest has brought about the incredible technological end for the Twentieth-Century, and has provided a great burst of momentum for the Twenty-First Century. At some point, however, the very principles of Physics suggest limits on this exponential growth. After all, at the atomic level, we have to recognize limitations to finite entities in closed systems. Certainly principles of Economics also suggest limits, for it is one matter to spend research money in university laboratories exploring new materials and processes but quite another to start a fabrication facility. Perhaps, too, ethics itself begs for limits. We, if we so choose may have the reason and the will to set limits.
Physical Limits:
The micro dimensions already employed by Semi-conductor manufacturers certainly suggest limits from the tooling angle––especially considering limitations of the lithographic process. Consider Intel’s current process for the P4 family that etches CMOS transistors with a thirty nanometer (thirty billionth of a meter) distance between logic gates. Shrinking the distance between gates allows for more transistors in a give area, i.e. greater density. Researchers at the University of Wisconsin, Madison, Center for Nanotechnology, have invented a technique that can etch twenty nanometer features by modifying improving lithographic techniques. The researchers envision full functionality of the “Bright-Peak Technology” by 2010, which would extend Moore’s Law.
Yet, with increased density, the industry faces the challenge of power leakage and heat dissipation. In the not to distant past, processors required a heat sink to dissipate heat. Now, however, liquid coolants, extensive fanning systems and air conditioning are recommended for complex multi-processing environments. While discussing heat dissipation, one wag quipped, “You could fry an egg on a PII.” Even Gordon Moore himself, when commenting about density challenges, had quipped. “Approaching the density of a nuclear reactor.” While developments in new materials suggest breakthroughs in speed and density, the threshold for the standard CMOS transistor appears to be slightly below the nine-nanometer area. Beyond that threshold, the instability of electron flow brings functionality to question.
Economic Limits:
So far, the economic aspect of Moore’s Law (an increase in density yields a proportionate decrease in price) has held true. In the Nineteen Fifties, a single transistor had manufacturing costs between five dollars to forty-five dollars. With our current mass production capabilities, a transistor now cost less than $0.00001 per each. Even so, Moore’s Law has a second component––doubling density directly doubles fabrication costs. While the ability to mass produce chips has increased, so have fabrication costs. To build a fabrication plant in 1995 cost about $1 billion. By 2010, start-up costs could average between $30 billion to $50 billion. At some point, investors would see increased risks and diminishing ROI. For now, it appears, manufactures attempt to reduce costs by locating manufacturing in low-wage, low-restriction nations. If we stay with the silicone CMOS paradigm, we may reach a profitability plateau.
Ethical Limits:
At the 2003 Infosec Security Conference in London, Phil Zimmerman, the developer of the PGP (Pretty Good Privacy) encryption product discussed concerns about technology escalation. Observing the “millions” of CCTV cameras in Great Britain, he was quick to note the invasion against our civil liberties. “You have millions of CCTV cameras here. Every citizen is monitored, and this creates pressure to adhere to conformist behaviour. The original purpose of cameras was to catch terrorists, but to my knowledge they haven't caught many terrorists using cameras." Zimmerman blamed the relentless pursuit of Moore’s Law as the main contributor to privacy invasion. If we are under constant surveillance, then we will have pressures to conform. Itself, that may not pose a problem, but who decides the standards of conformity?
Recently, in Maryland, a thirteen-year-old student was given an Internet assignment to research and to write a report on a major structure. He chose the Bay Bridge. With the help of his father, they researched the construction techniques, materials, costs, etc. In the process of the boy’s research, the Headmaster of the boy's school received a visit from the FBI’s Counter Terrorism Task Force. The incident came to a humorous end when they realized the lad was not a terrorist. Still, the power of technology to invade our privacy is beyond anything we have ever experience. Moreover, as Mr. Zimmerman points out, will the ever-increasing escalation of technology threaten our freedom in ways we dare not even imagine. If so, we could see legal limits on Moore’s law. As awareness and concerns mount, Congress could put limits and oversight on research spending. People could demand, and courts could enforce, our basic Constitutional rights to due process and reasonable cause. International governing bodies could impose restrictions on privacy invasion. The suggested legal restrictions could have a ripple effect of diminishing demand, hence, decreasing production of microprocessors.
How close are we to limits affecting Moore’s Law? Right now, the industry is doing everything in its power to stay on course. Researchers work intensely to discover new models that they hope will become the next paradigm. Yet, for most of us, it will simply mean faster, processors to come out “sometime next year.”
|
