Has Moore’s Law Reached Its Final Days?

aNewDomain.net — Due to Moore’s Law the computer that sits on your desk, your lap or even on your palm is twice as powerful as the one you bought just two years ago. Your PC will be substantially more powerful than anything back in the Clinton Administration, and light-years beyond that original PC of 1982.

Image Courtesy: Wiki Commons

Will the unstoppable progression of processing power continue forever?

In short, yes. But there is more than just processing power when it comes to computers.

To be clear, Moore’s Law is not a law in the normal context. It does not pertain to physical nature or governmental progress — it is a prediction based on the relevant behavior of semiconductors in the technology industry. Gordon Moore, the executive director of Intel, created the “Law” on April 19th, 1965 in Electronics magazine, which is now defunct. He stated that the number of components (transistors) able to fit on a circuit (chip) had doubled every year since 1958, while each year’s prices remained the same. The Law predicts that this will continue — he later amended the concept to an interval of two years. Simply put, more components on a circuit equals more power.

In an off-hand way the same article also foresaw the progression of computing that we have seen, namely that home computers would become commonplace with “automatic controls for automobiles, and personal portable communications equipment.” Everyone laughed at the time while the editors inked a cartoon of someone hawking home computers.

Critics hounded the statement, anticipating the speedy death of Moore’s Law the second it was created. Even now, after a half-century of doubts and no visible slowdown, pundits say the industry cannot continue to reduce the size of transistors, that silicon and its Valley cannot keep up the pace.

The Complicated Part

Today, amidst our home computers, computerized cars, and smartphones, “We are still on the curve (defined by Moore’s Law) and we are still talking about the fact that it will run out,” said Rob Enderle, technology analyst and head of the Enderle Group. “We always seem to find a way to stop it from running out.”

But this time things are more complicated, says microprocessor industry analyst Nick Tredennick.

“Before, when you shrank the components, everything got better—heat dissipation went down and voltage leakage was inconsequential. Everything was good,” recalls Tredennick. “You got more transistors on the chip and everything got faster. But with recent generations the transistors have been getting smaller but their current leakage is getting significant and heat dissipation is not dropping. It turns out that leakage and heat dissipation will limit what can be done with Moore’s Law.”

“Meanwhile, mobile devices are obviously the future, and have changed customer demand so that design requirements have moved from price-performance to price-performance per watt,” Tredennick noted.

In other words, instead of X-Y graphs that plot price-performance over time, we need X-Y-Z graphs that plot price-performance and wattage over time, expanding our conception of Moore’s Law, he indicated.

Meanwhile, instead of spending hundreds of millions of dollars to upgrade their facilities to produce the next, smaller generation of silicon components, the vendors can be expected to explore alternate approaches, Tredennick added. These include FinFETs, wafer stacking, and carbon nanotubes. Enderle added memristors to the list.

A FinFET is an on-chip transistor with a vertical rather than a traditional two-dimensional layout, so that the resulting components take up less space. The name means “field-effect-transistor with a fin.”

Wafer-stacking takes the 3D concept to the next level. Instead of printing a circuit on a wafer die, putting the die in a chip, and putting the chip on a board where it is connected to a bunch of other chips, you stack multiple wafer dies atop each other, which are then put into a chip and put on a board. The idea is to stack memory on top of a processor, eliminating the processing delays that occur when a signal has to transit between chips.

Carbon nanotubes could, in theory, be used to make transistors that respond to single electrons, opening new horizons in terms of high speed and low wattage by freeing the industry from silicon. Enderle said that adopting carbon nanotubes could actually accelerate Moore’s Law. A prototype carbon nanotube processor was demonstrated recently at Stanford University.

Memristors (memory resistors), meanwhile, are resistors whose level of resistance change according to the nature of the last current that flowed through them, and the memory is retained after the power is turned off. They can be used to replace multiple transistors, leading to higher circuit densities for both processor and memory chip, Enderle noted. Various vendors are currently working on the technology.

Futuristic Technologies

Enderle also noted that, in the future, different technologies will probably be adopted for different environments, with different Moore’s Law curves for each.

For instance, for processors used in the cloud (presumably in vast server farms) physical size won’t matter, but their use will be limited by the bandwidth in and out of the server farms, and the uncertain location of those farms could generate privacy issues. Processors used on the desktop will have to fit size and power limitations, while those used in mobile devices will face extreme size and power limitations and additionally have breakage and theft issues, he indicated.

What Enderle was absolutely certain of is that, no matter how much processing power is available, we will find some use for it. Self-driving cars, as Moore predicted, are already on the horizon, as are decision engines which can answer complicated questions, or intelligent machines that perform repetitive tasks, he noted.

For aNewDomain.net, I’m Lamont Wood.

Based in San Antonio, Texas, Lamont Wood is a senior editor at aNewDomain.net. He’s been covering tech trade and mainstream publications for almost three decades now, and he’s a household name in Hong Kong and China. His tech reporting has appeared in innumerable tech journals, including the original BYTE (est. 1975). Email Lamont at Lamont@anewdomain.net or follow him @LAMONTwood.