2 min read

Speed over silicon

Intel is dabbling in chips without silicon waffers

Moore’s Law has lasted remarkably well. Forty-one years after Moore originally stated it in Electronics Magazine, the law still more or less holds. (Depending on how you define it. The law is often defined as saying the number of transistors on a chip doubles every 18 months, Intel’s own Web site says the law is that “the number of transistors doubles about every two years,” while the original Electronics article actually said that “the complexity for minimum component costs has increased at a rate of roughly two per year,” in other words a doubling every year. Take your choice.)

But this isn’t one of those natural laws, like gravity, that just happens. It’s more like the speed limit on a freeway – sticking to it is hard work. Chipmakers like Intel are continually trying out new technologies they hope will help them keep up the pace.

One thing they keep exploring is new semiconductor materials. And the latest development on that front is the demonstration in December of a prototype transistor using indium antimonide.

Indium antimonide is one of an assortment of compound semiconductors chipmakers have played with over the years. Compound semiconductors have been used in chips, though not so far in microprocessors.

According to Rob Willoner, a technology analyst at Intel, indium antimonide can deliver about a 50-per-cent speed advantage over silicon while consuming about one tenth as much power. That means Intel’s research on indium antimonide isn’t purely about keeping up with Moore’s Law. It’s also about more computing power for the same amount of electrical power – a goal Intel has been talking about recently.

Steve Smith, vice-president of Intel’s Digital Enterprise Group, says his company has been looking for ways to build energy-efficient chips for notebooks since the 1980s, but “in the last five years or so, we increasingly recognized the need for performance per watt in other segments.”

IT departments are increasingly concerned about the power and heat issues as they try to pack more computing power into the limited space of data centres.

And both ergonomic and economic issues argue for more energy-efficient desktops.

Intel demonstrated its indium antimonide prototype in co-operation with QinetiQ Ltd., a British defence and electronics firm.”We’re not experts at com-pound semiconductors,” Willoner says, “that’s not our strength.”

In a way, compound semiconductors are an old story. They are made up of elements from two columns of chemistry’s periodic table – the columns on either side of the one containing silicon. In fact, indium and antimony come right before and right after silicon in the periodic table.

Look up one row from each, and you find gallium and arsenic. In the early 1980s, there was a lot of interest in the possibilities of gallium arsenide as a semiconductor material.

As it happens, gallium arsenide is used as a substrate in the prototype that Intel and QinetiQ demonstrated in December.

Eventually, Willoner says, Intel hopes to use silicon as a substrate for commercial indium antimonide chips.

And, when will that be? Not soon. We may still only be about two thirds of the way from those early experiments with gallium arsenide to commercial compound semiconductor chips. Willoner says Intel is targeting production in “the second half of the next decade.”

Moore’s Law dictates pretty fast growth in computing power, but sometimes the work behind that progress takes a long time.