IEEE Spectrum November, 2013 - 27

you'd get by switching from an older-generation chip to a 0.35-µm
processor. The term "0.35-µm node" actually meant something.
But around that same time, the link between performance and
node name began to break down. In pursuit of ever-higher clock
speeds, chipmakers expanded their tool kit. They continued to use
lithography to pattern circuit components and wires on the chip,
as they always had. But they also began etching away the ends of
the transistor gate to make the devices shorter, and thus faster.
After a while, "there was no one design rule that people could point
to and say, 'That defines the node name,'" says Mark Bohr, a senior
fellow at Intel. The company's 0.13-µm chips, which debuted in 2001,
had transistor gates that were actually just 70 nm long. Nevertheless,
Intel called them 0.13-µm chips because they were the next in line.
For want of a better system, the industry more or less stuck to the
historical node-naming convention. Although the trend in the measurements of transistors was changing, manufacturers continued to
pack the devices closer and closer together, assigning each successive chip generation a number about 70 percent that of the previous
one. (A 30 percent reduction in both the x and y dimensions corresponds to a 50 percent reduction in the area occupied by a transistor,
and therefore the potential to double transistor density on the chip.)
The naming trend continued as transistors got even more complex. After years of aggressive gate trimming, simple transistor
scaling reached a limit in the early 2000s: Making a transistor
smaller no longer meant it would be faster or less power hungry.
So Intel, followed by others, introduced new technologies to help
boost transistor performance. They started with strain engineering, adding impurities to silicon to alter the crystal, which had the
effect of boosting speed without changing the physical dimensions
of the transistor. They added new insulating and gate materials.
And two years ago, they rejiggered the transistor structure to create the more efficient FinFET, with a current-carrying channel that
juts out of the plane of the chip.
"WhaT dO YOu mEan bY 14 nm?" WhEn i askEd an sTEEgEn ThaT
Through all this, node name numbers continued to drift ever
question at an industry conference in July, she smiled and let out
a wry, knowing laugh. "Ah...what's in a name?" asked Steegen, downward, and the density of transistors continued to double from
senior vice president for process technology development at Imec, generation to generation. But the names no longer match the size
of any specific chip dimension. "The minimum dimensions are getthe Belgian research center. "Actually, not that much any more."
It's a state of affairs that has been nearly two decades
ting smaller," Bohr says. "But I'm the first to
in the making. Once upon a time, the node name told
admit that I can't point to the one dimension
you practically everything you needed to know about a
that's 32 nm or 22 nm or 14 nm. Some dimenchip's underlying technology. If you trained your microsions are smaller than the stated node name,
and others are larger."
scope on microprocessors made by a handful of differThe switch to FinFETs has made the situaent companies using a 0.35-micrometer process, you'd
tion even more complex. Bohr points out,
find that their products were all remarkably similar.
for example, that Intel's 22-nm chips, the
In the mid-1990s, when such chips were the state of the
current state of the art, have FinFET tranart, 0.35 µm was an accurate measure of the finest features that could be drawn on the chip. This determined
sistors with gates that are 35 nm long but
dimensions such as the length of the transistor gate, the
fins that are just 8 nm wide.
electrode responsible for switching the device on and off.
That is, of course, the view from a chip
Because gate length is directly linked to switching speed,
manufacturer's side. For his part, Paolo
you'd have a pretty good sense of the performance boost
Gargini, the chairman of the International
-Mark bohr, intel
It's actually become a fairly common refrain among industry
experts. The practice of attaching measurements to chip generations has "been hijacked by marketers to an enormous extent," one
chip-design expert told me. "A lot of it's really smoke and mirrors,"
says analyst Dan Hutcheson of VLSI Research in Santa Clara, Calif.
It's "spin," he says, that's designed to hide widening technological
gaps between chip companies.
The nanometer figures that Hu discussed are called nodes, and
they are, for want of a better term, the mile markers of Moore's Law.
Each node marks a new generation of chip-manufacturing technology. And the progression of node names over the years reflects
the steady progress that both logic and memory chips have made:
The smaller the number, the smaller the transistors and the more
closely they are packed together, producing chips that are denser
and thus less costly on a per-transistor basis.
But the relationship between node names and chip dimensions is
far from straightforward. Nowadays, a particular node name does
not reflect the size of any particular chip feature, as it once did. And
in the past year, the use of node names has become even more confusing, as chip foundries prepare to roll out 14-nm and 16-nm chips,
custom-made for smartphone makers and other customers, that
will be no denser than the previous 20-nm generation. That might
be just a temporary hiccup, a one-time-only pause in chip-density
improvement. But it's emblematic of the perplexing state of the field.
Moore's Law, when reflected through the steady march of node
names, might seem easy and inexorable. But today a plague of
intense manufacturing and design problems is forcing compromises that are sometimes sobering. And some analysts suggest that
regardless of what we call the next generation of chips, the transition from old to new no longer provides nearly the kind of payoff-
in cost or performance-that it used to.

"There was
no one
design rule
that people
could point
to and say,
'That defines
the node
name.' "

SPectrUm.ieee.orG

north american

nov 2013

http://SPectrUm.ieee.orG

Table of Contents for the Digital Edition of IEEE Spectrum November, 2013