The Magazine of IEEE-Eta Kappa Nu July 2017 - 17

The IEEE Rebooting Computing Initiative

I. Introduction - Moore's Law and Computing
Electronic computers and transistors both had their
origins around 70 years ago, but starting about 50
years ago, with the silicon integrated circuit (the "chip"),
these two technologies really took off together. This was
quantified by Gordon Moore of Intel, who projected in
1965 that the number of transistors integrated on a
chip would double every two years [1]. This exponential
improvement has been maintained or exceeded for
50 years, a remarkable period of growth unmatched in
any other technology. While the scale of transistors is
still shrinking, this is reaching its practical limit. As the
atomic limit is being approached, chip manufacturing is
becoming more difficult and expensive.
The core of a digital computer is the central processing
unit or CPU. A CPU generally has a master clock that
synchronizes operations of different portions of the
circuit, with a clock frequency or clock speed, where one
clock cycle corresponds to a single logical operation or
switching event. For circuit reliability, the maximum clock
speed is typically somewhat less than the reciprocal
of the characteristic switching time for the devices. For
example, if the switching time is 100 ps, the clock speed
might be about 1 GHz.

Starting around 1970, a simple CPU was integrated on
a single chip, known as a microprocessor. The resulting
microcomputer provided the basis for the personal
computer revolution, which in turn led to the internet.
Smaller transistors permit more complex circuits on a
chip, but also switch much faster, enabling computers
to run much faster. So Moore's Law also led to scaling
of computer speeds. Figure 1 shows the trends in clock
speed of state-of-the-art microprocessors [2] starting
around 1985, on a log-plot of clock speed and a linear plot
of years. This corresponds to exponential increase in speed
until around 2005, when speed saturated at a few GHz.
Computer performance continued to increase through
the use of multi-core chips, with two or more CPUs on a
single chip. This enabled further increases in speed, at least
for problems that are amenable to parallel computation.
However, this approach is not very scalable, since the power
is excessive, and portions of the chip must be turned off
much of the time to avoid overheating. (This is known as
the "dark silicon" problem.) For continued improvement in
the future, entirely new approaches are needed.

II. IEEE Rebooting Computing Initiative and
Roadmapping the Future
Changing the direction of the entire computer industry
is a major undertaking, particularly when the new

Fig. 1. Trends in clock speeds of microprocessors from 1985 to present, showing saturation of clock speeds at a few
GHz after about 2005. Data from Stanford University VLSI Research Group.
THE BRIDGE // Issue 2 2017

Table of Contents for the Digital Edition of The Magazine of IEEE-Eta Kappa Nu July 2017