Moore’s
Law has become an endangered species. No longer is it a safe assumption
that the number of transistors in processors doubles every two years. Intel
abandoned this expectation at
the ISC 2016 Conference in Frankfurt, Germany. Processor performance barely
increases even though they become more energy efficient, and the growth of
transistor density has slowed down remarkably. Since considerable time, the
industry is also looking for alternatives to silicon as physical
limits to its useful further miniaturization are approaching.
Enter carbon
nanotubes. The University
of Wisconsin at Madison has reached a major milestone
in manufacturing transistors out of carbon nanotubes that leave its silicon
cousins way behind in terms of conductivity. An increase
of 90% in current was measured by pitting a 140 nm carbon nanotube transistor
against a 90 nm silicon P-channel MOSFET transistor. The carbon nanotube FET
did well even under the disadvantage of a larger node, but the jury is still
out on a comparison to current 14 nm FinFET or Tri-Gate transistors.
Carbon nanotubes consist
of rolled single-atom layers. They are among the most highly conductive
materials known to man. Carbon-based transistors might support a five-fold
increase of performance compared to silicon while energy consumption will
decrease to one-fifth of present levels.
But manufacturing had always run into problems with minuscule metal
contaminations that massively affected conduct of electricity, and thus
performance. New technology developed at UW Madison resolved that by filtering
impurities with the assistance of a polymer that leaves 99.99
percent pure carbon layers.
This permitted a new production process that places carbon nanotube
transistors on a 1x1 inch wafer. But to render the technology interesting and
affordable for commercial use, this process needs to become scalable to produce
larger wafers and higher transistor densities. First experiments appear to have
been successful. Still, several years will likely pass before carbon nanotubes
will appear in commercially available processors, since chip manufacturers will
also need to invest heavily in adjusting their manufacturing processes and
factories.
RAM might take this leap a lot sooner. Fujitsu
is already working on commercialization of nano-RAM (NRAM) based on
carbon nanotubes and plans to initiate mass production based on licensed technology
of Nantero, the world leader in carbon nanotube electronics, by 2018.
Along
the concept, if not the formula, of Moore’s Law, somewhere between 20-50 years
out, further miniaturization of IT elements and devices (but also their
transition and fusion into biotechnology and transhumanism) will be dominated
by a transition from software into hardware, a confusion of the two so complete that it will
literally become impossible to know where the boundary between them ought to be
drawn. And that is precisely the point: perhaps there shouldn’t be a boundary,
especially if the next step or, rather, quantum leap, as some conjecture,
should be DNA
computing.