[FoRK] Welcome to the era of radical innovation

Eugen Leitl eugen at leitl.org
Wed Jan 8 02:18:29 PST 2014


Welcome to the era of radical innovation

Why the end of Moore's Law may be a good thing for innovation

By Patrick Thibodeau

January 7, 2014 06:45 AM ET

Computerworld - Moore's Law created a stable era for technology, and now that
era is nearing its end. But it may be a blessing to say goodbye to a rule
that has driven the semiconductor industry since the 1960s.

Imagine if farmers could go year to year knowing in advance the amount of
rainfall they would get. They could plant crops based on expected water

That's the world that device makers, who are gathering this week in Las Vegas
for the Consumer Electronics Show (CES), have long been living in, and every
year has been a good one. Droughts haven't been part of the forecast, yet.

The tech industry has been able to develop products knowing the future of
processing power, meaning device makers could draw up product road maps based
on microprocessor performance gains that could be reliably anticipated.

In sum, the technology industry has been coasting along on steady,
predictable performance gains.

But stability and predictability are also the ingredients of complacency and
inertia. At this stage, Moore's Law may be more analogous to golden handcuffs
than to innovation.

Technology innovation, particularly in the past decade, has been "a
succession of entertainment and communication devices that do the same things
as we could do before, but now in smaller and more convenient packages,"
wrote Robert Gordon, an economist, in a recent paper for the National Bureau
of Economic Research that addressed the question of whether U.S. economic
growth is over.

Moore's Law, first described by Intel co-founder Gordon Moore in 1965, states
that the number of transistors on a chip would double approximately every two
years. But the law was never meant to hold true indefinitely, and today
microprocessors are reaching a point where they can shrink no more.

The 14-nanometer silicon chips that are now heading to mobile phones and
elsewhere may eventually reach 7nm or even 5nm, but that may be it.

When the European Commission looked at the changing landscape in
high-performance computing and the coming end of Moore's Law, it saw
opportunity. No longer will "mere extrapolation" of existing technologies
provide what is needed, but, instead, there will be a need for "radical
innovation in many computing technologies," it said in a report this year.

And in a recent budget request, the U.S. National Science Foundation said
that radical innovation beyond Moore's Law will require "new scientific,
mathematical, engineering, and conceptual frameworks."

The NSF sees a need for new materials that can work in quantum states, or
even "molecular-based approaches including biologically inspired systems."

That new technology could be carbon digital circuits made of nanotubes, which
could perform 10 times better than today's technologies, as rated by metric
that considers both performance and energy usage. A nanotube is a rolled-up
sheet of graphene.

Another emerging technology that may replace or more likely augment
microprocessors is quantum computing, something both NASA and the NSA are
working on, as are most other major nations.

The end of Moore's Law was a topic of discussion at the recent SC13
supercomputing conference. Experts see instability and much uncertainty ahead
now that the technology we rely on today can no longer be expected to improve
at a regular, predictable pace.

Marc Snir, director of the Mathematics and Computer Science Division at the
Argonne National Laboratory, and a computer science professor at the
University of Illinois at Urbana-Champaign, told SC13 attendees (see slides)
that alternate technologies are not yet ready.

Christopher Willard, chief research officer at Intersect360 Research, said
that the era of buying commercial off-the-shelf products to assemble a
high-performance system is coming to an end. "The market should then be
entering a new phase of experimentation, and computer architecture
innovations," he said.

The demise of Moore's Law is already evident in the high-performance
computing world.

If Moore's Law continued to hold true, the U.S. would have an exascale system
in 2018, instead of the early 2020s, as now predicted.

A 1 gigaflop system was developed in 1988 and nine years later work was
completed on a 1 teraflop system. In 2008, work on a 1 petaflop system was
finished. A petaflop is a thousand teraflops, or one quadrillion
floating-point operations per second.

The end of Moore's Law isn't as urgent of a concern for the device makers at
CES as it is for supercomputing researchers.

But there is a shift in themes at CES -- the focus has moved away from
smaller, faster, better gadgets to the Internet of Things. The underlying
message is: True computing power is measured by the ability of a mobile
platform to control and track a multitude of physical and virtual objects
over a network. But that message might work for just so long.

The problem that high-performance computing faces in reaching exascale will
also eventually confront the device makers at CES, which was launched in
1967, two years after Gordon Moore delivered the paper outlining Moore's Law.

The problem the device makers at CES face is that Moore's Law ends for

Patrick Thibodeau covers SaaS and enterprise applications, outsourcing,
government IT policies, data centers and IT workforce issues for
Computerworld. Follow Patrick on Twitter at Twitter at DCgov, or subscribe to
Patrick's RSS feed Thibodeau RSS. His email address is
pthibodeau at computerworld.com

More information about the FoRK mailing list