3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

Intelligent Machines

Computing at the Speed of Light

Replacing metal wiring with fiber optics could change everything from
supercomputers to laptops.

by Tom Simonite August 4, 2010

The world of computing could change rapidly in coming years thanks to

technology that replaces the metal wiring between components with
faster, more eùcient fiber­optic links.

“All communications over long distance
are driven by lasers, but you’ve never had
it inside devices,” says Mario Paniccia,
director of Intel’s photonics lab in Santa
Clara, CA. “Our new integrated optical
link makes that possible.”

Paniccia’s team has perfected tiny silicon
chips capable of encoding and decoding
laser signals sent via fiber optics. Today,
Seeing the light: A chip in the center of this circuit board
when data arrives at a computer via a fiber contains four lasers that convert electrical signals into light
optic connection it has to be moved from a pulses. The pulses travel at high speeds along a fiber-optic
link.
separate photonic device to an electronic
circuit. This new system promises to speed things up because everything
works in silicon.

Last week, Paniccia’s team demonstrated the first complete photonic
communications system made from components fully integrated into
silicon chips. Electronic data piped into one chip is converted into laser
https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 1/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

light that travels down an optical fiber and is transferred back into
electrical signals a few fractions of a second later. The system can carry
data at a rate of 50 gigabits per second, enough to transfer a full­length
HD movie in less than a second.

The silicon photonic chips could replace the electronic connections
between a computer’s key components, such as its processors and
memory. Copper wiring used today can carry data signals at little more
than 10 gigabits per second. That means critical components like the
central processing unit and the memory in a server cannot be too far
apart, which restricts how computers can be built.

The new Intel setup has four lasers built into its transmitter chip that
shine data into a single optical fiber at slightly diûerent wavelengths, or
“colors.” Chips with even more lasers should make it possible to
communicate at 1,000 gigabits per second.

“Having a chip the size of your fingernail that can deliver a terabit per
second changes the way you can think about design,” says Paniccia. Such
chips could make a big diûerence inside the sprawling data centers
operated at great expense by Web giants like Google, Microsoft, and
Facebook. “Data centers today are big piles of copper–that imposes the
limits on how you arrange components inside a server,” Paniccia says.

“If I could just move the memory a foot away [from the processors], I
could add a whole board of memory for a single CPU instead,” says
Paniccia, whose team is experimenting with prototype servers to work
out how to build them with photonics links inside.

Moving a server’s memory away from the CPUs would also make
ventilating them easier. Since roughly half the cost of running a data
center, used for everything from services like Facebook to banking
records, comes from cooling, that could have a significant impact.

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 2/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

Further savings may come from the fact that optical links require less
power to operate, says Keren Bergman, who leads a silicon photonics
research group at Columbia University. “With electrical wires, the longer
you go, the more energy you spend in an exponential fashion,” she says.
Optical fiber allows low­power signals to travel farther faster. Bergman’s
group has used data on the performance of computers at Lawrence
Berkeley and MIT’s Lincoln Laboratories to simulate how systems with
optical interconnects might perform. “You can get an order of
magnitude gain in energy eùciency,” she says, with the largest gains
seen for applications such as high­bandwidth image processing and
video streaming, she says.

Data centers aren’t the only things that may see their insides lit up with
lasers. “We’ve developed this technology to be low­cost so we can take it
everywhere, not just into high­performance computing or the data
center,” says Paniccia. The components of the Intel system, including the
lasers, are made with the same silicon­sculpting methods used to
construct computer chips in vast quantities. “I’m drafting Moore’s law,”
says Paniccia. “We’ve enabled the benefits of using light with the low­
cost, high­volume, scalability of silicon.”In consumer computers like
laptops, that would allow innovations in industrial design. I could put
the memory in the display instead, and change the design of the whole
thing.”

This could make it easier to swap in new components without having to
open up a machine. It would also allow core components to be installed
in peripherals. Extra memory could, for example, be hidden in a laptop
or smart phone dock to increase a portable device’s computing power
when plugged in.

Fully exploiting the benefits of the optical age will, however, means
changes to the components being linked up. “It’s not just a case of whip

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 3/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

out the electrical wires and replace them with optical fiber,” says
Bergman.

Ajay Joshi, an assistant professor at Boston University, who is also
exploring design options for high­performance computers with optical
interconnects, agrees. “If we speed up the channel between logic
[processors] and memory, we need to rethink the way you design that
memory.”

The speed gap between processors and optical links is smaller, but
ultimately, that too will likely change. “It would be nice to also see
processors that work optically instead of electronically,” Joshi says.

The latest Insider Conversation is live! Listen to the

story behind the story.
Subscribe today  
Already a Premium subscriber? Log in.  

More from Intelligent Machines

Artificial intelligence and robots are transforming how we work and live.

01 Is Artificial Intelligence Stuck in

A Rut?
The former director of Uber’s AI lab says the field
is in danger of losing sight of its long-term goals.
by Will Knight

02 Apple’s AI Director: Here’s How

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 4/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

02 Apple’s AI Director: Here’s How

to Supercharge Deep Learning
Ruslan Salakhutdinov, who leads Apple’s AI
efforts, says emerging techniques could make the
most popular approach in the field far more
powerful.
by Will Knight

03 AI-Powered Drone Will Follow

You Around and Take Pictures
Startup Skydio is prepping a consumer drone
that uses machine vision to keep track of you
while avoiding obstacles.
by Tom Simonite

More from Intelligent Machines

Want more award-winning journalism? Subscribe to

Insider Basic.

Insider Basic $29.95/year*

Six issues of our award winning magazine and daily delivery of The Download, our newsletter of
what’s important in technology and innovation.

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 5/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

Subscribe

See details+

*Prices are for U.S. residents only

See international prices

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 6/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 7/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 8/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 9/10
3/30/2017 Computing at the Speed of Light ­ MIT Technology Review

https://www.technologyreview.com/s/420082/computing­at­the­speed­of­light/ 10/10