Raphael ThysBuilding a better supercomputer is something many tech companies, research outfits, and government agencies have been trying to do over the decades. There’s one physical constraint they’ve been unable to avoid, though: conducting electricity for supercomputing is expensive.
Not in an economic sense—although, yes, in an economic sense, too—but in terms of energy. The more electricity you conduct, the more resistance you create (electricians and physics majors, forgive me), which means more wasted energy in the form of heat and vibration. And you can’t let things get too hot, so you have to expend more energy to cool down your circuits.
Any gamer or regular laptop user is familiar with overheating problems. Supercomputing deals with the same issues on an exponential scale, with energy use similarly enlarged and thus a significant cost concern (there’s the economic bit). That’s why consumer supercomputers that try to control these issues will run you at least $6,000, and why supercomputing as a whole has been reserved for code-cracking spy agencies, really Big Data crunching at various companies and governments, and mega-funded research institutions.
A new nanowire breakthrough could be the first in a wave of similar innovations set to change all that.
Superconductors Gone Miniature
One of the holy grails in supercomputing development—and in electronics/physics at large—has been developing a resource- and cost-effective superconductive material. Unlike regular electric conductors, superconductors transmit electrons (i.e. “electricity”) without any resistance. They output exactly the same amount of energy that was inputted and do not dissipate “waste” energy in terms of heat or sound. This means they require much less energy to run and no energy to cool…sort of. While superconductors don’t generate any heat while conducting electricity, all of the superconductive materials yet discovered and developed have to be cooled below a critical temperature to gain their superconductive powers. This has made them more expensive and unwieldy in most applications, including supercomputers, than using more traditional materials and cooling systems. But researchers at the University of Illinois at Urbana-Champaign have just created a working superconductive nanowire memory cell that could prove to be a game changer.
“An SEM image of Device 7715s1. Two carbon nanotube templated Mo75Ge25 wires lay across a roughly 150 nm wide trench, 2.5 μm apart. The two wires have similar dimensions, but are not identical. (b) An SEM image of Device 82915s2. The Mo75Ge25 (dark) is patterned into two geometrically different nanowires sitting 150 nm apart. The right wire has a non-uniform width.”
Get all that? Good.
The cell consists of a nanowire loop and a couple of electrodes and is programmed simply by subjecting it to an initial positive or negative charge, which sends electrons around the loop either clockwise or counterclockwise. This binary option equates to the 0s and 1s needed for computing, and the cell’s memory is preserved with no additional energy applied—that is, the electrons keep moving as long as the wires stay superconductive.
Keeping nanowires at the cool conditions needed for superconducting—especially when they don’t generate any heat of their own—could prove far more affordable and far less space-intensive than current supercomputing solutions demand. And when things get smaller and cheaper, they open up whole new markets and usher in a wave of further innovations.
Putting a superconducting supercomputer onboard a self-driving car could become a reality, making them vastly more responsive, reliable and more affordable, both to build and to operate. Space-bound supercomputers could proliferate, with all kinds of academic and commercial applications. We’re not likely to see even a nano-sized cooling system in our smartphones anytime soon, but desktop supercomputers could be markedly reduced in upfront and long-term costs.
Quantum computing, which uses relatively low energy to change the quantum states of electronics to preserve memory, is undoubtedly going to hit the market first, and we’ll see some major leaps in computing power, size, and efficiency. Nanotech is keeping supercomputers on the map for a variety of applications, though, and we’re eager to see just how small things can get.