Researchers at the University of California at Berkeley have declared the problem of the high energy consumption required by today's high-performance computing hardware solved: by switching from electronics to magnetics in future processors.
Rather than the modern silicon-based processor, which shuffles electrons around its labyrinthine circuits to perform its calculations, the researchers propose a switch to nanometre-sized bar magnets in place of the more typical transistors. The result, they claim, would be a processor dissipating a tiny 18 millielectron-volts of electricity.
That's a million times less than the energy dissipated by today's processors, and the bottom limit allowed by the second law of thermodynamics - known as the Landauer limit - below which it is thought to be impossible to go.
"Today, computers run on electricity; by moving electrons around a circuit, you can process information,” explained UC Berkeley graduate student Brian Lambson. "A magnetic computer, on the other hand, doesn’t involve any moving electrons. You store and process information using magnets, and if you make these magnets really small, you can basically pack them very close together so that they interact with one another. This is how we are able to do computations, have memory and conduct all the functions of a computer."
The work of Lambson and his colleagues Jeffrey Bokor and David Carlton builds on the calculations of Rolf Landauer from the 1960s. Landauer used what was then cutting-edge information theory breakthroughs to calculate the minimum possible energy a ligical operation would dissipate given the limitation of the second law of thermodynamics. That law, for those who fell asleep during that particular lesson, states that an irreversible process dissipates energy that cannot be recovered.
In his work, Landauer calculated that the absolute minimum a circuit could dissipate and still perform and irreversible process would be 18 millielectron-volts, and it's this limit that the team is hoping to reach with the switch to magnetic computing.
The team's work involves the careful use of tiny nanomagnets to produce memory and logic devices just 100 nanometres wide and 200 nanometres long. While that's significantly above the size of a traditional electronic component - modern processors are built on a process size closer to 25nm than 200nm - it's a promising start for what is, at its heart, a fundamental rethink of what makes up a computing system.
Lambson's work has already proven that an operation on a simulated magnetic memory component can be performed close to or at the Landauer limit, and the team is now working on developing physical prototypes of its system in the hope of delivering a Turing-complete computing system that operates at a tiny power draw. Such a system would require little in the way of cooling, and draw a fraction of the power of even the most economical portable CPU on the market today.
Lambson isn't the only one investigating the possibility of magnetic computers, however. Back in 2006, researchers at the University of Notre Dame constructed a three-input majority logic gate from 16 coupled nanomagnets, and were able to demonstrate a logical operation successfully.
There's also the small matter of actually programming the nanomagnets themselves. Currently, electricity is used to generate a magnetic field in order to erase or change the polarity of the nanomagnets, resulting in a system which is many times less efficient than straightforward electronic circuitry and nowhere near the Landauer limit. That's a problem which is going to take time to overcome.
"We are working now with collaborators to figure out a way to put that energy in without using a magnetic field, which is very hard to do efficiently," Bokor admitted. "A multiferroic material, for example, may be able to control magnetism directly with a voltage rather than an external magnetic field."
Although it's going to be a while before we're running our smartphones on magnetic processors, the team is confident that the research solves real-world problems. "The magnetic technology we are working on looks very interesting for ultra low power uses," claimed Bokor. "We are trying to figure out how to make it more competitive in speed, performance and reliability. We need to guarantee that it gets the right answer every single time with a very, very, very high degree of reliability."
The team has published its results in a paper for the Physical Review Letters.