Racetrack Memory On the Move
Caption: A new design for comuter memory would send information shooting along wires like cars along a racetrack. Image source: Jina Lee, Wikimedia Commons.
The word racetrack has quite a history in American culture. It conjures up images of horse racing, with beautiful animals, colorful spectators, cheap hot dogs, and expensive gambling. Perhaps more prominently, it brings automobiles to mind. The Ford Model-T made cars affordable, and Prohibition catalyzed the development of brawny speedsters that could outmuscle the police cruisers, but it was the automobile’s marriage to the racetrack that finally transformed stock car racing into the booming phenomenon that it is today. Now a team of researchers at IBM lead by Stuart Parkin would like us to start associating the word with computer memory.
Racetrack Memory, as the research team has coined it, is a new idea that could compete with some of the most popular memory devices in use today. In an article published in the December 24th issue of the journal Science, Parkin and collaborators have measured some key features of magnetism that bring the idea one step closer to viability.
Devices today almost all store their memory using either FLASH or magnetic hard drives. Both technologies have undergone dizzying improvements recently in storage capacity, cost, and reading and writing speed. Still, they have a few inherent drawbacks. Traditional magnetic hard drives are inexpensive and have an enormous storage capacity. However, they require moving parts, which costs a device energy and consequently battery life. FLASH memory has no moving parts, but it takes a relatively long time to write information, and FLASH continues to be more expensive than magnetic hard drives.
Racetrack Memory is proposed to be a third alternative. Information would be stored magnetically in a similar manner to a magnetic hard drive. However, instead of storing the information in stationary bits on a disk, it would be stored in thin nanowires, and if the proper signals are applied, the information could be zipped through the wires from one place to another — like a car shooting down a racetrack — with no need for moving parts. IBM has big ambitions for the new memory architecture, claiming that it would enable a single portable device to hold more than 2,000 movies while running on a battery that would last for weeks.
Particularly exciting is the fact that the basics of Racetrack Memory are becoming rapidly understood. “In the past ~5 years we have demonstrated that the physics underlying the Racetrack Memory that I proposed in 2004 or so works!” Dr. Parkin said. “Thus, there are no fundamental roadblocks to making Racetrack Memory a reality.” This is a contrast with some of the more exotic ideas out there for computer memory (see my last blog post on multiferroics, for example).
So why can’t I seem to find a media player that doesn’t conk out after half a day’s use yet? Parkin said that while the physics of Racetrack Memory architecture works, many engineering hurdles still need to be overcome. He was unwilling to put a definite time estimate on when devices might begin to show up in the marketplace. “Of course,” he joked, “if you write a large check, we can build you a prototype Racetrack Memory in just two to three years (depending on the size of the check!).”