TORONTO – Its name implies speed, and as a memory technology, Racetrack holds a great deal of promise in terms of performance, but IBM’s experiment is slow leaving the lab.
EE Times caught up with Dr. Stuart Parkin, an IBM Fellow and manager of the magnetoelectronics group at the IBM Almaden Research Center in San Jose (see Figure). Parkin, who is also director of the IBM–Stanford Spintronic Science and Applications Center, has been working on the principles of Racetrack since 2002.
Dr. Stuart Parkin, an IBM Fellow and manager of the magnetoelectronics group at the IBM Almaden Research Center in San Jose, has been working on the principles of Racetrack since 2002.
Racetrack is also known as domain-wall memory (DWM) and is non-volatile. It uses a spin-coherent electric current to move magnetic domains along a nanoscopic permalloy wire about 200 nm across and 100 nm thick. As current is passed through the wire, the domains pass by magnetic read/write heads positioned near the wire, which alter the domains to record patterns of bits.
In an interview from Germany, where he also serves as director of the Max Planck Institute for Microstructure Physics, Parkin said a 3-bit version was demonstrated in 2008 and two years ago, the first fully integrated Racetrack prototype memory was built with 90nm CMOS- based technology with integrated circuits for reading, writing, and moving the domain walls. “We’ve basically proven in that work that the fundamental concept works and that we can build a prototype,” he said.
Within the past three years, Parkin said an entirely new mechanism for moving magnetic information in the Racetrack has been developed. “It’s extremely exciting because it works even more efficiently than by how we were first moving the domain walls,” he said. “The fundamental principle is the same.”
What’s different, Parkin explained, is magnetization is perpendicular to the Racetrack and no longer in its plane. “It enables you to make the domain walls much smaller,” he said, perhaps as much as 50 times narrow, so now the domain walls can be packed densely with each other and can be moved at greater speeds from the same current.
Some analysts view Racetrack as a storage replacement for both spinning disk and solid state drives, but Parkin sees Racetrack as memory that could have a wide range of applications, including replacing disk drives as well as well as typical memories, such as most DRAM. “It’s very versatile,” he said. “It’s more like a memory than storage device, but some people call it a memory storage device.” In general, Parkin said Racetrack could make systems much simpler, more compact and require less energy.
While Parkin emphasized that Racetrack is memory, not storage, Gartner analyst Brady Wang said it shows potential for replacing hard-disk drives, as well as NAND flash memory. One of the benefits of Racetrack, he said, is it solves the problem of the problem of leakage and wear-out that is common to solid state storage and provides greater endurance and better scalability.
Wang said it’s too early to judge the likelihood of Racetrack’s commercial success, but if it remains true to the theory and can be moved out of the lab, it will provide huge improvements to the speed and capacity of current memory technologies.
Jim Handy, principal with Objective Analysis, said Racetrack does have the potential to be a significant technology someday, but the question is how far can it get? “Will it compete on cost with some of the other memories?”
Once advantage Racetrack may have is that it should entail a lot less complicated process than 3D NAND, said Handy, but like many emerging memory technologies that have been around for a while, including some that date back to the 80s, he said it could be a very long time until Racetrack sees the light of day.
Parkin said Racetrack is at the point where it needs investment to carry out engineering and integration for the next prototype with vertical Racetracks, which is necessary to achieve ultimate density.