IBM's 'racetrack' could change the world of computer memory

Earlier this week IBM scientists had a sort of mini-expo at the Institute of Electrical and Electronics Engineers International Electron Devices Meeting. Among the stuff they demonstrated was their progress toward a new kind of memory that is both higher in capacity and faster than anything currently on the market.

Originally announced as a practical possibility in 2008, this “racetrack” architecture, once perfected, could change the computer industry for the better, which is, I suppose, a scientist’s job, after all. Ultimately, the work could lead to data-centric computing that would allow access to massive amounts of stored information just like that – in less than a nanosecond, IBM says.

This type of memory uses nanowires made of a nickel-iron alloy, which are not made the way larger wires would be, by drawing hot metal through smaller and smaller dies. These are far too small for that – about 20 nanometers thick.

Wires this small are actually made the same way integrated circuits are made – by depositing a layer of metal onto a silicon wafer, and etching away metal until only the wires are left. This is kind of like Michelangelo’s methods, but with microscopic metal instead of marble.

Pulses of electrons can be delivered through the wire by a write device to make electromagnetic “stripes,” which represent the data. When they are spinning one way, the resulting magnetic field is treated as a “1,” and when it is going the other way, it is a “0.” A current moves the stripes along the wire so a read device can read the stripes in succession, hence the racetrack moniker.

Data can be written to and read from each stripe in less than a nanosecond –a billionth of a second – and since the wires are so small, you can have quite a lot of them on one chip.  

The only limitation is how many of these stripes they’ll be able to cram on to each wire. For the recent demonstration, they had exactly one on each, but that was sufficient to demonstrate that the concept was working. Now it’s just a question of finding the metal alloy that has the right magnetic properties to maximize stripe capacity.

So, someday soon, we’ll be using computers with huge amounts of really fast memory. Maybe they’ll also be powered by grapheme-sheet batteries so we can really feel like we’re in the future. Of course, by then, we will be.

About the Author

Greg Crowe is a former GCN staff writer who covered mobile technology.

inside gcn

  • virtual reality hardware used by NASA to rehearse work on the International Space Station. (Photo by NASA)

    Accelerating virtual reality in government

Reader Comments

Fri, Dec 9, 2011 SoutheastUS

That means the next "bottleneck" for tremendously fast computing will, once again, be the interconnections between computers in a cluster or grid. The network bandwidth will become the bottleneck once memory access speeds up enough to be able to keep the CPUs busy. Now might be a good time to invest in companies that seem to be making great strides in the "next generation" of networking appliances (in terms of transmission latency and packet routing/switching speeds). Then when IBM's memory is deployed, the demand for network bandwidth will jump and the companies ready to supply the next generation of networking hardware will be poised to profit handsomely.

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above

More from 1105 Public Sector Media Group