You've probably heard of flash memory, a non-volatile form of computer storage that's used primarily for memory cards, USB flash drives and solid-state drives. It doesn't take up as much space as a hard drive and can allow people to store a lot of data in a small space. But here's the down side of flash memory: The laws of physics hinder how small a flash memory drive can be. It has to do with the ability of those materials to trap electrons.
Scientists are investigating a different kind of memory that gets even smaller. It relies on phase-change materials (PCMs), which are heat sensitive. Instead of ones and zeros being stored in the electrons, they're stored by changing the material resistance, which moves from low to high as the temperature rises. Samsung and Micron Technologies both unveiled devices using PCM last year, but it hasn't been widely adopted yet.
A new data-encoding scheme could cut down the amount of power that PCM memory uses, potentialy making this kind of memory more widely available.
The method, invented by a team from UCLA and Rice University, uses a specially designed algorithm to read the bits in groups and change only those that need to be changed. For example, of a string of bits is 0110 and one wants to make it 0111, only the last bit needs to be different. But ordinarily a computer will change all four bits, even though it's only changing three of them to the same value. It's similar to editing an audio tape by re-recording the whole thing even though the only change needed is a few seconds at the end.
Lead researcher Farinaz Koushanfar, assistant professor of electrical and computer engineering and of computer science at Rice, told Discovery News that flash memory can't pick out individual bits at all, so the amount of energy needed to change anything is larger. In PCM, since one can limit the number of bits changed, it's possible to cut down the amount of power necessary.
"We're not identifying a new technology," she said. "We're optimizing an existing one."
Besides reducing the energy used by as much as 30 percent, the new method also boosted the number of cycles that a given memory cell could go through before becoming exhausted.
The algorithm is also applicable to other new memory technologies, such as spin-transfer torque random-access memory, which stores bits in magnetic fields rather than trapped electrons.
Image: Rice University