Friday, July 01, 2011 ... //

IBM: phase-change memory, PCM

Phys Org (and those who follow them) informs about a new breakthrough of IBM that could lead to a two-order-of-magnitude improvement of flash-like memories. The technical paper is here:

Drift-tolerant multilevel phase-change memory (PDF)
The first author is Papandreou - clearly a chap from a family that brought Greece to the edge of bankruptcy and that is now trying to save it or at least delay its death by a year. How are the bits stored?

The basic idea is very simple. To "poke" a bit, an alloy is heated up and transformed to a low-resistivity crystallic phase or a high-resistivity amorphous phase. The resistivity is measured if you want to "peek" at the bit. I hope that some TRF readers have spent some time with BASIC, an old programming language.

Those systems of bits are much more durable than the flash memory. Also, the physicists are trying to store more than one bit at a place - which is possible because one may distinguish many levels of resistivity. On the other hand, the resistivity tends to "drift" over time. This unwelcome process is being fought against by storing the bits in a balanced way.

It can't be excluded that within 5 years, we will have memories that are 100 times faster and have many other advantages. But note that there have been several similar promises. For example, we're waiting for the amazing new generation of computers that are based on memristors.