Infineon, IBM collaborate on new memory tech
The two companies will research phase-change memory, material that retains data by changing its structure.
The two companies, along with Taiwan's Macronix, will dedicate about 25 employees to research materials and structures that could be used for these devices. The work will take place at IBM's Yorktown Heights, N.Y., and Almaden, Calif., labs.
Phase change memory involves changing the underlying structure of a material, often by heating it rapidly, so that it's crystalline one moment and amorphous the next. When a laser light is shown on the spot, the reflection will differ, depending on whether the spot it was shown on was crystalline or not. A computer records the different reflections as a 1 or 0.
Current memories work by capturing electrons in a cell. Phase change memory has an advantage in that data doesn't depend on an electrical charge, which can vanish. On the other hand, phase change memories can wear out mechanically.
Researchers around the globe are searching for materials and structures that will enable their companies to get off the hamster wheel of Moore's Law, which predicts exponential growth in transistor and memory capacity every couple of years. By switching from making chips out of silicon, companies hope to reduce their manufacturing costs while devising chips that will be faster, consume less energy and fit into smaller spaces.
Chips made of these new types of materials are expected to come out over the next two decades; still, there is no guarantee that any of them will work in mass production. Memory will likely be one of the first markets where these new ideas will be tried.
But it's not easy. Intel has backed Ovonics Unified Memory, a form of phase-change memory, for years. Intel co-founder Gordon Moore predicted Ovonics would become popular 30 years ago, but it hasn't happened yet. Intel has delayed the release of Ovonic chips a couple of times. Philips also has a phase change program.
Infineon and IBM also collaborate on spintronics, a type of memory chip that reads data depending on the spin of electrons contained inside a material. Like Ovonics, spintronics has a future that has yet to translate to actual results.