The Future of Computer Memory: MRAM v. FRAM v. OUM

Not long ago, solid-state memory was slow, limited in capacity and, worst of all, volatile (whatever data the memory contained would vanish when the power went off). Volatility spelled vulnerability. The earliest PDAs, left untended and unplugged, eventually lost everything?files, the operating system, applications and whatever else resided in the memory-on-a-chip that permitted the device’s compact form. Then came Flash memory, which wrought a revolution in pocket computing. The new memory held data indefinitely without the need for external power. Best of all, it was cheap. As a result, Flash provided the technical foundation for today’s PDAs and smart phones, and by extension, for the whole mobile data industry.

But Flash is relatively slow compared with the volatile memories it replaced; it degrades over time and uses considerable energy?hardly an ideal situation for products often powered by a couple of batteries. Those limitations quickly prompted the search for successors.

Three such next-generation technologies are currently vying for the huge Flash market. MRAM (magnetic read only memory), FRAM (ferroelectric RAM) and OUM (ovonic unified memory) all claim similar performance advantages: low voltage operation, long life span and very high speed. They reach those goals in different ways, though in each case an innovative material technology lies behind the performance breakthroughs.

MRAM, pioneered by Motorola and IBM, stores bits of data as changes in electrical resistance in the memory cells?produced by exposing certain exotic materials to magnetic fields. OUM, invented by researchers at Ovonyx, in Troy, Mich., also stores data as variations in resistance but creates those variations by changing the crystalline structure of substances known as chalcogenide alloys. FRAM, developed by Colorado Springs, Colo.-based Ramtron, and licensed to Fujitsu, Hitachi, Texas Instruments and Toshiba, is significantly different from the other two: It stores bits as fixed electrical potentials (voltages) within ferroelectric materials.

Only FRAM is produced today, though the low memory capacity of current chips (64KB) has restricted it to telemetry applications. MRAM has already surpassed FRAM in capacity, achieving a 256KB figure a scant three years into its development cycle, but fabrication issues still need to be sorted out. OUM’s prospects are even less clear. While its basic materials technology has already proven successful in recordable optical discs, the one major chip manufacturer committed to OUM development?Intel?isn’t saying much.

Industry analysts generally believe that MRAM has the edge, though they admit that FRAM and OUM are still in the running. But everyone agrees that perfecting any of them would have significant consequences, effectively obsolescing Flash, permitting the fabrication of complete systems-on-a-chip (making for even smaller devices) and ultimately even replacing all but the fastest solid-state memories.