The Next Step In RAM Tech

A note for all those who have wondered why computers take so long to start up: In five years—possibly less—expect to see “instant-on” systems that come to life the moment you hit the switch and that consume a lot less power than today’s systems.

Holding out that promise is a developing technology called magnetic random-access memory (MRAM) that could eventually replace today’s dynamic RAM (DRAM) as the standard memory in computing devices.

Derived from a technology that’s already used in storage devices, MRAM is in various stages of development by several big-name vendors, including IBM, Motorola Inc., Honeywell International Inc., Hewlett-Packard Co. and Intel Corp.

Driving much of the companies’ interest is the potential size of the market opportunity – projected to be approximately $40 billion by 2005, according to Fred Zieber, an analyst at Pathfinder Research Inc. in San Jose.

Gauging Its Potential

“MRAM certainly holds tremendous promise,” Zieber says. “There is a way to go before you see MRAM in [commercial systems].” But if the technology shapes up right, he predicts, “it will replace DRAM.”

MRAM depends on magnetic polarity to store data, rather than on electricity like DRAM does, explains Stuart Parkin, a scientist at IBM’s Almaden Research Center in San Jose and a pioneer in MRAM research.

In DRAM, a memory bit is stored as an electrical charge on a capacitor. Because this charge on a DRAM chip constantly leaks, it needs to be supplied with a near-continuous electrical current to refresh the memory. When the power is turned off, DRAM chips lose their data, Parkin says.

In contrast, MRAM bits are stored in thin magnetic layers. Since the memory bits are stored magnetically rather than as charges, the memory is nonvolatile and doesn’t need power to retain its data.

The difference is crucial in how computers come to life when they’re powered up, analysts say.

In current-generation computers, the operating system and all application software are stored on the hard disk. When a computer is powered up, it loads a working copy of the operating system along with any other start-up software to the DRAM, from which the data can be quickly accessed by the microprocessor. This boot-up process can take several minutes.

MRAM devices, on the other hand, retain data without using electricity. Because working copies of the operating system and other applications can be stored on MRAM, it eliminates the need for lengthy boot-up times.

Flash memory—the kind found in digital cameras, for instance—already offers this kind of instant-on capability. But those memory devices are far too slow and degrade too quickly to be of much use in a commercial computing environment, says Parkin.

Another kind of memory available today is static RAM (SRAM), which provides faster access to data than DRAM does. Though it needs power to retain data, it doesn’t need it constantly like DRAM does. But SRAM chips are more expensive than DRAM chips and therefore aren’t ideal for mass commercial markets, says Parkin.

MRAM devices are considerably cheaper to manufacture than semiconductor-based DRAM and SRAM technology. MRAM is also expected to substantially reduce the battery power drain for portable devices because it doesn’t need to be constantly refreshed like DRAM does, say analysts.

“You are looking at [mobile] computers that can go days and weeks, and cell phones going perhaps even months on standby operation” with MRAM, says Rick Doherty, an analyst at The Envisioneering Group in Seaford, N.Y.

So essentially, the benefits of MRAM technology include the small size of DRAM, speed and performance comparable to that of SRAM, and the nonvolatility and inexpensiveness associated with flash memory, Parkin says.

MRAM Basics

MRAM devices, just like those with DRAM technology, are solid-state devices with no moving parts.

At its most basic, an MRAM device consists of two layers of magnetic material separated by a thin, nonmagnetic metallic layer through which electrons can tunnel. A set of parallel conducting lines are laid on the top half of this ferromagnetic sandwich.

A similar set of conducting lines are laid perpendicularly to the first set on the bottom half of the sandwich, resulting in a grid of conducting lines.

Each point where the bottom and top lines cross represents a bit. Data can be written on or read from such a device by passing currents through wires above and below the device. Once data is entered into the bit, the bit retains the data until the system erases or rewrites it, Parkin explains.

Limited production of such memory devices could begin in high-end systems in about two years, with mass production starting in about five years, predicts Doherty.

What companies are trying to do is “wed the speed of semiconductor memory with the nonvolatility of magnetic materials,” Doherty says. “You are talking about the possibility of a unified memory that combines the best of flash, DRAM and SRAM within a single device.”