Advances in computing and storage technology continue to shape how we will handle information in the form of electronic data. Those that remember the first electromagnetic storage technologies, called "magnetic core memory" that in the early computers of the '60s and '70s used magnetic fields created by iron rings, may appreciate where random-access memory (RAM) is headed. Back then, you could pull the memory board and set it on a shelf; then come back six months later and have that board retain the data (without having to maintain a continuous power supply). Core memory has long since vanished, replaced by flash, EEPROM and other memory technologies. Yet, in more recent times, those past principles of magnetic cores are coming full circle and may alter computing and storage once more.
This relatively new memory technology could sidestep all current RAM technologies. The low-power, nonvolatile micromemory-cell technology is called magnetoresistive random access memory or MRAM. In concept, when a metal that shows a slight change in electrical resistance is placed in a magnetic field, it is referred to as "magnetoresistive." Data is stored in MRAM by using magnetic charges instead of the electrical charges used by dynamic random-access memory (DRAM). In MRAM, data is stored without electricity by employing the principle of magnetic polarity to store data.
NO FLASH IN THE PAN
MRAM research began around 1995 when the U.S. Defense Advanced Research Projects Agency (DARPA) funded the private sector to develop a general-purpose memory with high-density, high-speed and low-power consumption. The consortia, headed by IBM, Motorola and Honeywell, among others, employed two sciences in the development of MRAM: spin electronics, the science behind giant magnetoresistive heads used in disk drives and tunneling magnetic resistance, or TMR. The latter is expected to be the basis of future MRAM technology, which is said to be much faster and less expensive to make than today's nonvolatile Flash memory. Flash memory cells are damaged each time a bit is written in the memory; because of this, Flash, which is well-suited for consumer electronics, has a life of about 10,000 read/write cycles before the memory cells fault or fail altogether. MRAM, which is a better choice for desktop or mission-critical systems, could last indefinitely.
Besides retaining its state when power is removed, the one-transistor, one-magnetic tunneling junction (1T-1MTJ) architecture exhibits no degradation in resistance as it moves beyond 10 billion read/write operations. Researchers believe that accelerated testing of MRAM shows in excess of 10 years mean-time-to-failure, and is likely to exceed the numbers achieved by Flash and ferroelectric memory.
MRAM uses smaller-sized memory cells and functionality is simpler to manage when compared to other nonvolatile storage technologies. Charge pumps, necessary with Flash, are eliminated in MRAM. With MRAM's very fast read and write times - on the order of a few tens of nanoseconds - opportunities abound. Already there are some one sub-1-MB chips developed as a three-volt MRAM and feature address access times of about 15 nanoseconds. Motorola, for example, in June 2002 demonstrated a 1 MB MRAM chip with a 50-nanosec access and program time. Serious production is expected sometime in 2003 and, if successful, we should see the transition to MRAM in full swing around 2004.
BRICK WALL
Among the numerous opportunities offered by using magnetic memory storage are the nonvolatile storage of critical boot data - potentially creating "instant-on" computing, portable systems with less power consumption and the ability to retain the OS, all your working data and even some of the applications in the event of a crash or other power-critical failover. Embedded systems for portable communications (including cellular telephones) seem to be another set of applications for MRAM.
Based on parallel advances in RAM technologies, we find that the amount of data that can be stored on a square inch of "surface" continues to increase. Bit density - called areal density in recording technology terms - has grown tremendously, with advances in magneto-resistive (MR) and giant MR (GMR) heads, surpassing the areal density of optical recording more than seven years ago. In the one year between 1999 and 2000, areal density (in the laboratory) went from 35.3 GB to 63 GB per square inch and headed toward 100 GB per square inch in 2001 - resulting in IBM's claim of a greater than 400 GB single disk drive in the not-too-distant future.
However, the physical limits of magnetic areal density are expected to be reached in the next five years, given the 60 percent compound growth rates experienced in the previous three to five years. Eventually the magnetic "brick wall," known as the super-paramagnetic limit, will be reached whereby the magnetic particles become so closely packed that they actually interfere with each other. At that point, a new technology known as atomic-force microscopy (AFM), sometimes called "probe-recording" may take over, whereby read/write densities could reach 400 GB per square inch. AFM uses a plastic substrate and a superheated probe to mark small pits in the surface, but demonstrations so far have been confined to write-once/read-many (WORM) capabilities only.
Still, there is yet another hold card in the hand of magnetic recording. Disk drive technology currently employs longitudinal recording, which is principally how data is deposited and read back from the tracks of a revolving disk surface. Bits recorded on the media must meet a strong signal-to-noise ratio, thereby imposing limitations in concert with the areal density equations. A new and different approach, called "perpendicular recording," could push areal densities even further. A switch to perpendicular recording could most likely occur when the 100- and 200-GB-per-square-inch areal density is reached. This form of recording aligns the magnetization of the disk perpendicularly to the surface (i.e., standing the bits on end), producing higher magnetic fields by depositing the media on a soft magnetic layer. This process doubles the recording field by producing an image of the recording head on the surface and creating regions of upward- or downward-directed magnetization that in turn represent the digital 1's and 0's of the data.
Researchers believe that when the combinations of technologies, such as GMR and perpendicular recording - before near superparamagnetization limits are approached - we could see a single 3.5-inch disk drive storing in excess of 1 terabyte of data.
