Freescale Semiconductor Inc. last week announced the commercial availability of a chip based on Magnetoresistive Random Access Memory (MRAM), a move that will significantly advance MRAM technology, although barriers to its mass adoption remain, iSuppli Corp. believes.

The company said its new MRAM device combines all of the desirable traits of today's incumbent memory technologies into a single 4-Mbit density device. MRAM stores data using magnetic polarization, as opposed to charge storage and maintenance in the memory bit cell. With MRAM, the binary state is detected as a change in resistance, rather than as a specified level of charge, as in other mainstream memory technologies such as SRAM, DRAM and flash.

Freescale said its new device has balanced read and write speeds of 35 nanoseconds (ns) — similar to a DRAM or a flash-memory device. The device also has a cell density similar to DRAM and flash, but with no leakage issues like DRAM. It also has unlimited endurance like DRAM or SRAM, combined with non-volatility, a key feature of flash.

MRAM as a technology has been around for several years, but has failed to achieve major commercial success. There are varying reasons for this, but two of the major causes have been MRAM's uncompetitive cost per bit and the difficulty in integrating the technology into standard CMOS processes.

Freescale has addressed the cost issue with the use of an innovative MRAM cell structure incorporating what it calls a "toggle" bit, which stabilizes a cell in either the one or the zero state without requiring additional control transistors, thereby offering an optimized bit cell solution.

The process integration issue was addressed by incorporating the MRAM module late in the process flow in order to minimize any effect on the standard CMOS logic processing. According to Freescale, the lack of disruption to standard semiconductor processes by the MRAM module will facilitate future MRAM technology versions as the base CMOS technologies advance.

iSuppli feels this initial commercialization of MRAM by Freescale is an impressive accomplishment and represents a significant advance. However, it does not yet solidly position MRAM for mass adoption.

Freescale has overcome some technology barriers that have plagued its competitors in the past. However, Freescale still has a long way to go to before its product is capable of being a mainstream memory.

iSuppli believes the real short-term value of Freescale's technology lies not in standalone memory replacement, where the competition over cost-per-bit is extreme. Rather, the opportunity for this product is to serve as a key element in upcoming System-on-Chip (SOC) designs.

As an SOC element, Freescale's technology can be used to incorporate sufficiently high-density memory onto highly integrated processing units and can be leveraged into solutions by Freescale and with licensing arrangements by others.

The International Technology Roadmap for Semiconductors says that today's SoC designs use up to 50 percent of their die area on embedded memory, normally SRAM. This figure is expected to surpass 70 percent in the not-too- distant future. Currently, the Freescale MRAM's read/write speeds of 35ns on 180nm technology are too slow for Level- One (L1) cache purposes, but could suffice for lower-level caching on some microprocessors. But as the technologies advance, performance should be elevated to L1-capable levels.

iSuppli certainly is not discounting the validity of the standalone memory potential of Freescale's MRAM product. However, its present pricing structure and relatively low density will keep it limited to niche applications, like battery- backed SRAMs, and for controlling mission-critical backup systems where its cumulative advantages of high performance, endurance and non-volatility make it a compelling alternative to today's high-priced solutions.

MRAM is only one of many technologies now under development that shows promise to become the mainstream memory of the future. Such technologies must offer high density, performance, endurance, non volatility and extremely low cost.

This achievement by Freescale drives MRAM beyond the "great potentialbut" category and places MRAM squarely in the fray to become a universally-accepted memory technology. It will be extremely interesting, as its MRAM development efforts continue, to follow not only Freescale's progress, but the reactions and responses to this product from its many competitors focused on the same long-term goal of developing a universal memory.