MRAM Moves to Higher Densities

Everspin introduced a 16-megabit MRAM product today focused on all applications requiring both SRAM performance and nonvolatile data storage. This latest product also addresses MRAM’s traditional high reliability and high-performance market segment, while advancing the technology to a higher density.

The new MR4A16B is a 3.3-volt parallel I/O non-volatile memory product featuring 35ns access times and unlimited read/write cycles. The new memory device is pin and function compatible with asynchronous SRAMs and specifically targets industrial applications, network and data storage systems, as well as other systems traditionally requiring SRAM-like performance. The new 16Mb-family of products supports both commercial temperature operating conditions as well as extended industrial temperatures with up to -40° C to +85° C operating range options. Production is scheduled for July of this year with samples immediately available.

Saied Tehrani, Everspin’s chief operating officer, said in the press release, “Everspin continues to rapidly expand its MRAM portfolio, enabling more customers to differentiate their products. Our product roadmap will deliver MRAM at increasingly higher densities, while maintaining MRAM’s unique characteristics in a cost effective way.”

Even as the discussions continue regarding if and when new memory technologies will compete with the cost-per-bit of high-volume production of DRAM or NAND, Everspin continues to grow MRAM production requiring a different set of performance and cost attributes. MRAM’s steady growth is mistakenly undervalued simply because the existing product line doesn’t indicate competition with DRAM or NAND in existing high-volume applications.

I find two questionable assumptions in that argument. The first assumption is that market conditions will remain unchanged and performance targets for high volume applications will not shift. Market conditions for memory products admittedly change slowly, but I believe that there is enough information available to validate that the trend-setting target applications are definitely shifting toward nonvolatile memory and away from the dominance of desktop PC applications.

The second assumption I question is that the historic rate of technologic progress will not change in the future. Postulating that existing memory technologies can continue the cost and performance trends indefinitely into the future is risky in light of the growing number of issues related to sub-20nm technologies. With previous announcements from IMFT about its 25nm NAND production and Samsung’s press release promising eminent 20nm NAND production, it still seems to me that existing charge storage memory technologies are rushing toward whatever challenges ahead at a faster rate than other segments of semiconductor production. Anticipating smooth sailing ahead while weathering historic declines in the bit cost of the current memory technologies just does not seem prudent.

This second assumption also contains a permutation of the tired argument that new memory technologies cannot compete with established memory technologies because the higher production of older technologies benefits from being so far ahead on the learning curve. But doesn’t that same argument suggest that the science and physics of any technology in production is likely accelerating at a faster rate than a technology not yet in production.

MRAM is in production and continues to advance in density and production volume while target applications increasingly view nonvolatile memories as the technology driver of the future. This sounds like a successful program to me.