SNIA MRAM alliance looks to unite a broader ecosystem to encourage MRAM adoption.
The momentum behind magnetoresistive random access memory (MRAM) continues to grow as the semiconductor industry aligns around the technology’s long-term potential.
A recent article published by EE Times explores the launch of the SNIA MRAM Alliance SIG and the broader effort to accelerate MRAM adoption across the semiconductor ecosystem.
Magnetoresistive random access memory (MRAM) has hit a tipping point, spurring the formation of a new group to help expand its use.
The MRAM Alliance Special Interest Group (SIG), recently launched within the Storage Networking Industry Association (SNIA), aims to unite a broad range of stakeholders, including foundries, chip makers, memory manufacturers, equipment suppliers, and system companies.
“We believe there is a window of opportunity for MRAM because it has reached a maturity level which is much higher than what it is today for resistive RAM or ferroelectric RAM,” Jean-Pierre Nozières, SIG co-chair, told EE Times in a briefing.
The new SIG is an expansion of an existing working group that already included several leading companies and research organizations in the field, he said, and the aim is to grow that membership to encompass the semiconductor value chain as well as several application verticals to improve ecosystem alignment.
STT MRAM is now well established as a core technology within major foundries, including TSMC, Samsung, UMC, and GlobalFoundries, Nozières said. “It’s intersecting with a real demand from the market side.”
Some demand signals are clearer than others. NXP is adding embedded MRAM to its automotive MCU platform, Nozières noted, and because it is generally much more radiation-tolerant than charge-based memories, MRAM is finding its way into space-bound systems.
But there are some preconceived notions about the emerging memory, he said, which in part led to the formation of the SIG.
Most notable is a concern about MRAM’s magnetic immunity, which Nozières described as “an elephant in the room.”
There is worry, especially in the consumer electronics segment, that some MRAM generations and package types can be upset by unusually strong external magnetic fields during handling, read, or write operations. Nozières said qualms remain about how devices such as wearables could be affected by magnetic fields.
Most STT-MRAM parts are rated well above typical ambient magnetic exposure, and vendors often add package-level shielding, so the practical risk for most devices is low in normal environments. “If you design it properly, this won’t be a problem at all,” Nozières said.
IEEE has a formal effort to standardize MRAM magnetic immunity testing, which specifies how to verify the maximum static magnetic field an MRAM device can tolerate while still meeting a target bit error rate for read, write, power-off, and standby modes.
Nozières added that MRAM, like any other memory, is not always suitable for the application at hand. “No technology can be deployed everywhere in every mission profile in every environment.”
He said the aim is to quell concerns over magnetic immunity with education, as well as tout MRAM’s advantages over other memory options, including flash and resistive random-access memory (ReRAM). Nozières said the SIG is “a bit of everything.” Aside from education and communication around MRAM benefits, the aim is to tackle some technical problems and, perhaps in the long run, generate some standards.
The value of being under the umbrella of SNIA is that it is an organization of end users, Nozières said, not just semiconductor players who understand MRAM’s capabilities. “It’s very important for us to get end users involved.”
MRAM is already making inroads into the microcontroller embedded world, including automotive, by replacing NOR flash, he said. But further work is needed to drive adoption across the data storage, hyperscale, and AI customer segments.
Nozières said MRAM deployed at the edge could alleviate memory bottlenecks in the data center by allowing more inference to be done locally. “You need fast, non-volatile, and infinitely endurant memory,” he said. “I don’t see any other emerging NVM today which has these three attributes.”
While standards aren’t the core focus of the MRAM Alliance SIG, Nozières said the development of a single JEDEC-like MRAM interface standard would make it easier for users to integrate MRAM in their system, just as they would drop any other type of memory. “Standards are an important piece of what we want to deliver.”