Spin Memory, ARM, Applied Materials Ink Joint MRAM Agreement
In the world of alternative memory technologies, commercialization and broad adoption are eternally 3-5 years away. This is less a dig at the various companies developing technologies based on non-standard DRAM or non-volatile storage architectures and more an acknowledgment of how difficult it is to establish a new standard in the first place. Intel and Micron’s joint venture in 3D XPoint / Optane is the closest thing to a new RAM technology we’ve seen in-market in years, and Optane volumes are a fraction of, say, the NAND or DRAM markets.
With all of that said, a new joint agreement between ARM, Spin Memory (formerly known as Spin Transfer Technologies), and Applied Materials could signal a larger push into MRAM development in the future. Earlier this year, we discussed how Spin Memory had recently announced a breakthrough in MRAM endurance and performance that could push the standard into wider deployment. The problem with MRAM, historically, is that it offered excellent performance, data retention, and endurance — but not at the same time. Fast MRAM wasn’t very power efficient and didn’t last very long, while slow MRAM offered good power and endurance characteristics but low performance.
Spin Memory claimed in April to have solved this problem with a branded capability they’re now calling Endurance Engine. ARM is licensing the capability for inclusion into future products, according to the companies involved:
Under the licensing agreement, Spin Memory will provide Arm its innovative Endurance Engine design architecture to develop a new line of embedded MRAM design IP. This MRAM design IP will address static random-access memory (SRAM) application in SoCs, with denser and lower power solutions than typically achieved with the current 6T SRAM cell-based IP.
The Applied Materials collaboration is on the manufacturing side, where the two companies will collaborate to create an IP stack that allows customers interested in adopting MRAM to quickly deploy it for both non-volatile and SRAM-style replacements as well. This technology is expected to be available for licensing in 2019. MRAM is potentially interesting as an SRAM replacement, particularly on lower-power products, because it offers improved density and power scaling compared with relatively power-hungry SRAMs, which don’t necessarily scale all that well on FinFET processes to begin with.
This type of alternative approach — exploring differences in physical memory types or intrinsic tech advantages as opposed to simply relying on transistor scaling to provide additional benefits year on year — can be expected to increase as the benefits of scaling to new transistor nodes dwindles and the costs of developing on those nodes continue to increase. Manufacturers looking to deliver generational improvements can’t depend on node scaling any longer, and many low power and IoT devices aren’t necessarily built on leading-edge nodes anyway. That’s potentially important for technologies like MRAM, which don’t tend to use the latest nodes either.
Continue reading
Researchers Devise a ‘Honey Memristor’ for Neuromorphic Chips
Researchers have devised a way to use honey as a memristor, paving the way for sweet new forms of neuromorphic hardware.
New MRAM Breakthrough Could Revolutionize CPU Designs
A new breakthrough in MRAM could finally see the technology emerge as a competitor to SRAM.
Scientists Stuffed a Crocodile Into an MRI Machine to See If It Likes Bach
Do crocodiles like Bach? Hard to tell. Do they process Bach in equivalent neurological structures to those of birds and mammals? We've got evidence on that issue now.
MIT Neural Network Accelerates MRI Image Processing by 1,000 Times
It can take hours for a computer to match all the locations in a 3D map, but researchers from MIT have developed an algorithm that could cut that time to less than a second.