January 13, 2022
MRAM is non-volatile memory that retains information even if the device is turned off. Samsung Electronics, today demonstrated the world’s first functional computing system based on MRAM (magnetoresistive random access memory). We are facing a new paradigm that opens the door to a whole new generation of memory chips that can store data and process it.
The paper on this “in-memory computing” system was published online by the scientific journal Nature on January 12 (GMT) and will appear in the next print edition of Nature. Titled ‘A crossbar array of magnetoresistive memory devices for in-memory computing‘, this document shows the effort Samsung for merging memory semiconductors and next-generation artificial intelligence (AI) processors.
The research was led by the Samsung Advanced Institute of Technology (SAIT) in close collaboration with the Samsung Electronics Foundry Business and the Semiconductor R&D Center.
In standard computer architecture, data is stored on memory chips and data computation runs on separate processor chips. In contrast, in-memory computing is a new computing paradigm that seeks to perform both data storage and data computation on a memory network.
Since this scheme can process a large amount of data stored within the memory network itself without having to move the data, and also because the data processing in the memory network runs in a very parallel fashion, the consumption of power is substantially reduced. Therefore, in-memory computing has become one of the promising technologies to realize next-generation low-power AI semiconductor chips.
For this reason, research on in-memory computing has been intensively carried out all over the world. Nonvolatile memories, particularly RRAM (Resistive Random Access Memory) and PRAM (Phase Change Random Access Memory), have been actively used to demonstrate in-memory computing.
On the contrary, until now it has been difficult to use MRAM, another type of non-volatile memory, for in-memory computing despite the merits of MRAM such as speed of operation, endurance, and large-scale production. This difficulty stems from the low endurance of MRAM, because MRAM cannot enjoy the power reduction advantage when used in standard in-memory computing architecture.
Researchers at Samsung Electronics have provided a solution to this problem through an architectural innovation. Specifically, they succeeded in developing an MRAM array chip that demonstrates in-memory computing, by replacing the standard ‘current-sum’ in-memory computing architecture with a new ‘resistance-sum’ in-memory computing architecture, which addresses the problem of small resistors of individual MRAM devices.
Subsequently, the Samsung research team tested the performance of this compute chip on MRAM memories by running it to perform AI computing. The chip achieved 98% accuracy in classifying handwritten digits and 93% accuracy in detecting faces in scenes.
By bringing MRAM, memory that has already reached commercial-scale production embedded in system semiconductor manufacturing, into the realm of in-memory computing, this work pushes the frontier of next-generation low-power AI chip technologies. generation.
more like the brain
The researchers have also suggested that this new MRAM chip can not only be used for in-memory computing, but can also serve as a platform for offloading biological neural networks. This is in line with the neuromorphic electronics vision that Samsung researchers recently presented in a perspective article published in the September 2021 issue of the journal Nature Electronics.
“Computing in memory resembles the brain in that in the brain, computation also occurs within the biological memory network, or synapses, the points where neurons touch each other,” said Dr. Seungchul Jung, first author of the report. “Indeed, while the computation performed by our MRAM network for now has a different purpose than the computation performed by the brain, such a solid-state memory network may be used in the future as a platform to mimic the brain by modeling the brain synapse. connectivity.”
As highlighted in this paper, by building on its advanced memory technology and merging it with system semiconductor technology, Samsung plans to continue to expand its leadership in next-generation computing and AI semiconductors.
[+] Videos de nuestro canal de YouTube