Samsung Electronics has successfully demonstrated the world’s first in-memory computing based on MRAM (Magnetoresistive Random Access Memory) and publicized it on the Nature, the company said on Thursday.
The new achievement showcases the tech giant’s leadership in memory chip technology and its commitment to merge memory and system semiconductors for next-generation artificial intelligence (AI) chips.
In-memory computing is a new computing paradigm to enable both data storage and computing in a memory network. Since this architecture can process a large amount of data within the memory network in a highly parallel manner without having to move the data, power consumption is substantially reduced. For this reason, in-memory computing has emerged as one of the promising technologies to realize next-generation low-power AI semiconductor chips. In the standard computer scheme, data is stored in memory chips and computing is executed in separate processor chips.
In-memory computing has been intensely studied worldwide with a focus on the use of non-volatile memories, in particular RRAM (Resistive Random Access Memory) and PRAM (Phase-change Random Access Memory). But it has so far been difficult to use MRAM, another type of non-volatile memory, for in-memory computing due to its low resistance despite MRAM’s merits such as high data stability and operation speed.
Samsung Electronics researchers challenged this problem by replacing the standard, ‘current-sum’ in-memory computing architecture with a new, ‘resistance sum’ in-memory computing architecture.
In a subsequent performance test, Samsung’s MRAM in-memory computing chip achieved an accuracy of 98 percent in classification of hand-written digits and a 93 percent accuracy in detecting human faces from scenes.
This research is significant in that it can expand the frontier of the next-generation low-power AI chip technologies by ushering MRAM into the realm of in-memory computing.
The researchers also suggested that this new MRAM chip not only can be used for in-memory computing, but it also can serve as a platform to download biological neuronal networks.
The research was led by Samsung Advanced Institute of Technology (SAIT) in collaboration with Samsung Electronics Foundry Business and Semiconductor R&D Center. The first author of the paper is Dr. Jung Seung-chul, Staff Researcher at SAIT, and the co-corresponding authors are Dr. Ham Don-hee, Fellow of SAIT and Professor of Harvard University and Dr. Kim Sang-joon, Vice President of Technology at SAIT.
[ⓒ Pulse by Maeil Business Newspaper & mk.co.kr, All rights reserved]