Samsung has just revealed the world’s first in-memory computing based on MRAM (Magnetoresistive Random Access Memory). The team described these innovative findings in a paper published by Nature.
Attempting to merge memory and data computing is not a new idea, but it has never been implemented with the use of MRAM before. This breakthrough MRAM breakthrough is the result of Samsung’s effort to combine memory and system semiconductors into one, all with artificial intelligence in mind.
In Press release Which describes Samsung’s research, the company explains, showing that memory chips are typically used to store data, while data computing is executed in separate processor chips. In-memory computing is a technology that performs both functions within a memory network.
The use of in-memory computing allows the processing of large amounts of data that is stored within the memory network itself. This means that data does not need to be transferred and is processed in a parallel manner, resulting in faster transfer speeds and significantly lower power consumption.
In-memory computing has been attempted with the use of other non-volatile memory types, notably resistive random access memory (RRAM) and phase-change random access memory (PRAM). However, the use of MRAM has proved difficult, and Samsung explains that this is due to the low resistance of this type of memory. When used in standard in-memory computing architectures, MRAM does not benefit from the desired power reduction.
Despite the difficulties, MRAM remained an attractive option for in-memory computing due to its various advantages, such as high operation speed, the fact that it is mass-produced, and its stamina. This encouraged Samsung researchers to continue with the technology and eventually led them to a breakthrough that made MRAM-based in-memory computing a reality.
Addressing the above problem of low resistance presented by individual MRAM devices, the research led to the development of MRAM array chips capable of in-memory computing. The array chip was then tested by performing various AI computing tasks, and it performed excellently. Samsung reports that the chip hit 98% accuracy in categorizing handwritten digits and 93% accuracy in being able to tell faces apart from scenes.
The research that led to this discovery was the result of a collaboration between the Samsung Advanced Institute of Technology, the Samsung Electronics Foundry Business, and the Semiconductor R&D Center.
Using MRAM for in-memory computing is certainly a big step forward for Samsung. This technology could be one of the most viable options for the next generation of power-efficient AI processing. Samsung said it plans to move forward and continue to build on its MRAM in-memory computing success.