At the starting of 2022, Samsung Electronics revealed that it has realized the world’s first MRAM-based in-memory computing. The new development is in the form of the Human Brain is even capable of interference.
The company is expected to be able to keep ahead in the fast-growing AI chip industry with this new AI chip technology. In-memory computing, the work is to store data and carry out calculations with the chip at the same time.
By using the technique, a large amount of information in memory is computed in parallel in the memory without any movement, and as a result to which the power consumption is reduced to a large extent.
It is very useful in developing low-power consumption chips. As we all know the global demand for this kind of Intelligent chip is soaring now, as the technology is largely being used by the industry for a wide variety of applications.
According to some popular blogs the average annual growth of the global AI chip market for the period of 2022 to 2025 is estimated to be over 30 percent.
Advantage of MRAMs
- The MRAMs provide a high level of data stability
- High processing speed
- Low resistance value hindering the realization of in-memory computing
Samsung Electronics successfully achieved a low power consumption design by overcoming the limitation and have reached the accuracy level of 93 to 98 percent in the number classification and face detection, respectively.
The Samsung research is particularly meaningful in the above-mentioned particulars and is the world’s first in-memory computing-based realization of MRAMs that can be mass-produced in the combination with the non-memory chip manufacturing process.
Samsung Electronics has earlier suggested the possibility of the MRAM chip platform being used as a neuromorphic platform in near future.
Neuromorphic computing includes the usage of these MRAMs that provide us with the capability to store, compute and analyze the data that works similar to neurons in the human brain. The benefits of such computing are
- 10x faster processing capability
- 60x more inter-chip bandwidth
- One million neurons flowing with 15x greater resource density
- 3D scalable and have Ethernet support
- Enhanced learning and adaptation capabilities
This will be further used for next-generation intelligent devices and autonomous systems. Samsung has been already working on the prototype, and we could soon expect a launch of a human-AI Chip, by the end of the year 2022, but nothing has been cemented from its side and is still uncertain where they are going to apply it