Microchip cooperates with Korean intelligent hardware company IHWK to develop an analog computing platform to accelerate edge AI/ML inference

【Lansheng Technology News】In order to adapt to the rapid development of artificial intelligence computing and related inference algorithms at the network edge, Korean intelligent hardware companies are developing neuromorphic computing platforms for neurotechnology devices and field-programmable neuromorphic devices. Microchip Technology Inc., through its subsidiary TPV Semiconductor, participated in the development and provided an evaluation system for the platform’s SuperFlash? memBrain? neuromorphic memory solution. The solution is based on Microchip's industry-proven non-volatile memory (NVM) SuperFlash technology and is optimized to perform vector matrix multiplication (VMM) for neural networks through simulated memory computing methods.

The memBrain Technology Evaluation Kit is designed to allow IHWK to demonstrate the sheer energy efficiency of its neuromorphic computing platform for running inference algorithms at the edge. The ultimate goal is to create ultra-low-power analog processing units (APUs) for applications such as generative AI models, autonomous vehicles, medical diagnostics, speech processing, security/surveillance, and commercial drones.

Since current neural network models used for edge inference can require 50 million or more synapses for processing, purely digital solutions require sufficient bandwidth to use off-chip DRAM, which creates a bottleneck for neural network calculations and reduces overall computing power. . In contrast, the memBrain solution stores synaptic weights in on-chip floating gates in ultra-low-power sub-threshold mode and uses the same memory cells for calculations, significantly improving energy efficiency and reducing system latency. Compared with traditional digital DSP and SRAM/DRAM-based methods, this solution can reduce the power consumption of each inference decision to 1/20 to 1/10, and can significantly reduce the overall bill of materials.

To develop the APU, IHWK also collaborated with the Korea Advanced Institute of Science and Technology in Daejeon for device development and Yonsei University in Seoul to provide device design assistance. The final APU is expected to optimize system-level algorithms for inference and run between 20-80 TeraOPS per watt, achieving the best performance for an in-memory computing solution designed for battery-powered devices.

Mark Reiten, vice president of SST, Microchip's technology licensing business unit, said: "By using mature NVM to perform neural network calculations and store weights instead of alternative off-chip memory solutions, Microchip's memBrain in-memory computing technology has the potential to eliminate the need to perform artificial intelligence at the edge of the network. Handling the large-scale data communication bottlenecks faced. Cooperation with South Korea's leading R&D company IHWK, relevant universities and early users further proves the advancement of Microchip's neural processing technology and will also promote Microchip's development in the field of artificial intelligence."

Lansheng Technology Limited, which is a spot stock distributor of many well-known brands, we have price advantage of the first-hand spot channel, and have technical supports.?

Our main brands: STMicroelectronics, Toshiba, Microchip, Vishay, Marvell, ON Semiconductor, AOS, DIODES, Murata, Samsung, Hyundai/Hynix, Xilinx, Micron, Infinone, Texas Instruments, ADI, Maxim Integrated, NXP, etc

To learn more about our products, services, and capabilities, please visit our website at https://www.lanshengic.com