Microchip Technology Inc., through its subsidiary Silicon Storage Technology (SST), has announced that its SuperFlash memBrain neuromorphic memory solution has solved this problem for the WITINMEM neural processing SoC, the first to be mass-produced to enable systems with sub-1 mA current draw to reduce speech noise and recognize hundreds of instruction words in real time, immediately after power-up.
Microchip has worked with WITINMEM to incorporate Microchip's memBrain analog memory embedded computing solution, based on SuperFlash technology, into WITINMEM's ultra-low-power SoC. The SoC is based on embedded memory computing technology for neural network processing, including speech recognition, voiceprint recognition, advanced speech noise reduction, scene detection, and health monitoring. WITINMEM is working with several customers to bring their products based on this SoC to market in 2022.
“WITINMEM is breaking new ground with Microchip’s memBrain solution to meet the computationally intensive requirements of real-time AI-powered voice processing at the network edge, based on advanced neural network models,” said Shaodi Wang, CEO of WITINMEM. “We were the first to develop an in-memory compute chip for audio in 2019, and now we have achieved another milestone by mass-producing this technology in our ultra-low-power neural processing SoC, which speeds up and improves the performance of voice processing in smart products for voice and healthcare applications.”
“We are pleased to be collaborating with WITINMEM and are excited for the company to enter the expanding AI edge processing market with a superior product that uses our technology,” said Mark Reiten, Vice President of Licensing Vision at SST. “WITINMEM’s SoC demonstrates the value of memBrain technology in creating a single-chip solution based on a neural processor with in-memory computing that eliminates the problems of traditional processors that rely on digital DSP and SRAM/DRAM-based techniques to store and execute machine learning models.”
Microchip’s memBrain neuromorphic memory product is optimized for vector matrix multiplication (VMM) in neural networks. It enables processors used in battery-powered, embedded, and advanced edge devices to deliver the highest possible AI inference performance per watt. This is achieved by storing the neural model weights as values in the memory array and using the memory array as the neural computing element. The result is 10 to 20 times lower power consumption than other alternative solutions, as well as lower processor bill of materials costs because external DRAM and NOR chips are not required.
The permanent storage of neural models within the memBrain solution's processing element also incorporates real-time neural network processing capabilities. WITINMEM leverages the non-volatile nature of SuperFlash floating-gate cells to deactivate its integrated memory computing macros during idle states, further reducing power leakage in demanding IoT applications.
Bulgin Buccaneer circular power connectors for harsh environments #circularconnector #bulginbuccaneer
Available from stock through Powell Electronics are several series of Bulgin's popular Buccaneer range of sealed, circular power connectors for harsh environments.
The Standard Buccaneer range is a line of waterproof connectors. Rated IP68 and IP69K, these...
