memBrain solves edge processing challenges

Computing-in-memory technology is poised to eliminate the massive data communications bottlenecks associated with running Artificial Intelligence (AI) voice data in edge devices, but requires an embedded memory solution that simultaneously performs neural network computation and data storage.

Microchip Technology, through its Silicon Storage Technology (SST) subsidiary, announced that its memBrain SuperFlash neuromorphic memory solution has solved this problem for the WITINMEM neural processing SoC, the first in volume production that enables sub-mA systems to reduce background noise and recognise hundreds of command words, in real time and immediately after power-up.

Microchip has partnered with WITINMEM to incorporate memBrain's analogue in-memory processing solution, based on SuperFlash technology, into WITINMEM's ultra-low power SoC.

The SoC features computing-in-memory technology for neural network processing, including voice recognition, voiceprint recognition, deep speech noise reduction, scene detection and health monitoring.

WITINMEM, in turn, is working with multiple customers to bring products to market during 2022 based on this SoC.

memBrain is optimised to perform vector matrix multiplication (VMM) for neural networks.

It enables processors used in battery-powered, deeply embedded edge devices to deliver the highest possible AI inference performance per watt.

This is achieved by storing the weights of the neural model as values in the memory array and using the memory array as the neural calculation element.

The result is 10 to 20 times lower power consumption than alternative approaches together with lower overall processor bill of materials (BOM) costs because no external DRAM and NOR are required.

The permanent storage of neural models within the processing element of the memBrain solution also supports the instant-on functionality for real-time neural network processing.

WITINMEM has taken advantage of the non-volatility of SuperFlash technology's floating gate cells to power down its computing-in-memory macros during the idle state to further reduce power dissipation in demanding IoT use cases.

Comments

"WITINMEM is breaking new ground with Microchip's memBrain solution to meet the high-intensity requirements for real-time AI speech processing in edgesystems based on advanced neural network models," said Shaodi Wang, CEO of WITINMEM. "Wewere the first to develop a computing-in-memory chip for audio in 2019 and now we have reached another milestone with volume production of this technology in our ultra-low power neural processing SoC that optimises and improves speech processing performance in intelligent voice and health products."

"We are excited to have WITINMEM as our primary customer and are pleased with the company for entering the expanding AI edge processing market with a superior product using our technology," said Mark Reiten, vice president of the license division at SST. "The WITINMEM SoC demonstrates the value of using memBrain technology to create a single-chip solution based on a neural computing-in-memory processor that eliminates the problems of traditional processors that use digital DSP and SRAM/DRAM-based approaches for storing and executing machine learning models."

For further information click here