Microchip emulates embedded SuperFlash technology

     Computing in memory technology is expected to eliminate a large number of data communication bottlenecks caused by artificial intelligence (AI) voice processing at the edge of the network, but it requires an embedded memory solution that can simultaneously perform neural network computing and storage weights. Microchip Technology Inc announced through its subsidiary TPV Semiconductor (SST) that its SuperFlash memBrain neuromorphological memory solution has solved this problem for WITINMEM neural processing SoC. This is the first mass-produced SoC, which can reduce voice noise and recognize hundreds of instruction words in real time immediately after the sub milliampere system is turned on.

     Microchip cooperates with Zhicun Technology to integrate Microchip's memBrain analog memory computing solution based on SuperFlash technology into Zhicun Technology's ultra-low power consumption SoC. The SoC uses memory computing integration technology for neural network processing, including speech recognition, voiceprint recognition, deep speech noise reduction, scene detection and health status monitoring. Zhicun Technology is currently cooperating with multiple customers and will launch products based on the SoC to the market in 2022.

     Wang Shaodi, CEO of Zhicun Technology, said: "Zhicun Technology is using Microchip's memBrain solution to solve the real-time AI voice computing intensive challenges at the edge of the network based on advanced neural network models. We took the lead in developing an all-in-one memory computing chip for audio in 2019, and now has achieved another milestone. We have adopted this technology in batch in ultra-low power neural processing SoC, simplifying and improving the voice processing performance of intelligent voice and health products.

     Mark Reiten, Vice President of SST Licensing, said: "We are very pleased that Zhicun Technology has become our main customer. We are grateful that Zhicun Technology has chosen to use our technology to launch such excellent products and enter the expanding artificial intelligence edge processing market. Zhicun Technology SoC has demonstrated the value of using memBrain technology to create a single chip solution based on an integrated memory and computing neural processor, which eliminates the traditional processors using digital DSP and SRAM/DRAM based methods to store and execute machines Problems in the learning model. "

     Microchip's memBrain neuromorphological memory product has been optimized to perform vector matrix multiplication (VMM) for neural networks. It enables processors for battery powered and deeply embedded edge devices to provide the highest possible AI reasoning performance per watt. This is achieved by storing the weight of the neural model as a numerical value in the memory array and using the memory array as a neural computing element. The power consumption is 10 to 20 times lower than other methods. At the same time, since external DRAM and NOR are not required, the overall bill of materials (BOM) cost of the processor is also lower.

     Permanently storing the neural model in the processing element of the memBrain solution can also support the instant opening function of real-time neural network processing. By making use of the non-volatile nature of the floating grid storage cell of SuperFlash technology, Zhicun Technology can further reduce the static leakage power consumption of demanding IoT applications by turning off the memory computing macro in idle state.