Computing-in-memory Powering Smart Wearable Applications

On August 18th, the 2022 Asia Smart Wearable Exhibition hosted by I Love Audio Network ended successfully at Shenzhen Futian Convention and Exhibition Center. At the three-day cutting-edge audio technology event, WITMEM demonstrated the powerful advantages of the computing-in-memory chip technology and the WTM2101 chip in intelligent voice applications to more than 100 smart wearable, Bluetooth audio, and auxiliary listening companies. At the Asian Smart Wearable Conference held at the same time, Wang Shaodi, founder and CEO of WITMEM, shared in detail the outstanding performance of WTM2101 in computing power, power consumption, energy efficiency and application with the title of "Integration of computing-in-memory to Help Smart Wearable Applications" , to help more AI applications to be implemented, and to upgrade the experience of smart wearable products.









As user-side requirements become more and more diversified and subdivided, smart wearable products have higher and higher requirements for computing power and power consumption, and the problem of storage walls under the traditional Von Neumann architecture has become increasingly prominent. WITMEM has been deeply involved in storage and computing integration technology for nearly 10 years, and innovatively uses Flash memory to complete the storage and computing of neural networks, which can perfectly solve the problem of AI storage walls, improve computing efficiency and reduce costs.



In January this year, after WITMEM Technology mass-produced the world's first computing-in-memory SoC chip WTM2101, the first to enter the smart wearable market, this market has also become an important place to test the achievements of computing-in-memory chips. WTM2101 has both ultra-low power consumption and high computing power. Compared with NPU, DSP, and MCU computing platforms, the AI computing power of WTM2101 is increased by 10-200 times. At the same time, in terms of low power consumption, the WTM2101 inference operation deep learning network process consumes only 50uA-3mA. In addition, WITMEM provides a wealth of software development tools, which can help more developers develop mature solutions based on WTM2101. These advantages enable WTM2101 to better meet the market demands of smart wearable products and have ease of use. At present, the WTM2101 chip has been implemented in many categories such as smart watches and wireless earphones, and the annual sales volume can reach one million.



Wang Shaodi, founder and CEO of WITMEM, believes that innovation and differentiated competition will become the focus of the smart wearable field in the next few years. The WTM2101 chip can increase AI computing power by 100 times under the power supply commonly used by existing wearable devices, enabling more AI functions and experience upgrades, such as intelligent noise reduction, specific sound enhancement, sound effect processing, AI health, and motion gesture recognition. , command and control, environmental recognition, and more. It is precisely these functions that will likely bring new market growth points for Bluetooth audio and smart wearable devices.

https://en.witmem.com/