Saratoga, CA, July 29, 2021 – Aizip, Inc., a Silicon Valley startup company providing leading artificial intelligence (AI) models for the Internet of Things (IoT), the so-called tiny machine learning (TinyML), announced that it has reached an exclusive license agreement with Rice University for their Processing-In-Memory (PIM) technology.
Over the last decade, AI has proven to be of tremendous value in improving quality of life on many fronts. One of the most critical areas in AI-technology development is power efficiency. In the AI-inference process in an existing chip, most of the power is wasted in the data transfer between memory and computing units. Analog computing inside the memory – that is, PIM – is widely considered to be the ultimate solution to solve this challenge, especially in the IoT market where low power consumption and low chip costs are essential.
Professor Kaiyuan Yang, a recognized leader in PIM research at Rice University, has done pioneering research on the design of PIM chips for low-power and high-throughput AI applications. In a paper recently published in IEEE Journal of Solid-State Circuits, Yang’s group, in collaboration with Northeastern University, presented the results of the first PIM semiconductor chip fabricated with industry standard 6T-SRAM technology. At the core of their innovation is a cell cluster structure based on capacitive analog computing. The device has been fabricated and tested to show one of the smallest reported 0.35-LSB root-mean-square computing errors, and superior area and energy efficiencies over other schemes being pursued worldwide.
“The successful demonstration of 6T-SRAM based PIM hardware is a major breakthrough in the path to ultra-low power and low-cost AI accelerators,” noted Gene Frantz, a professor at Rice University, a member of National Academy of Engineering, and fellow of IEEE. Frantz, who has been called the “father of digital signal processing,” served as a Principal Fellow at Texas Instruments. “The work by Professor Yang and his team demonstrates that PIM can be realized in CMOS semiconductor technology with excellent performance and efficiency, showing considerable potential for its transition into commercial applications.”
“We’re very excited to work with Rice University to bring this innovation to the market,” said Yuan Lu, co-founder and President of Aizip. Aizip believes PIM is a cornerstone in the future of AIoT, or TinyML, and “we have developed extensive IP in AI algorithms and architectures,” according to Lu. He further stated that “the combination of Rice’s IP in hardware and Aizip’s IP in application and software will provide a complete solution in the fast-growing TinyML market.” “I’m glad to see the enthusiasm from industry on our research results and to see a path towards real-world applications,” remarked Professor Yang.
Many semiconductor companies have invested in PIM or have shown strong interest in PIM in recent years. Power consumption is expected to reduce by 100x or more in IoT application, which should stimulate a new breed of applications, especially in battery-powered mobile scenarios. Aizip plans to work with IC partners to design and develop the next generation low-power, analog PIM for AIoT applications.
* * *
About Aizip, Inc.
Aizip, Inc., develops AI models for edge applications. Based in Silicon Valley, Aizip provides models with superior performance, quick turnaround time, and excellent ROI. These models can be used in a wide range of applications for an intelligent, automated, and connected world. For more information, visit www.aizip.ai.