Pimchip Technology is a pioneering company of processing-in-memory (PIM) technology bringing a vision that the universe can be better perceived by AI computing. The PIM technology, combining processors and data storage, revolutionizes the computing architecture paradigms by breaking through the inherent limitations of traditional chip architecture (Von Neumann), known as “memory wall” problem, and hence enables high-performance AI computing engines at low cost. The PIM chips allow computationally demanding machine learning algorithms to run on local devices, while enjoy lower power consumption and smaller size than Von-Neumann based chips. Not limited to AI, PIM in fact can be tailored to accelerate any given tensor/vector calculation.
AI acceleration units and chips developed by Pimchip Technology based on SRAM processing-in-memory technology. PIMCHIP-N300 is a new processing-in-memory neural network processing unit (NPU) that provides a dedicated acceleration core for machine learning and artificial intelligence fields. It processes machine learning algorithms such as artificial neural networks and deep learning models with higher efficiency and lower energy consumption, facilitating the implementation of AI in multiple scenarios; PIMCHIP-S300 specifically used in intelligent perception and decision-making, The demonstrated high computation efficiency beats the current GPU solutions by at least an order of magnitude. Beyond, Pimchip offers customized solutions to accelerate the construction of universe perception, opening a new eco-system in the era of AI.