Pingxin Technology Launches Two AI Chips in Beijing
TapTechNews August 9th news, Pingxin Technology held the event of "Storing and Computing in the Chip, Initiating the Future" in Beijing yesterday (August 8th), launching two AI chips, PIMCHIP-S300 and PIMCHIP-N300.
PIMCHIP-S300
The official introduction said that this AI chip is a multi-modal intelligent perception and decision-making chip, with the characteristics of high energy efficiency, small area, low power consumption, and low cost.
This chip is equipped with a SRAM-based in-memory computing acceleration unit, and the computing core energy efficiency ratio is as high as 27 TOPS/W, allowing computing to occur inside the memory and effectively reducing data movement during the computing process, and the energy efficiency of the core computing unit is improved by dozens to hundreds of times.
TapTechNews learned from the report that the PiMCHIP-S300 is equipped with a lightweight MCU processor to achieve real-time control and scheduling, and supports the access of audio, video and various sensors to achieve multi-modal fusion perception.
The PiMCHIP-S300 can achieve "zero movement" of data processing. While greatly improving the computing efficiency, it significantly reduces the energy consumption and improves the computing efficiency in fields such as smart wearable devices, smart security, embodied intelligence, AI large models, and health data analysis.
PIMCHIP-N300
PIMCHIP-N300 is a fully self-developed in-memory computing integrated innovative IP, a dedicated acceleration core for the field of machine learning and artificial intelligence. It is specially optimized for a large number of neural network models involved in AI computing, and can process machine learning algorithms such as artificial neural networks and deep learning models with higher efficiency and lower energy consumption to facilitate the landing of AI in multiple scenarios.
Computational Power
N300 provides a single-core 0.5 TOPS computing power, as well as a Cluster cluster architecture composed of multiple NPU cores.
Accuracy
Supports mixed-precision computing, supports integer 4-bit, 8-bit and floating-point 16-bit computing to better balance power consumption, computing density and computing accuracy.
Compatibility
Supports custom operators to meet the needs of various model deployments, and provides configuration plans and special optimizations for application scenarios such as voice monitoring, eye tracking, active noise reduction, and environmental perception.
Related Reading:
"Alumnus of Tsinghua University, Chen Yiran and Yang Yue formed a team to enter the AI chip market, established Pingxin Technology, and the latest Pre-A round won nearly 10 million US dollars."