Cambricon today unveiled its Cambricon 1M chip for edge computing, and the MLU100, the first in a new chip series for cloud computing.
Cambricon 1M is the company’s third generation AI chip for edge devices. With its TSMC 7nm technology, the AI chip provides efficiency of 5 TOPS/Watt for 8-bit computing. The 1M chip is available in 2, 4, and 8 TOPS versions to support a range of AI applications.
Like the company’s 1H and 1A chips, Cambricon 1M supports deep learning models such as CNN, RNN, SOM; and supports SVM, k-NN, k-Means, decision tree and other algorithms. This is also the world’s first AI processor supporting local machine learning training.
Cambricon MLU100 is the first generation of Cambricon’s new product series supporting cloud computing. MLU100 adopts Cambricon’s latest MLUv01 architecture and TSMC 16nm technology. This processor has the capability of providing 128 TFLOPS/166.4 TFLOPS in balance mode and high-performance mode respectively.
Cambricon and its partners demonstrated several use cases: Lenovo’s new ThinkSystem SR650 server is based on the MLU100, as is Sugon’s new PHANERON series of products. iFlytek has also announced a collaboration with Cambricon.
Author: Victor Lu| Editor: Tony Peng, Michael Sarazen