» Current | 2018 | 2017 | 2016 | Subscribe

Linley Newsletter

Cambricon Leads China Into AI Chips

March 6, 2018

Author: Mike Demler

Cambricon Technologies is a two-year-old Beijing-based startup that develops intellectual-property (IP) cores and processors for deep-learning acceleration. In 2017, it received $100 million in funding from the Chinese government. Among its first customers is Huawei, which uses a Cambricon core in the Kirin 970 smartphone chip.

The company withheld details of its production deep-learning accelerators (DLAs), but its published prototype implements a RISC-like load/store architecture that’s 64 bits wide and has just 43 scalar, vector, and matrix instructions. These instructions are optimized for neural-network operations. The design combines a small scalar CPU with a variable-width vector/matrix coprocessor. In the Cambricon prototype, the vector execution unit comprises 32 INT16 adders, dividers, multipliers, and transcendental blocks (e.g., exponential, logarithmic, trigonometric, and anti-trigonometric). The matrix unit is an array of 32 identical computation blocks connected together on an H-tree bus. An H-tree is an efficient method for arranging a binary number of elements, since the area is proportional to the number of nodes.

Cambricon is ramping its sales force, but thanks to its $100 million in funding, it’s poised to take on the incumbent DLA-IP suppliers. Its design win at Huawei brings to market a DLA that surpasses the performance of the neural-processing engine in the Qualcomm Snapdragon 845. The relationship with Huawei’s processor-design team will help Cambricon develop production-ready offerings, but the startup is unlikely to license its DLAs beyond its home market. We expect it will focus on its stated mission, which is to build processor chips for intelligent cloud servers, robots, and smart terminals.

Subscribers can view the full article in the Microprocessor Report.

Subscribe to the Microprocessor Report and always get the full story!

Purchase the full article


Linley Spring Processor Conference 2018
April 11 - 12, 2018
Hyatt Regency, Santa Clara, CA
Register Now!
Linley Fall Processor Conference 2018
Covers processors and IP cores used in embedded, communications, automotive, IoT, and server designs.
October 31 - November 1, 2018
Hyatt Regency, Santa Clara, CA
More Events »


Linley Newsletter
Analysis of new developments in microprocessors and other semiconductor products
Subscribe to our Newsletter »