Cambricon unveils three autonomous driving chips in development


At the 2022 World Artificial Intelligence Conference (WAIC) on September 1, Chen Tianshi, president of Chinese AI chip company Cambricon, revealed that the company is developing three self-driving chips covering L2-L4 functions, which could meet different computing power requirements from 10 to 1000 TOPS.

The three chips under development are “Si Yuan 590”, a next-generation cloud smart drive chip, “SD5223”, an integrated chip for L2+ automated driving and parking, and “SD5226”, a SoC (system on chip) for Advanced L4 Automated Driving Multi-Domain Convergence Platforms.

Si Yuan 590 adopts the all-new architecture of MLUarch05, and its measured drive performance has been significantly improved compared to the company’s flagship products on sale. It offers larger memory capacity and higher memory bandwidth, and its input and output and inter-chip interconnect interface has also been significantly improved over previous generations.

SD5223, an integrated automated driver chip for the L2+ market, provides 16 TOPS computing power and higher DDR bandwidth. It supports automatic quality visual processing and natural low-power heat dissipation, and is suitable for 8M IFC/5V5R/10V5R and other multi-product forms.

SEE ALSO: Cambricon plans to raise $395 million for R&D via private placement

In addition, based on the changing market demand, Cambricon has updated the specifications of its SD5226 SoC chip, a high-level self-driving multi-domain convergence platform for the L4 market, and its computing power will reach more than 400 TOPS. It adopts 7nm process technology and advanced AI architecture to adapt to changing algorithms.

(Source: Cambricon)

At the conference, Chen Tianshi said in his speech, “It is predictable that over the next five years, we will see four major trends in automated driving. First, L2+ automated driving systems will be popularized quickly and will be around for a long time. Over the next five years, the overall penetration rate of automated driving systems L2+ and above could exceed 50%. L4 automated driving in restricted scenarios will begin to be implemented, and L2+ to L4 automated driving will exist in parallel. The second trend is that automated driving algorithms are more complex, the amount of data processed is increasing exponentially, and the demand for computing power is constantly increasing. The third trend is that vehicles, roads and cloud computing cooperate to achieve a closed loop of big data, and the driving experience continues to improve. The fourth trend is that to meet the individualized needs of consumers and enhance the differentiated competitiveness of manufacturers, the demand for self-learning from the automotive side is constantly increasing, and cars will be very different from each other in terms of systems. ”

At the conference, Cambricon took “Vehicle, Road and Cloud Computing Collaboration” as the theme, showcasing intelligent processors and chip products covering full scenarios, numerous industry solutions based on the development of Cambricon chips and diversified AI hardware products. It has demonstrated basic technical strength and implementation capabilities in the field of smart chip design.

Previous Strategy World War II (SSM) has been set until 2023
Next "Encore Webinar Series" Parish Development Forum Focuses on Ministry