Nio's current use of four Nvidia Drive Orin chips is fine for autonomous driving algorithms, but the next generation smart driving chip should do better, an executive said.
(Image credit: CnEVPost)
Nio (NYSE: NIO) continues to warm up for the upcoming Nio Day 2023, with one executive hinting that the company may unveil an in-house developed autonomous driving chip.
Nio's current use of four NVIDIA Drive Orin chips is fine for smart driving algorithms, but the next-generation smart driving chip should do better, the company's vice president for hardware business, Bai Jian, said today on Weibo.
"Is there such a chip? I think there is, stay tuned for Nio Day!" Bai wrote.
A year ago, there were media reports that Nio was working on autonomous driving and LiDAR chips, and by then it had a chip team of 300 people.
Nio's chip team was formed in the second half of 2020 under Nio's smart driving hardware division, which is headed by Bai, who was previously general manager of Xiaomi's chip business and OPPO's hardware director, according to an October 2022 LatePost report.
Nio's chip team was working on two chips at the time, a high-level self-driving chip and a LiDAR chip, the report said.
The company's self-driving chips could start being used in production cars as early as around 2024, according to the LatePost report.
At the Nio IN 2023 Innovation Day event on September 21, Nio unveiled its first in-house chip, a LiDAR main control chip. William Li, the company's founder, chairman and CEO, said at the time that the chip would be in mass production in October.
Nio's models based on the latest NT 2.0 platform come standard with the Nio Adam supercomputer, powered by four Nvidia Drive Orin systems-on-chips (SoCs) that deliver a combined 1,016 Tops of computing power.
The company's first generation of assisted driving uses Mobileye's EyeQ4H chip, with all perception algorithms provided by Mobileye and Nio's algorithm team responsible for planning and controlling algorithm development, according to Bai's Weibo post today.
To improve development efficiency and iteration speed, Nio started to develop its own sensors, domain control hardware, and autonomous driving algorithms since the NT 2.0 platform, according to Bai.
The speed at which Nio's city pilot-assisted driving feature now adds coverage of road miles and cities proves the importance of the four Orin and Aquila super-sensing systems that come standard, said Bai.
With plenty of computing power, Nio Adam will be the top-of-the-line domain control hardware for the next five years, Bai said.
Nio's algorithm team can upgrade the latest algorithms on the Adam, fix long-tail bugs, and continue to improve the smart driving experience to stay ahead of the curve, he said.
In the next few years, smart driving algorithms will evolve towards end-to-end and big models, so core components such as CPUs, NPUs, and MSPs will need to meet the needs of BEVs, occupancy networks, and end-to-end perception planning and control, but also need to think about the future and match the end-to-end big model algorithms, Bai said.
"The current 4 Orins, of course, can meet the demand, but the next generation smart driving chip, should do better," he wrote.
Nio teases core features of ET9: 900 V high-voltage architecture, wire-controlled chassis
