Bai wrote that intelligent driving algorithms are set to evolve towards end-to-end and large model directions in the near future. This necessitates that core components like CPU/NPU/MSP not only cater to current needs, such as enabling Bird’s Eye View (BEV), network occupancy, and end-to-end perception control, but also future-proof themselves to seamlessly integrate with large model algorithms.
At present, NIO vehicles come equipped with four Orin chips, which, according to Bai Jian, are “more than enough to meet” the existing demands. However, he hinted that the next-generation smart driving chips are expected to outperform their predecessors. His post concluded with an exciting invitation to the audience to “stay tuned for NIO Day”, which will be held this coming Saturday.
Earlier this year, during an event in September, Li Bin, the founder of NIO, announced the mass production of NIO‘s first proprietary chip in October. Touted as the industry’s first self-developed LiDAR master control chip, the Chinese-named “Yang Jian” chip boasts an 8-core CPU, 8 sampling channels, 9Bit sampling depth, and a sampling rate of 1GHz. It is claimed to reduce power consumption by 50%, latency by 30%, and has the capacity to process 8 million points cloud per second.”