Why NVIDIA Computing is The ‘TOP’ Choice For Chinese Automakers NIO, IM, XPeng & Li Auto

Many Chinese automakers are super-powering their technology platforms using NVIDIA chips, AI, technology and infrastructure. Ahead of NVIDIA’s GTC 21 tech event, NVIDIA’s Tim Wong and industry analysts explain why NVIDIA DRIVE is driving the future of many automakers in China.

Li Auto, Xpeng, IM and NIO will all be deploying NVIDIA technology. Chinese EV automaker Li Auto will be deploying NVIDIA DRIVE AGX Orin.

Xpeng has been developing on NVIDIA DRIVE since 2018. The XPilot 3.0 level 3 autonomous driving system uses NVIDIA DRIVE AGX Xavier. XPilot was developed in Xpeng’s data centre, with NVIDIA’s AI infrastructure for training and testing self-driving deep neural networks.

SAIC and Alibaba’s new premium EV brand, IM, will use NVIDIA DRIVE Orin compute platform for its premiere electric sedan and SUV. These models will have multiple NVIDIA Orin SoCs (system-on-a-chip) at the core of a centralized computer system, achieving 500 to 1,000+ trillion operations per second (TOPS) of performance for automated functions, autonomous capabilities and in-cabin personalisation.

Due in 2022 the fully autonomous NIO ET7, will be packed with NVIDIA to deploy advanced automated driving technology. NIO’s in-house developed autonomous driving algorithms will be running on four NVIDIA Orin processors. The system is backed with a new NVIDIA-powered supercomputer, called Adam that achieves over 1,000 trillion operations per second (TOPS).

Even with such great demand, NVIDIA, unlike some of its competitors, has plenty of semiconductor stock to go around.

“We have not been affected by the automotive chip shortage. We have more than enough stock ready for all our customers to meet the demand,” says Tim Wong, Technical Marketing for Autonomous Vehicles, NVIDIA.

A voice saying in 100 feet we will change lanes is reassuring to the customers.

Wong explains why NVIDIA is popular with Chinese electric automakers. One reason is the headroom needed to upgrade software.

Wong recently bought a Tesla Model Y. After a few days, he received software updates and an offer to spend an extra $2,000 for faster acceleration.

The trend of software-defined cars is happening in China too and t’s not just for software updates. The automakers need headroom to be able to upgrade their vehicles in the future, says Wong.

He gives the example of the Orin SoCs. They can be used at half power or quarter in the NIO ET7.

“There is a lot of room to add code or add features in the future,” says Wong. “The chips can easily work with the eight-megapixel cameras, multiple sensors and handle the data needed for image recognition.”

Although NVIDIA no longer makes chips for Tesla, it can top out Tesla’s current chips.

“Tesla hardware runs on a 160 TOPS chip while the Orin is 254 TOPS (trillion operations per second), including CUDA TensorCore Ampere architecture GPU, 21.3 Billion Transistors, 12 Cortex-A78 ARM64 CPUs and 200 GB/s Memory.”

NVIDIA provides a platform for different aspects of automotive that is much easier to upgrade, says Wong.

There are three parts of the NVIDIA platform – there is the in-car computer, the enterprise-level computer central AI and then the simulation aspect. Different automakers can customize their systems in different ways for different types of vehicles.

NVIDIA systems can also talk to the driver. “In autonomous mode, we have found that a voice saying in 100 feet we will change lanes is reassuring to the customers,” adds Wong.

New Ways to Raise Revenue

Analysts suggest NVIDIA technology is gaining popularity in China due to Tesla envy, software and profits.

Many Chinese OEMs are trying to follow a similar strategy to Tesla. The company used NVIDIA in its first-generation cars before bringing the silicon design in-house, says Phil Amsrud, Senior Principal Analyst, IHS Markit.

“For Chinese companies looking at Tesla for inspiration, selecting a high-compute-based architecture makes sense. It lends itself to a more centralized electrical architecture,” says Amsrud.

“Any software-defined car requires the ability to update the software after the car is in the field. For example, NIO’s scalable platform based on NVIDIA’s Orin covers 10 – 1000 TOPS of performance. That will support a lot of software reuse. It allows the software to be updated as lessons learned that can be applied across the entire platform,” adds Amsrud.

There may be cultural reasons why NVIDIA is so popular in China, says Chris Schreiner, Director of Syndicated Research, Strategy Analytics.

Chinese people are more accepting of new technology. They have almost completely converted to contactless payments. Everyone of all ages uses contactless payments on their phones. In the West, people are more hesitant to adopt new technology. They don’t like to risk their privacy, notes Schreiner.

“Advanced technology also makes it easier for automakers to sell software-as-service or software upgrades. Over-the-air-updates and unique software is a way for automakers in China to keep the vehicle modern,” says Schreiner.

He points out that automakers are looking for ways to make extra money by selling new infotainment or other features. A longer battery range, faster acceleration or entertainment are ways for them to raise revenue.