In the next ten years, the fastest-growing and most core component in the car will no longer be the engine, but the autonomous driving chip.
Ideal Cars is clearly aware of this.
On September 22, at the Beijing Ideal Auto Delivery Center, Li Auto signed a strategic contract with NVIDIA and Desay SV.
Signing ceremony site
In 2022, Li Auto will become the first automaker of NVIDIA DRIVE AGX Orin, NVIDIA’s next-generation software-defined platform for autonomous driving and robotics.
In short, cars powered by Nvidia’s next-generation self-driving chips will be debuted by Li Auto. Li Auto previously announced that it will start L4 autonomous driving research and development ahead of schedule, and the investment scale will reach 1 billion. Taking the lead in adopting NVIDIA’s latest self-driving chip can be said to be another major move for Li Auto’s rush to L4.
At the same time, Wang Kai, the new CTO of Li Auto, also made his debut at the signing ceremony.
Dual Orin chips can achieve L4 autonomous driving
At the signing ceremony, Li Auto announced that Nvidia’s latest Orin self-driving smart chip will be installed on the second car launched in 2022.
Li Auto’s second vehicle, positioned as a full-size SUV, will continue to use extended-range hybrid technology.
As consumers pay more and more attention to the technological attributes of automobiles, it can become the first car equipped with NVIDIA’s new chips, and the ideal car can be regarded as a big crab.
At the GTC conference in December 2019, NVIDIA released the next-generation software-defined platform for autonomous driving and robotics – NVIDIA DRIVE AGX Orin, and the carrier of this platform is the Orin autonomous driving smart chip.
Orin is currently the most powerful autonomous driving smart chip. Just the number of transistors is enough to see the power of this chip. Orin has 17 billion transistors. For comparison, Nvidia’s previous generation self-driving chip, Xavier, had only 9 billion transistors. Orin has nearly doubled the number of transistors compared to Xavier.
In terms of specific parameters, the single-chip computing power of the Orin chip reaches 200TOPS (200 trillion times per second), which is 7 times the computing power of the previous generation chip.
Three main functions are integrated inside Orin, NVIDIA’s next-generation GPU architecture and Arm Hercules CPU cores, as well as new deep learning and computer vision accelerators.
Using a single Orin chip, it can realize L2+ level advanced assisted driving functions. If dual Orin chips are used in a car, L4-level autonomous driving functions can be realized. In the future, further expansion of computing power can be achieved through dGPU, which can theoretically achieve 2000TOPS computing power, which reserves hardware and computing power redundancy for L5-level autonomous driving.
In the Orin chip, the Arm Hercules CPU is also integrated, which is ARM’s next-generation smart device CPU, which means that Orin is not only an autonomous driving chip, but can also provide the computing power required for smart car machines.
A single Orin chip can achieve the computing power of two Xavier+820A chips, which is obviously very attractive to car companies.
A single chip can meet the computing power required for smart cockpit and autonomous driving, and car companies do not need to separately arrange the chips required for autonomous driving and smart cockpit. In terms of cost and complexity of vehicle Electronic design, Orin’s single-chip solution has more advantages than the current two-chip solution.
Just a week before the signing ceremony, Nvidia announced a $40 billion acquisition of ARM. When talking about this issue, Zhang Jianzhong, NVIDIA’s global vice president and general manager of China, said, “NVIDIA’s acquisition of ARM is a very important thing.”
Zhang Jianzhong, Global Vice President of NVIDIA
In almost all IoT and mobile devices, ARM’s computing platform is widely used, and NVIDIA’s investment in AI is huge, so NVIDIA hopes that the results achieved in AI can be put into a wide range of computing architecture platforms. The cooperation of ARM enables AI to be applied to every mobile processor, enabling all ARM users and potential ARM users to benefit from the benefits of AI.
After NVIDIA completes the acquisition of ARM, Orin and NVIDIA’s software-defined platform for autonomous driving and robotics – NVIDIA DRIVE AGX Orin will be better developed.
On NVIDIA’s strong CUDA GPU, NVIDIA has been deeply involved for nearly 20 years and is currently the industry leader. On the RISC (reduced instruction set) processors required by IoT devices, ARM has already become the industry hegemon and is far ahead. The merger of the two companies provides a fundamental guarantee for the follow-up research and development of the intelligent driving platform.
It is not accidental to change the door
Nvidia’s rival, longtime rival in the personal computer space, is Intel.
Intel’s development in the field of autonomous driving was achieved through the acquisition of Mobileye, which cost Intel $17 billion.
The latter’s Mobileye Q4 chip is the intelligent assisted driving chip used by the Ideal ONE.
Mobileye’s advantage is that the system is mature, but maturity itself may also be one of the reasons that hinder the development of car companies.
In terms of loading speed, Mobileye has an advantage. The Mobileye Q4 chip can adapt to different sensors and has more advantages in data accumulation, and Mobileye also has more advantages in the development of autonomous driving software.
The combination of these advantages is that car companies do not want to do or do not want to do automatic driving software and hardware platforms, Mobileye has helped car companies do a good job, car companies can load vehicles at the fastest speed, and realize L2-level intelligent assisted driving functions.
But from another point of view, Mobileye will become a constraint for car companies with ambitions for autonomous driving.
Mobileye’s mature system, for the sake of stability, leaves little room for car companies to play freely. And the more mature the system is, the more requirements are placed on car companies, and car companies will inevitably become vassals of the autonomous driving R&D platform and lose control over autonomous driving R&D.
In terms of business cooperation, because Mobileye has too many partners, Mobileye is inevitably stretched when allocating R&D resources to car companies.
Under the general trend of software and hardware integration, how much resources an autonomous driving chip hardware company can allocate to a car company often means how high the autonomous driving of the car company can achieve.
In this regard, the ideal car is not as good as the old European car companies. For example, the first car equipped with Mobileye’s next-generation autonomous driving chip Mobileye Q5 has been confirmed to be launched by BMW in 2021.
Therefore, for a company with great potential for development in the field of smart cars and autonomous vehicles, the advantages that Mobileye can provide are not exactly equal to the needs of ideal cars.
The next goal is to achieve NOA
Regarding the change of family, Wang Kai, the new CTO of Ideal Auto, said that NVIDIA and Desay SV can provide a more open cooperation pattern and achieve a win-win situation.
Ideal Auto CTO Wang Kai
Wang Kai said that the current autonomous driving perception part of Lili ONE is the technology provided by Mobileye, while the body control and adjustment are the responsibility of the internal team. In the future, the technology of the visual perception part will be self-developed by Ideal.
To achieve self-development of perception capabilities, replacing the autonomous driving R&D platform is a key step.
Therefore, it is not difficult for us to understand why the ideal car changed its door from the Mobileye platform to the NVIDIA DRIVE AGX Orin autonomous driving platform.
But far cannot quench the thirst of the near, ideally the next model, there are still 2 years before the announcement, and the delivery time to consumers will be longer.
Whether the existing Ideal ONE owners can enjoy the performance improvement of NVIDIA Orin chip in advance, Wang Kai did not give a positive solution.
What is certain is that the ideal autonomous driving function of the moment is NOA. For ideal ONE owners, NOA’s desserts seem to be more fragrant than Yuanshui, which can’t quench their thirst.
As for the effect of switching from the Mobileye platform to the NVIDIA platform, it will be left to the future to judge.