Urban auxiliary driving war on the eve of the battle: Nvidia chip is restricted, domestic car companies are looking for "Plan B"

As the new forces of cars are building from high -speed scenes from high -speed scenes to urban roads to urban roads, more and more car companies are actively entering this track.

Recently, Changan Automobile announced that it will achieve mass production of urban auxiliary driving in 2024. Prior to this, Wei brand released the first domestic mass-produced city auxiliary driving system and took the lead in loading on the We brand Mocha DHT-PHEV lidar version. Even Xiaomi cars, which have not yet been "launched" by mass production cars, also displayed the city -aided driving function that did not intervene in the test car.

Nvidia chip is restricted, domestic car companies are looking for "Plan B"

When major car companies are competing for urban road development auxiliary driving functions, Graphic chip giants Nvidia suddenly announced a few days ago that the latest two -generation flagship GPU computing chips A100 and H100 need to apply for new permits to export the latest two -generation flagship GPU computing chips. At present, domestic Weilai, Xiaopeng and other companies have created autonomous driving training centers based on Nvidia A100 to achieve higher -level auxiliary driving functions.

Some analysts believe that with the popularity of urban auxiliary driving, the demand for large computing chips and large computing platforms will also be higher. This round of Nvidia chips will directly affect the layout of enterprises and host factories in the autonomous driving field.

"PLAN B" outside Nvidia

Public data shows that more than 25 car companies and autonomous driving companies in the world choose to use Nvidia smart driving chips. Nevertheless, the car company that has tasted the suffering of "lack of cores" has not put the eggs in a basket this time, but is looking for more "Plan B" (that is, when the first plan is not feasible, use The second solution).

For example, Weilai and Xiaopeng chose to cooperate with Nvidia, Mobileye and Qualcomm at the same time. In addition, car companies such as FAW, Changan Automobile, Great Wall Motors, and SAIC Group choose to cooperate with local autonomous chip manufacturers represented by Black Sesame Smart and Horizon.

From the current point of view, in the large computing chip market, there are many options for Nvidia, and the speed of mass production is also accelerating. The Extreme Fox Alpha S · Hi version that has been delivered and the Avita 11 delivered during the year is also equipped with Huawei MDC 810 smart driving computing platform. In addition, the Mocha DHT-PHEV lidar version equipped with a 7nm Qualcomm Snapdragon 9000 chip is expected to mass production at the end of September this year and be delivered within the year.

"In terms of comprehensive computing power, the computing power of Qualcomm Chip 360TOPS has already appeared redundant, which is enough to cope with the urban auxiliary driving function at this stage." Wei brand CMO Qiao Xinyu said in an interview with the reporter of "Daily Economic News" that Qualcomm Snapdragon Ride Ride The average power consumption of the platform is as low as 5TOPS/W, and the power consumption comparison advantage with similar performance products is obvious. In addition, Qualcomm's solution advantage is also more obvious in terms of expandability, power consumption, efficiency, and local support teams.

The reason why more and more car companies find Plan B are another consideration to keep their words. Earlier, due to the poor openingability of Mobileye, the OEMs lacking the right to speak with it, the accumulated data difficult to support the rapid optimization of the algorithm, which caused iteration to iterate in the intelligent driving function. At present, Tesla, Weilai, Xiaopeng and other car companies have chosen full stacks to open up data-algorithm closed-loop channels.

CICC Research Report believes that with the continuous improvement of the self -developed software algorithm of the OEM, in the future, car companies may find suitable chips for their own software algorithms through self -developed+cooperation, and improve the adaptation efficiency of software algorithms and chips. To maintain product advantages.

Pay attention to the light map routes are favored

In addition to looking for "Plan B", autonomous driving companies and car companies to realize urban auxiliary driving also need to solve problems such as high -precision maps.

The so -called urban road auxiliary driving is based on the route set by the user to help users realize from point A to urban B point B, but compared to semi -closed and road structured high -speed scenes, urban road conditions have high complexity, pedestrians, non -machinery, and non -machinery There are many road factors such as vehicles and motor vehicles, so the assistance driving function on urban road development is more difficult to mass production.

At present, many car companies use high -precision maps to master the location and instructions of the roadplaces, instructions, etc., which is equivalent to superimposed people's experience and intelligence to the map, marked the left -turn lane, right -to -right -handed lane, and universal areas. However, due to factors such as national security, regulations, and policy approval, it is very difficult for high -precision map solutions to achieve landing.

In this context, more and more companies such as Xiaopeng, Huawei, and Wei brand are reducing their dependence on high -precision maps. "Under the technology that is aware of it, you can achieve urban auxiliary driving functions without a high -precision map." Pan Xing, the technical director of Zhixing, told reporters.

To improve perception, it means that the vehicle will be equipped with more laser radar. For example, Wepai's city -assisted driving system with mass production has 2 125 -line laser radar, 5 millimeter wave radar, 12 ultrasonic wave radar and 12 high -definition cameras; Lidar, 6 millimeter wave radar, 13 cameras and 12 ultrasonic wave radar.

"The distance perception and 3D configuration brought by the laser radar will complement the advantages of other perceptions. Under the blessing of the algorithm, we can avoid strengths and avoid shortcomings, bringing safer and more reliable urban auxiliary driving." Pan Xing said.

CICC Research Report believes that with the gradual popularization of high -speed auxiliary driving, the urban auxiliary driving is accelerated, the market has also switched from functional popularization to interaction and user experience.The key to constructing differentiated competition advantages.Daily Economic News

- END -