laitimes

Interview with NavInfo: Balancing NPU and GPU computing power to ensure market competitiveness

author:EV Vision

On April 25, NavInfo released the NI in Car intelligent integrated solution at the Beijing International Auto Show with the theme of "Driving Intelligence". After the meeting, NavInfo's Senior Vice President Bi Lei was interviewed by EV Vision and elaborated on future market development and other issues.

Interview with NavInfo: Balancing NPU and GPU computing power to ensure market competitiveness

It can be said that NavInfo has quickly completed the transformation from map service to overall solution, and in the future, whether the company has a clear market goal and what market size it will achieve within a specific time period has become the focus of attention in the industry.

Bi Lei clearly pointed out that the company's strategic planning is closely focused on the main products in the market. NavInfo's technical strength is already in the leading position in the industry, but in terms of market landing strategy, the company will focus on those consumer groups who have both demand and certain cost affordability, especially those in the price range of 10-200,000 yuan. Bi Lei emphasized that the company's goal is to meet the needs of this group and gradually advance the technological development, from the current NOP (Navigation Assisted Driving) to more advanced autonomous driving functions in the future, leading the market shift from high-end models to mid-range models.

In terms of technical route, NavInfo has chosen a fusion perception solution based on visual perception, which gradually reduces the dependence on maps under the BEV (Bird's-eye View) + transformer architecture. Bi Lei revealed that the company has taken the initiative to adjust the technical direction, from high-precision HD maps to HD Lite, and has reserved relevant technologies for future needs. He emphasized that maps are not only functional parts, but also safety parts, which can reduce risks under certain conditions. NavInfo is reducing costs and improving security through technology optimization to meet users' concerns about overall functional experience and security.

In the field of perception algorithms, NavInfo has demonstrated its R&D strength. Bi Lei introduced that the company has both its own research team and cooperation with partners to ensure the competitiveness and cost control of products. At present, the size of the self-developed perception algorithm team is about 50 people.

Interview with NavInfo: Balancing NPU and GPU computing power to ensure market competitiveness

The use of data is another focus of NavInfo. Bi Lei said that the company pays more attention to transforming data into leading functions and product technologies, rather than simple data monetization. After the data is closed, a large amount of data can improve the effect of algorithms and vehicle-end applications. At the same time, the company is also exploring business model optimization and data trading solutions in the industry, although there is no mature solution yet.

In the field of chips, NavInfo's layout and progress are equally eye-catching. Bi Lei revealed that the company is currently focusing on two directions: on-board cockpit SOC chips and on-board control MCU chips. The company's 8025 chip has been designated in a large number of global international car manufacturers, and although the product has not yet been launched, the relevant contracts have been signed. In addition, the company is also planning the 8035 high-performance chip, which will be decided to be put into research and development according to the market's acceptance and scale of high-performance chips.

Interview with NavInfo: Balancing NPU and GPU computing power to ensure market competitiveness

Bi Lei also mentioned that there is a trend in the market from the integration of travel and parking to the integration of cabin and parking. NavInfo can use the vehicle's existing sensors to realize the driving and parking function, and after the optimization of computing power, it can combine the computing power of the cockpit chip and the intelligent driving chip to further reduce the pre-control cost.

In terms of cost control, Bi Lei predicts that through optimization, the cost can be reduced by at least 20%, and may even reach a reduction of 30% to 40%. He also pointed out that the configuration preferences of OEMs for computing chips are changing, and cost performance has become a key consideration.

In terms of the strategy of going overseas, Bi Lei said that NavInfo's 8015 cockpit chips are selling well, and the company will follow the pace of OEMs to provide lower-cost solutions for the needs of the overseas market.

Finally, Bi Lei looked forward to the future development direction of the company.

He mentioned that although the current living space of single domain control may be challenged, NavInfo will continue to focus on providing complete solutions while maintaining flexibility to meet the individual needs of different OEMs. In the process of intelligent driving integration, the company will balance the computing power requirements of NPU and GPU according to market demand to ensure the market competitiveness of products.