Car stuff (public number: chedongxi)
Author | The Way
Edit | Juice
Chedong March 23 news, today's Chief Intellectual Officer Conference and Machine Heart AI Technology Annual Conference opened, this conference around the intelligent car 3.0 era, intelligent driving data, bicycle intelligence, cloud edge end collaboration, China's RoboTaxi commercialization and other themes, but also invited the guests around the era of large computing power chips and autonomous driving commercialization how to mature to discuss, dry goods.
Among them, the organizers of the conference invited Yang Yuxin, CMO of Black Sesame Intelligence, Wang Ping, EXECUTIVE PRESIDENT of Cambrian Xingge, Wang Kai, director and CEO of Core Qing Technology, and Li Bo, vice president and head of the intelligent driving business line of Lotus Technology, to participate in the roundtable debate on "Chip Challenges in the Era of Big Computing Power" to discuss the problems faced by the chip industry in the era of large computing power.
Everyone discussed issues such as chip computing power, chip shortage and chip self-development, and then talked about the introduction of intelligent cockpits and L5 level automatic driving.
▲Guests participating in this roundtable debate
First, the chip computing power is sufficient Reserve computing power is also necessary
In the roundtable debate session, Yang Yuxin, CMO of Black Sesame Intelligence, said that in the short term, the chip computing power is lacking, and the situation of insufficient computing power will not appear, in theory, the chip computing power is enough for the use of car manufacturers. He said that at present, automatic driving is still in the early stage of development, which can be said to be the second half of the first half, and everyone pays more attention to the mass production and user experience of the car, and is not mature enough in user communication and other scenarios. From another point of view, car companies need to help users and help the market recognize the computing power of chips. Some car companies have raised the computing power very high, but the current chip computing power is enough.
Later, he said that hash rate is a necessary but not sufficient condition. While hash rate is a critical point, chips can be multi-core heterogeneous. Therefore, the consideration of chips is multifaceted, and the computing power is the most concerned. In the end, the hash rate may enter the bottleneck. As chip manufacturers, they are also working hard to help users achieve more functions through smaller computing power, reducing power consumption and cost while improving computing power.
▲Black Sesame Intelligent CMO Yang Yuxin
Li Bo, vice president of Lotus Technology and head of the intelligent driving business line, believes that it is necessary to reserve enough computing power to prepare for the upgrading and updating of the later automatic driving function.
Li Bo said that it is possible to use fewer sensors to achieve automatic driving in the future, but in the current era, the first car that can achieve high-level autonomous driving will definitely use multiple sensors. The price is in the range of 400,000-500,000, and there is a basis for fully autonomous driving models, and it is also necessary to reserve enough computing power.
He added that software defines the car, while hardware defines the ceiling of software. When asked whether he would cooperate with Geely's partner Core Engine Technology in the future, he said that he would maintain communication with Core Engine and looked forward to future cooperation opportunities.
▲ Li Bo, vice president of Lotus Technology and head of the intelligent driving business line
Second, the chip self-research difficulty is high The shortage may continue until next year
For Xiaopeng, Tesla and Weilai full-stack self-developed chips, Wang Kai, director and CEO of CoreTek, said that the architecture of the car is constantly upgrading and developing to low power consumption. Software-defined cars, soft and hard fusion is the general direction of the car. OEMs are also playing an increasingly important role, with teslas and Ford developing their own chips.
He believes that OEMs have incentives to carry out chip self-research, but there are also difficulties. With the improvement of automobile intelligence, the number of automotive chips and performance requirements have increased, and the chips supplied by traditional chip manufacturers such as Japan and the United States are not enough to meet the performance requirements of OEMs, and the cost is increasing. Therefore, self-developed chips occupy a great advantage.
However, the threshold for high-end chip research and development is high, and there is a shortage of talents. If the car company does not do well, it will face the problem of financial loss and planning imbalance, and the return is low. Automotive chips and consumer electronics chips are different, with high thresholds, high investment, and long time required for vehicle regulation certification. It is difficult for any one company to succeed, it needs to be cooperated, and the products provided should meet the needs of multiple enterprises.
Wang Ping, CEO of Cambrian Xingge, said that Cambrian Xingge will launch a 7nm process with AI computing power of more than 400 TOPS chips for the L4 market. There is a lack of domestic solutions in this field, and customers expect to launch domestic chips. In addition, Cambrian Has established a team to develop vehicle-grade chips, which will get the chips on the car as soon as possible, and has also received a lot of help from the parent company.
▲Cambrian Xingge EXECUTIVE PRESIDENT Wang Ping
In addition, Wang Kai, director and CEO of CoreTek Technology, also said that the chip shortage problem did not appear suddenly, and it has been encountered before. The biggest reason for this is the epidemic, because in 2020, everyone's travel frequency is reduced, the demand for the automotive industry is reduced, and the upstream production of the automotive industry chain is reduced, and after the epidemic eases, the supply chain has not returned to the previous production capacity, resulting in a shortage of chips. He expects to be able to restore production capacity next year or in the first half of next year.
▲Wang Kai, director and CEO of Core Qing Technology
As for the chip shortage problem, Yang Yuxin, black sesame intelligent CMO, said that the relevant responsible persons in the automotive industry such as material suppliers and chip manufacturers are more concerned, and the industry consensus is that the chip shortage problem will continue this year and may continue until the beginning of next year. The first focus is whether the domestic chip production capacity can be on top, the second is whether the production capacity of foreign manufacturers can be improved, and the most important thing now is the production capacity of mature process chips.
At the same time, he said that the future may coexist with the chip shortage problem for a long time, one is because the chip industry production capacity is limited, and the other is because the chips equipped with cars are also increasing.
He stressed that domestic chip manufacturers need to speed up research and development to avoid another situation in the future. In addition, because foreign car companies generally support local chip manufacturers, it is difficult for domestic chip manufacturers to get foreign orders, so they hope to be able to obtain the support of domestic car companies with the technology of black sesame intelligence.
Third, the intelligent cockpit is gradually introduced To mention L5 level automatic driving is premature
Talking about the landing of the smart cockpit, Yang Yuxin, the CMO of Black Sesame Intelligence, said that the smart cockpit has higher requirements for AI performance, and the chip concentration of the cockpit is also high, which requires the continuous improvement of AI computing power. Wang Kai, director and CEO of Core Engine Technology, said that the first domestic 7nm process process vehicle specification intelligent cockpit chip "Dragon Eagle One" independently developed by Core Engine Technology was successfully released in Wuhan in December 2021, and will be mass-produced and mounted on the car as planned this year to help the intelligent cockpit introduce more models.
Talking about L5 level automatic driving, Wang Ping, executive president of autonomous driving Cambrian Xingge, said that it is necessary to talk about the algorithm used first, and first talk about the solution adopted. In addition, he agreed with Yang Yuxin and Li Bo's consideration of different indicators of the car in terms of computing power. The hardware part of the car is relatively fixed, while the software can be updated by OTA, and although part of the computing power of the chip has not been used, it can be planned in advance, such as the ET7 model of Nio.
Wang Kai, director and CEO of CoreTek, said that he agrees that hardware defines the ceiling of software, and hardware defines the degree of freedom of software, but hardware has costs. The most important thing in the chip design is its power consumption. When the automatic driving really reaches the L5 level, the chip computing power required is quite different from the current L2 and L3 automatic driving. In addition, different levels of autonomous driving algorithms are also different, and the computing power should be enough to match the algorithm. At the same time, all aspects of the chip must also keep up.
As a demand-side enterprise executive, Li Bo, vice president of Lotus Technology, believes that from the perspective of the entire development of automatic driving, it is too early to mention L5 level automatic driving, and there are two important indicators for the division of automatic driving level, one is to take over the mileage, and the other is to cover the mileage. To truly achieve L5 level automatic driving, the test mileage should reach hundreds of thousands of kilometers of untaken mileage. Compared to the long-term goal of L5, it is necessary to achieve the current goal first.
Conclusion: Multi-angle analysis helps the healthy development of intelligent driving in China
In the roundtable debate session, a number of big bulls from different positions in the automotive industry chain, shared a variety of views on whether the chip computing power is enough in this field, at the same time, also conducted a detailed analysis of the difficulties and pain points of chip self-development, encouraged chip manufacturers to increase research and development efforts, and also suggested that local car companies and local chip manufacturers establish cooperative relations to create a more stable supply chain, so as to alleviate the domestic chip shortage. In addition, there is also more understanding of the introduction of smart cockpits and L5 level autonomous driving.
This roundtable debate also built a good communication platform for the heads of many companies in the automotive industry chain, giving everyone the opportunity to discuss how to solve the problems in the field of intelligent driving, as well as the problems that need attention, and also broadened the company's business development direction and helped the healthy development of intelligent driving in China.