Written by / Qian Yaguang
Editor/ Liu Baohua
Design / Shi Yuchao
With the addition of large factories in the industrial chain, the future technology trend of automobile electrification and intelligence has also attracted more and more attention. In the past 2021, automotive chips have become a hot spot in the automotive industry, both in the supply chain and in terms of development speed.
According to data from the Korea Automotive Research Institute, the cumulative orders of the world's major automotive semiconductor manufacturers have exceeded 30% of the semiconductor production capacity in 2022, and these companies have to start accepting orders in 2023. That is to say, in 2022, the cumulative orders for semiconductor chips from major global automakers have exceeded the production capacity of semiconductor chips, and chip shortages will continue.
Major global automakers are also starting to collaborate on semiconductor technology, internationalization, and supply chain management. Some car companies have replaced custom chips with generic chips to ensure supply flexibility, and some car companies have carried out vertical integration of technology through strategic cooperation with chip companies to cope with the growing demand for semiconductors.
The automotive semiconductor industry plans to focus on investing in the development of next-generation semiconductors and is expanding mass production to address the shortage of semiconductor supplies. Except for a few high-performance semiconductor chips, there is still a long-term shortage of conventional semiconductor chips.
Driven by automatic driving and the concept of "meta-universe", these high-performance semiconductor chips, in the current and future intelligent automotive industry, are becoming the main force of product marketing for the C-end of car companies, and the resulting sharp reduction in the types of automotive chips will also become a panacea to reduce chip shortages. Jen-Hsun Huang, CEO of Nvidia, predicted: "Chips will play a key role in the autonomous driving revolution in the automotive industry. ”
At ces 2022, which opened on January 4, which best reflects the development trend of consumer goods technology, the overall sound volume of the automotive industry has become smaller, and the role of related chip manufacturers has become more prominent, and they have pushed AI chips with better performance and continued to expand the circle of friends of the automotive industry, which has become a highlight of this CES.
A number of "cross-border" major chip giants have not only launched new chip products related to automobiles, but also announced the intention of cooperation with many automobile companies, including traditional car companies such as Volkswagen, Ford, and Toyota, as well as new car-making forces such as Weilai, Xiaopeng, and Ideal.
Judging from the new products released by various chip companies, it is not difficult to find the following characteristics:
First, the performance of the chip has been greatly improved compared with the previous generation, and both the computing power and the function have made great progress. For example, the performance of the EyeQ Ultra launched by Intel's Mobileye is equivalent to the sum of the performance of 10 EyeQ 5s, and the AI domain controller chip CV3 series released by Ambar has 42 times more AI processing computing power than the previous generation of vehicle-grade SoC CV2 series.
Second, each chip giant has many fans of cooperative automobile-related enterprises to jointly promote the improvement of products. For example, Polaris, Xiaopeng, Nio, Ideal, Desay SV, etc. cooperated with Nvidia; BBA, Toyota, Volkswagen, GM Stellantis, etc. cooperated with Qualcomm; Mobileye had Ford, Volkswagen and Extreme Kr.
Third, the product layout of each chip giant is more comprehensive, but in the use of car companies can be flexibly customized, each taking what they need. For example, Nvidia launched hyperion 8 software and hard automatic driving solutions; Qualcomm brought the overall solution of the family barrel smart car - Snapdragon digital chassis; Mobileye not only released a new chip, but also launched a map platform.
Let's take a look at the performance of chip companies at CES 2022:
Nvidia
Nvidia entered the autonomous driving market earlier, temporarily ahead of its major competitors in progress. At this ces, NVIDIA once again showcased the latest NVIDIA DRIVE platform, including the in-vehicle computer (DRIVE AGX) and the complete reference architecture (DRIVE Hyperion), as well as data center hosting simulation (DRIVEConstellation), deep neural network (DNN) and training platform (DGX), as well as a rich software development kit (SDK).
The eighth-generation DRIVE Hyperion 8 architecture, which was officially announced at gtcup in November 2021, can be summarized as NVIDIA's hardware- and software-integrated autonomous driving solution for automakers.
Drive Hyperion 8 features 12 external surround cameras, 3 internal cameras, 9 millimeter-wave radars, 12 ultrasonic radars, and 1 lidar. It provides two NVIDIA DRIVE Orin chips with hash rates, each with a hash rate of up to 254TOPS.
To work with powerful hardware, Hyperion 8 has built-in drive Sim replicator, an autonomous driving training, testing, and validation platform for machine learning.
This is a synthetic data generator for autonomous driving development, through the real physical world data collected by the vehicle, the virtual camera, lidar and other sensors in the simulation platform can be synthesized in multiple dimensions, such as various speeds, distances, occlusions and bad weather is difficult to generate true value data, generating a digital twin environment similar to the real world, in order to continuously exercise the ability to automatically drive.
In conjunction with this platform, NVIDIA also released NVIDIA DRIVE Concierge and DRIVEChauffeur, AI platforms designed to eliminate the stress and cumbersomeness of daily driving.
NVIDIA DRIVE Concierge for AI interaction can combine conversational speech assistance with driver monitoring and automatic parking technology, and can combine speech AI, computer vision, natural semantic understanding and search engines to accurately realize voice requirements and provide personalized customization for every driver and occupant on the vehicle.
DRIVE Chauffeur is an AI-assisted driving platform based on the NVIDIA DRIVE AV SDK that enables robust and efficient autonomous driving operations with a high-performance computing architecture and drive Hyperion 8 sensor set.
Combined with DRIVE Chauffeur, drive concierge can provide low-latency, high-quality 360-degree 4D visualization services, such as providing automatic search for parking spaces and valet parking capabilities. Using the "summon" function, drive Chauffeur can drive the vehicle to its side.
NVIDIA revealed that the company has reached agreements with a number of electric vehicle industry chain manufacturers.
Ali Kani, vice president and general manager of Nvidia's automotive division, mentioned in his speech that vehicle brands such as Mercedes-Benz, Polestar, Xiaopeng, NIO, Zhiji, Ideal, R Auto, as well as suppliers such as Desay SV, Flextronics, Quanta, Valeo and ZF, will also use its autonomous driving platform.
Mobileye
From the only option for car companies to mass-produce self-driving chips, to customers changing doors, Intel's Israeli chip company Mobileye has a sad path.
At this CES, Mobileye seems to be about to reorganize the mountains and rivers, not only releasing the most advanced and powerful EyeQ Ultra system integration chip built for automatic driving so far, but also revealing the news of a good relationship with a number of car companies.
EyeQ Ultra is based on Mobileye's proven EyeQ architecture, in line with the product concept of "achieving an excellent ratio of performance and cost", achieving an excellent performance balance through different accelerators and general-purpose processors, and co-designed with Mobileye software, so that the chip hash rate of 176 TOPS, equivalent to the sum of the performance of 10 EyeQ5 pieces, but the actual power consumption will be less than 100 watts, achieving a very high energy efficiency ratio.
The chip is based on the seventh-generation EyeQ chip technology architecture, using a 5nm process process, and contains four types of proprietary accelerator designs, each designed for a specific task. These accelerators, in conjunction with other CPUs, ISPs and GPUs, form an energy-efficient solution that addresses all the needs and application scenarios of L4 autonomous driving.
It can simultaneously process input data from two sensing subsystems (including a pure camera subsystem and a subsystem that integrates radar and lidar) as well as the vehicle's central computing system, high-definition maps, and driving decision software, providing the performance required for consumer-grade autonomous vehicles at a reasonable cost.
The EyeQ Ultra is expected to be available by the end of 2023 and fully operational in 2025.
The EyeQ Ultra was accompanied by two new EyeQ system integration chips for ADAS – EyeQ 6L and EyeQ 6H.
EyeQ 6L is the successor to EyeQ 4, its package size is only 55% of EyeQ 4, is currently Mobileye's most cost-effective product. This integrated windshield solution covers the needs of L2 assisted driving scenarios based on small size and high efficiency. Samples of the chip began last year and mass production is expected in mid-2023.
EyeQ 6H hashrates twice the power of the EyeQ 5, enables high-end ADAS and some autonomous driving functions, including visual parking and driver monitoring, and can perform better under heavy AI workloads, and is expected to be in mass production by the end of 2024.
As an extension of the EyeQ series, EyeQUltra will also integrate Mobileye's mapping platform technology to capture vehicle driving data in real time, assist vehicles in safe driving in unscripted route sections, and increase the upper limit of the driving assistance system.
Mobileye's road network collection management (REM) technology collects road information with the help of millions of vehicles that have been equipped with Mobileye chips on the road, and then creates a crowdsourced cloud database mobile roadbook with high-precision and high-definition maps, which provides relevant information about the driving route in front of it in the form of cloud access and real-time update, and automatically aggregates and generates autonomous vehicle maps in the cloud. Provide a truly global and scalable mapping solution for autonomous vehicles.
Volkswagen, Skoda and SEAT electric vehicles based on the Volkswagen MEB platform will soon introduce the Travel Assist navigation 2.5 system function equipped with the road book to improve the comfort of ADAS functions including lane keeping/centering.
Mobielye's REM mapping technology will be added to future versions of Ford's BlueCruise system and offer L2+ ADAS solutions for multiple brands and models.
On January 4, Mobileye and Extreme Kr announced that they will cooperate in the development of a new consumer-grade pure electric vehicle with L4 autonomous driving capabilities, and plan to launch the world's first L4 consumer-grade intelligent driving electric vehicle in 2024.
Krypton Auto said that the model will be built based on the SEA Haohan architecture, based on 6 EyeQ5 system chips, to achieve the efficient integration of redundant braking, steering and power functions of the Geely SEA Haohan architecture and mobileye True Redundancy sensing solutions and other technologies. It is expected to debut in China as early as 2024 and then launch it to the global market.
Qualcomm
Mobile phone chip giant Qualcomm actually entered the automotive field as early as 2002, but has been involved in the role of communication chips, by 2014, Qualcomm began to enter the field of smart cockpits, a few years later, 25 major automakers around the world, 23 have adopted Qualcomm's smart cockpit chip solutions.
After several years of hard work, Qualcomm's autopilot chip is about to be on the car in 2022, and at this CES, Qualcomm brought the overall solution of the family's barrel-style smart car - Snapdragon digital chassis , which supports high customization and consists of an open and scalable platform.
These include Snapdragon Ride, Snapdragon Smart Cockpit, Snapdragon Connected Platform and Snapdragon Car-to-Cloud Service.
The Snapdragon Driving Platform is an open, programmable autonomous driving platform that provides a comprehensive range of assisted driving and autonomous driving technologies that will focus on a wide range of customer autonomous driving needs, from scalable SoCs to integrated AD stacks, development platforms and tools, providing a comprehensive solution for L2-L3 autonomous driving.
At the hardware level, Qualcomm will provide manufacturers with a scalable portfolio of 4nm process chip (SoC) processors and accelerators for vision, central computing and high-performance autonomous driving needs, providing different levels of computing power, including 10 TOPS of computing power for automotive windshield ADAS cameras with less than 5 watts of power consumption, to more than 700 TOPS of computing power for fully autonomous driving solutions.
At the software level, based on Arriver's one-stop vision software stack, Qualcomm will support automakers and Tier-1 suppliers to create their own software functions related to autonomous driving, parking or driver monitoring, and also provide comprehensive support for advanced features, functional safety/ expected functional safety (SOTIF) and system architecture capabilities.
Snapdragon Smart Cockpit is the 4th generation Snapdragon Smart Cockpit platform launched by Qualcomm with 5nm Qualcomm SA8295P as the core, and its GPU 3D rendering performance has been improved by 3 times, and its AI computing power has reached 30TOPS. The smart cockpit provides a safer, immersive in-vehicle experience and provides manufacturers with a highly customizable and always-connected cockpit system-on-chip and software solution as required.
The 4th Generation Snapdragon Cockpit Platform addresses the computational, performance, and functional safety needs of smart cars for partitions or domains, delivering a wide range of capabilities such as high-performance computing, rich graphics and multimedia, highly intuitive AI experiences, situational awareness and security enhancements, and immersive audio.
Based on LTE, 5G, Wi-Fi and GPS, the Snapdragon Auto Connect platform will fully support the secure connection between the car and the cloud, other vehicles and the surrounding environment, meeting consumer demand for a safer and more immersive driving experience.
Snapdragon's cloud services provide automakers with a flexible feature portfolio and performance upgrades as well as new features for updates and fee-based services through pre-integrated software and service platforms designed for a new monetization model, and hope to open up a new monetization model.
It can be said that from software to hardware, from interconnection to cockpit, Qualcomm provides an open and scalable car platform that can provide automakers with a variety of tools to make cars more intelligent and connected.
At CES, Qualcomm also announced a series of new automotive partners, including Mercedes-Benz, BMW, Audi, Volkswagen Group, Volvo, Renault, Toyota, Honda, Nissan, GM, Ford, Stellantis, Hyundai and Mahinda and other foreign car companies, as well as domestic car companies such as Geely, Xiaopeng, Jietu, Weilai, Weima, SAIC and Jidu, integrating different Snapdragon digital cockpit platforms.
From the perspective of product layout, Qualcomm, which is famous for mobile phone chips, began to accelerate the layout of automotive products, but it is still mainly based on operating digital dashboards and infotainment system chips. In the fields of autonomous vehicle chips and cloud services, Qualcomm is still in its infancy.
Ampha
Ambarella is a chip manufacturer specializing in AI visual perception, and its products are widely used in the field of human and computer vision, including video security, advanced driver assistance systems (ADAS), electronic rearview mirrors, driving recorders, driver and cabin intelligent monitoring, automotive driverless and robotic applications.
At CES 2022, Ambarella released the latest SERIES of AI domain controller chips CV. Based on the scalable, energy-efficient CVflow architecture, the CV3 series of SoCs has a hashrate of up to 500eTOPS, which is 42 times higher than the previous generation of Autobard's automotive grade SoC CV2 series; at the same time, the CV3 is equipped with 16 Arm Cortex-A78AE CPU cores, which is 30 times higher than the CPU performance required for software applications that support autonomous driving systems than the previous generation chip CV2.
CV3 through a single chip integration of multiple sensors, centralized AI perception processing, multi-sensor deep integration and path planning of autonomous vehicles, through the built-in new generation OF ISP, as well as millimeter-wave radar calculation unit, CV3 can run Auco's 4D imaging radar AI algorithm, while supporting high-resolution cameras, millimeter-wave radar, lidar and ultrasonic radar, can help smart cars achieve a higher level of environmental perception accuracy under complex environmental conditions
With a single chip, the CV3 can be adapted to typical L2+ sensor configurations while providing a high-performance binocular stereo vision engine and dense optical flow engine for depth detection and motion sensing.
In addition to covering the entire autonomous driving domain, from ADAS driver assistance systems to L4 autonomous driving systems, the CV3 can also handle in-cabin perception simultaneously, including driver and in-vehicle perception systems. The advantage of this model is to help car companies reduce software development costs.
Aba said automakers no longer need to develop different software stacks for their entry-level, mid-range and high-end models because customers can use a unified CVflow platform across all models, saving engineering costs while being able to react more quickly to market changes.
In the Chinese market, the MAXIPILOT 1.0 L2 intelligent cruise system based on the Ambarco CV2x series has been delivered to passenger car customers for large-scale mass production, and the upgraded L2+ function is expected to be mass-produced in 2022.
At the same time, Ambarco has mass-produced and delivered a variety of models of in-cabin visual interaction, driving recorder, panoramic surround view and other front-loading solutions landed, the cumulative market shipments of more than 300,000 units.
NXP
At CES2022, Dutch automotive semiconductor supplier NXP Semiconductors announced that the industry's first 4D imaging radar processor S32R45 with a 16nm process launched at CES last year has been put into mass production, while the newly launched S32R41 is tailored for L2 and above autonomous driving functions, and has been adopted by 20 famous car companies around the world.
4D imaging radar will be able to achieve multiple functions such as mapping, positioning, object contouring, and object classification. Combined with cameras, or pattern recognition and machine learning, imaging radar systems can sense the surrounding environment at a high resolution of 1° azimuth and 2° elevation, with a ± of 60 degree FoV (also said to be 100 ° FoV), which enhances "perception ability" is extremely important for achieving full automatic driving in complex driving environments.
The flagship product of NXP's 6th Generation Automotive Radar Chipset enhances awareness and understanding of the complex urban environment, while providing short, medium and long-range 3-in-1 multi-mode radar sensing. Not only does it provide high-resolution images to ensure that all locations of the vehicle can be covered by radar, but it can also perform 360-degree safe surround detection of the car and support imaging radar target recognition and classification functions, which are essential for driving and automatic parking in urban environments.
The new S32R41 is the industry's first 16nm radar processor designed specifically for L2+ autonomous driving applications. With the transition from L2+ to L5 in the automotive autonomous driving level, with the blessing of the S32R41, NXP's 4D imaging radar is expected to cover more models.
NXP pointed out that L2+ scenes that cannot be properly handled by traditional high-resolution sensors can be solved by 4D imaging radar chips from up to 6 angles and 360 degrees of surround distribution. The automotive radar portfolio, combined with the NXP TEF82xx RFCMOS transceiver, provides automakers with the higher angular resolution, processing power, and application range they need.
Throughout this CES exhibition, the chip war around smart cars has witnessed the trend of "software-defined cars" from the side. In fact, in addition to the efforts of the above chip giants in the automotive market, domestic vehicle-grade chips - Huawei, Horizon, Black Sesame, Xinchi and other chip companies are also struggling to catch up, promoting the faster iteration and development of autonomous driving technology and intelligent network technology, and promoting the arrival of a new era in the automotive industry.