Is the future of cars in the intelligent cockpit?

Text | Semiconductor industry

165438+1October15th, Qualcomm announced that its fourth-generation Snapdragon intelligent cockpit platform has been produced. This generation of intelligent cockpit platform is named SA8295, which supports the new shipment of Peugeot 308. At the beginning of the year, the fourth-generation Snapdragon intelligent cockpit was launched by Qualcomm in an online activity, using 5nm technology and equipped with KryoCPU, HexagonDSP, AdrenoGPU and SpectraISP from Qualcomm. Qualcomm claims that CPU and GPU are more than half of the previous generation SA8 155.

Qualcomm announced the prototype of the fourth generation Snapdragon intelligent cockpit platform. Source: International Digital Technology Exhibition

On the 28th of last month, the first 7nm smart cockpit chip "_ Eagle-1" in China was successfully released. This chip will be tested on Geely Automobile and developed by Xinqing Technology. Xinqing Technology "_ Eagle No.1" adopts 7nm process and integrates 8.8 billion transistors, as well as CPU, GPU, ISP, DSP and LPDDR5.

Intelligent cockpit has gradually become the standard of modern cars, and both at home and abroad have attacked the city in this field. At the same time, one core and multi-screen become the main architecture form of intelligent cockpit, and the cockpit chip is driving in the fast lane.

Competition pattern of players in big factories

According to a data report from car home, the global smart cockpit market will reach 44.7 billion US dollars in 2020, 6,865,438+billion US dollars in 2030, while the smart cockpit market will only reach 3.9 billion US dollars in 20 16. According to the comprehensive data, the market scale of intelligent cockpit in China is expected to reach 654.38+003 billion yuan in 2025. At present, the penetration rate of new cars equipped with intelligent cockpit in China market is about 48.8%, and it is expected to exceed 75% by 2025, which is higher than the assembly rate in the global market.

Intelligent cockpit is reflected in the upgrade of ADAS and in-vehicle entertainment system, and under the premise of intelligent and digital car building, the car factory needs to expand and merge a single ECU.

Therefore, SoC master control including CPU, GPU, ISP and DSP has become the mainstream choice.

Before 20 15, the operation and control of on-board system were mainly based on MCU and SoC with low computing power, and the suppliers were Renesas, NXP and Texas Instruments. In the process of transition to the intelligent cockpit stage, the three companies still occupy a large share. After Intel acquired Mobileye, it enhanced the strength of intelligent driving, and Qualcomm was a blow. After the second generation of products, the market share rose steadily, becoming the largest manufacturer at this stage, and it took several years to become the first in the industry competition.

R-CarGen3 is a classic intelligent cockpit product of Renesas, which was launched on 20 18. It is based on the ArmCortex-A57/A53 kernel and uses the Arm64-bit architecture. Gen3 series includes products from R-CarV3M to R-CarH3, in which R-CarV3M is suitable for entry-level models and R-CarH3 is specially designed for high-end models. In 20021year, Renesas introduced new R-CarGen3e, including R-CarD3e, R-CarE3e, R-CarM3e, R-CarH3e and R-CARH 3e. Compared with R-Cargen 3, the CPU performance of Gen3e is improved to 50kDMIPS and 2GHz. Renesas continued to use the architecture of ArmCortex-R7 and successfully migrated gen3e and Gen3. At present, customers of Gen3e include Great Wall and Volkswagen.

In addition to Renesas, traditional auto parts manufacturers Texas Instruments and NXP also have market share, but the share of these traditional manufacturers is gradually declining. In 20 18, NXP mass-produced i.mx master chips, which has been 5 years since the release of 20 13. In the past five years, NXP lost the opportunity of intelligent cockpit. In the same year, Qualcomm announced that its acquisition of NXP was "stillborn" and eventually paid NXP a $2 billion breakup fee.

The unsuccessful acquisition may have a considerable impact on NXP's career. After i.mx8, NXP did not launch new high-end products. The shipment of i.mx8 is also unsatisfactory. At present, NXP's main intelligent cockpit chip is i.mx6 released 10 years ago, which has been widely used in domestic entry-level cars and occupies a place.

A smart cockpit chip on Texas Instruments is Jacinto6, which was released on 20 16. For cost-sensitive customers, it still uses the ARMCortex-A 15 kernel. Texas Instruments released the intelligent cockpit chip Jacinto6 in 20 16, and has been using the ARMCortex-A 15 kernel for cost-sensitive customers. At the beginning of 2020, Jacinto7 processor series was released, focusing on ADAS, which became another heavy product of Texas Instruments. The main customers of Texas Instruments intelligent cockpit chips are Japanese cars and German cars.

ApolloLake is an onboard processor launched by Intel in 20 16. Now it's the Atom series. Based on Intel's Goldmont architecture, the chip has 2-4 CPU cores and multiple video output interfaces, which can support cockpit experiences such as in-vehicle infotainment systems. It is Intel's flagship product in the cockpit.

Architectural drawing of Apollo Lake. Source: Intel

As for Qualcomm, which has the largest shipment at present, Qualcomm has landed three generations of intelligent cockpit platforms, and the third generation intelligent cockpit platform SA8 155 was launched two years ago. Qualcomm officials said that among the top 25 auto companies in the world, 23 have purchased the SA8 155 intelligent cockpit platform. SA8 155 can be regarded as a classic intelligent cockpit platform at present. Weimar W6, Geely Xingyue L and Tucki P5 have all been equipped with Qualcomm SA8 155 intelligent cockpit platform.

As the main control chip of the third generation intelligent cockpit, the operation speed of SA8 155P is three times that of the second generation intelligent cockpit chip 820A. In related reports, the computing speed of SA8 155P is 3.6 million times per second, and SA8 155P also supports a new generation of networking technologies, including WiFi6 and Bluetooth 5.0. Compared with its predecessor, 820A155 has smaller volume, lower calorific value and higher computing power. It is reported that Qualcomm's third-generation on-board chip 8 155 accounts for more than 80% of the intelligent cockpit of electric vehicles.

Qualcomm began to lay out the intelligent cockpit market in 20 14, and Qualcomm launched its first-generation intelligent cockpit platform-Snapdragon 602A in 20 14 1 year. 602A is based on Snapdragon 600 platform, which meets a series of requirements such as 4G, gesture recognition, car WiFi and so on.

In 20 16, Qualcomm successfully launched the second generation intelligent cockpit S820A, which became the main fighting force for Qualcomm to gradually annex the intelligent cockpit market. Qualcomm launched the second generation intelligent cockpit, which stood out from the fierce offensive from Intel, Renesas, Texas Instruments and NXP with its great technological progress and ecological capabilities. 820 chip adopts KryoCPU, quad-core, 64-bit. At present, S820A's customers in Qualcomm include Ideal, BYD, Audi A4, Toyota Accord, Volkswagen Golf and other domestic and foreign models.

Samsung is a manufacturer that only joined the smart cockpit SoC in the past two years. At this year's CES, Samsung demonstrated the driving scene of its smart cockpit. Samsung's smart cockpit SoC mainly includes Exynos8890 and Exynosatov 9, and the related organizations said that the overall performance of Exynosatov 9 is equivalent to SA8 155P. This year, Audi suddenly turned around, left its original partner, turned around and embraced Samsung Exynos8890, and entered into a long-term contractual relationship.

New people emerge one after another, each with his own merits.

In addition to the above semiconductor manufacturers, Huawei's semiconductor department has also been producing products in recent years.

Huawei's smart cockpit chips are mainly Kirin series, including 7 10A released last year and Kirin 990A released in April this year. Kirin 990A is based on ARMV7A, in which Kirin 990A can also replace SA8 155P, which has a higher AI instance. 990A has been used in some models of Arctic Fox Alpha S and BYD.

Huawei intelligent cockpit architecture diagram. Source: Hayes official website

At present, the chips in the intelligent cockpit are mostly CPU built by ARM, which has a certain market share in both consumer products and car regulations. Traditional automobile chip manufacturers have met the challenge of consumer chip manufacturers. At this stage, Huawei HiSilicon Kirin has become the leader of domestic intelligent cockpit chips, and Geely's core engine technology has become the representative of self-research by automobile manufacturers. Internet companies such as Baidu have also built Apollo's self-driving cars and put forward the business direction of "intelligent creation".

Some intelligent cockpit semiconductor companies also have a certain scale, and Nanjing Xinchi Semiconductor is one of them. Its self-developed X9 series processor is an automotive chip specially designed for the new generation of automotive electronic cockpit, with a built-in 64-bit ARMCortex-A55 core. According to the company, X9 finally reached 100KDMIPS, 3D performance reached 300GFLOPS, AI computing reached 1.2TOPS, and there were at most 8 full HD displays. In addition to hardware, Xinchi Technology also launched the UniDrive fully open autopilot platform to help the X9 chip land.

Autopilot companies settled in earlier.

Some companies specializing in autonomous driving technology are not behind. 20 18, Horizon released Journey 5 series ASIC chips. Based on 16nm technology, Journey can be used in dynamic driving and intelligent cockpit scenes, and supports interactive and equal energy functions. In-cabin perceptual computing such as DMS and multimodal voice interaction has become a bright spot.

In September, NavInfo mentioned on the investment platform that the intelligent cockpit chip AC80 15 independently developed by Hefei Jiefa Technology, a subsidiary of NavInfo, was mass-produced and shipped in the first half of the year. It adopts quad-core ARMCortex-A53 architecture, and 80 15i, AC80 15H and AC80 15M are included in 8065438. At present, the chip has passed the AEC-Q 100Grade3 certification, and the company said that the smart cockpit chip has entered the Tier 1 car factory catalogue.

Block diagram of AC80 15 chip system of jiefa intelligent cockpit. Source: Jiefa Technology

In addition to companies that focus on autonomous driving such as Jiefa Technology and Horizon, Minieye is one of them. This company was founded on 20 14, and mainly engaged in intelligent driving and in-vehicle applications. In 20 18, minieye introduced X 1, a pre-ADAS product, using XilinxFPGA technology. At present, it has been mass-produced and put into use in trucks, hazardous chemicals transportation companies and other fields. Minieye's intelligent cockpit pays more attention to safety scenes, which can detect the gaze direction, facial state and behavior posture of drivers and passengers in real time, judge passengers' fatigue, distraction and dangerous behavior, and create a new in-vehicle interaction experience. Minieye said that FPGA is more suitable for deep learning, and a mature R&D system can reduce costs and improve efficiency.

Mini-eye cabin technology. Source: minieye

The industry is still developing.

Player identity fusion. The research and development of intelligent cockpit chips is not difficult and the cost is not high, so it attracts a large number of semiconductor companies with different genes to join, and automobile manufacturers, autonomous driving, communication manufacturers and traditional chip manufacturers all flock, with different architectures, based on different usage scenarios, modes and consumption levels. China is the country with the largest number of cars, and domestic manufacturers are not backward in technology, but they are expected to lead.

At the same time, the intelligent cockpit technology is still changing. ECU stacking was the main mode of cockpit in the past. Now the intelligent cockpit and autonomous driving are developing in parallel, and the automotive electronic and electrical architecture is transitioning to centralization. Some analysts said that in the near future, automobile chips will form a dual-brain structure of intelligent cockpit domain main chip and automatic driving domain main chip, domain will become the main control form, and automobile chips will evolve to "central computing platform" in the future. After 2030, the electronic and electrical architecture of self-driving cars will develop into a strip architecture based on domain integration, and the intelligent cockpit and autonomous driving will be gradually integrated into the central computing chip.

In addition to hardware,

Software has also become a problem that must be considered in the intelligent cockpit at this stage.

In other words, the overall solution and ecological creativity of the intelligent cockpit have become an important starting point for increasing market share. Previously, ChenDexin, a senior analyst of automotive technology at IHSMarkit, predicted that the software related to electronic and electrical architecture would be more perfect and OTA would be widely popular. In the past, automobile manufacturers did not think too much about software, or regarded software as the core of manufacturing or defining cars. This situation will change from now on.

But on the whole,

The hardware of the intelligent cockpit still determines the ceiling of the product.

This is also their golden age.