An in-depth report on the automotive chip industry Altera
Being able to drive involves human-computer interaction, visual processing, intelligent decision-making, etc. the core is AI algorithm and chip. With the acceleration of automotive electronics, the growth of automotive semiconductors accelerated. In 2017, the global market scale was US $28.8 billion (+ 26%), which was much higher than the growth rate of vehicle sales (+ 3%), of which the functional chip MCU (US $6.6 billion, accounting for 23%), followed by power semiconductors (21%) and sensors (13%).
Automotive semiconductors can be divided into functional chips MCU (microcontroller unit), power semiconductors (IGBT, MOSFET, etc.), sensors and others. According to strategy analytics, among traditional fuel vehicles, the value of MCU accounts for the highest, accounting for 23%; In pure electric vehicles, MCU accounts for 11%, which is second only to power semiconductor. DIGITIMES predicts that the market scale of functional chip MCU is expected to steadily increase from US $6.6 billion in 2017 to US $7.2 billion in 2020.
▲ global automobile sales (10000 units)
▲ global automotive semiconductor market size (USD 100 million)
▲ fuel automobile semiconductors are classified by type
▲ pure electric vehicle semiconductors are classified by category
▲ market scale of automotive functional chips (USD 100 million)
The function chip of the traditional automobile is only applicable to the local functions such as engine control and battery management, and cannot meet the relevant calculation of intelligent driving with high data volume.
In recent years, with the increase of the penetration rate of intelligent driving, global chip giants have entered the automobile industry one after another and launched master chips with AI computing capability. The market scale of master chip is expected to grow rapidly. IHS predicts that it will reach US $4 billion in 2020.
The master chip giant has strong AI computing advantages, and the functional chip manufacturers have rich experience in the automotive industry chain. Mergers and acquisitions and alliance cooperation between the two camps occur frequently.
Up to now, NVIDIA has reached cooperation with more than 370 vehicle manufacturers and first-class suppliers in the world; Intel's acquisition of Mobileye entered the automotive industry; Qualcomm once intended to acquire NXP.
▲ automobile chip market pattern
Main control chip: computing power continues to grow
Intelligent driving involves human-computer interaction, visual processing, intelligent decision-making, etc. AI algorithms and chips are the core. According to the statistics of NXP, at present, a high-end car has carried more than 100 million lines of code, which is far more than that of aircraft, mobile phones, Internet software, etc. in the future, with the penetration and level of automatic driving, the number of lines of code carried by the car will show exponential growth.
The calculation amount of automatic driving software has reached the order of 10 TOPS (Tera operations per second). The computing power of traditional automobile MCU is difficult to meet the computing requirements of autonomous vehicle. AI chips such as GPU, FPGA and ASIC have entered the automobile market
The world's unmanned driving leaders include Google, Baidu, Tesla, Audi, etc. from the SOC chip architecture of these manufacturers' automatic driving main control module, you can see the development direction of automobile chips.
Google waymo: Intel CPU + Altera FPGA solution is adopted, and Infineon MCU is used as the communication interface. Google waymo's computing platform uses Intel Xeon 12 core or above CPU, with Altera's Aria series FPGA, and Infineon's aurix Series MCU as the communication interface of can or FlexRay network.
▲ Google waymo's computing platform architecture
Baidu Apollo: NXP / Infineon / Reza MCU + Xilinx FPGA / NVIDIA GPU. Baidu driverless prototype adopts IPC (industrial control computer) scheme, but the volume and power consumption of industrial control computer are difficult to meet the requirements of mass production. Therefore, Baidu has also launched a domain controller embedded scheme suitable for mass production. Connect the original data of each sensor to the sensor box, complete the data fusion in the sensor box, and then transfer the fused data to the computing platform for automatic driving algorithm processing.
Baidu's automatic driving special computing platform ACU (Apollo computing unit) defines three series of products: mloc (high precision positioning, MCU), mlop (high precision positioning + environment awareness, MCU + FPGA), mlop2 (high precision positioning + environment awareness + decision planning, MCU + GPU).
Tesla: from Mobileye ASIC to NVIDIA GPU. In 2014, Tesla released autopilot 1.0, which is equipped with a front camera, a rear reversing camera (not involved in auxiliary driving), a front radar and 12 ultrasonic sensors. The vision chip is mobileyeq3, and the main control chip is NVIDIA Tegra 3.
At the end of 2016, Tesla released autopilot 2.0, which is equipped with 3 front cameras (wide-angle, long focus and medium viewing angle), 4 side cameras (left front, right front, left rear and right rear), 1 rear camera, 1 front radar (enhanced version), 12 ultrasonic sensors (the sensing distance is doubled). The main control chip uses NVIDIA drive PX 2, and the processing speed is 40 times that of autopilot 1.0.
Audi: Mobileye ASIC NVIDIA GPU + Altera FPGA + Infineon MCU multi chip integration solution. The new Audi A8 has disclosed its own zfas controller scheme. Zfas has four high-performance processors: 1) Mobileye's eyeq3 is responsible for visual information processing, including traffic sign recognition, pedestrian recognition, collision warning, lane line detection, etc; 2) NVIDIA's Tegra K1 SOC is responsible for 360 ° panoramic image; 3) Altera's cyclone5 FPGA is responsible for sensor fusion, map fusion, auxiliary parking, etc; 4) Infineon's aurix Series MCU is used for traffic congestion control, auxiliary driving, etc.
In the field of automobile main control chip, GPU will still maintain the mainstream position of general automobile main control chip, FPGA as an effective supplement, ASIC will become the ultimate direction.
At present, artificial intelligence and intelligent driving algorithms have not been finalized. As a general accelerator, GPU is expected to maintain its mainstream position as the main control chip of automobiles for a long time; As a hardware accelerator, FPGA is expected to be an effective supplement to GPU; In the future, if all or part of the intelligent driving algorithms are solidified, ASIC will become the ultimate choice for optimal cost performance.
▲ trend chart of automobile main control chip
1. NVIDIA: GPU monopoly advantage, from intelligent cockpit to autopilot
NVIDIA's revenue and net profit grow rapidly, and the automobile is the long-term power. NVIDIA is the leader in the GPU field, maintaining a market share of more than 70% all the year round. NVIDIA's revenue in fiscal year 2018 (corresponding to 2017 natural year) was US $9.71 billion, a year-on-year increase of + 40.6%; The net profit was US $3.05 billion, a year-on-year increase of + 82.9%.
▲ global unique GPU market share (2009-2017)
NVIDIA digital cockpit computer drive Cx: use advanced 3D navigation, high-resolution digital instrument cluster, natural voice processing and image processing to realize driving assistance. The core of drive CX is TEGRA X1 SoC Based on Maxwell architecture. In addition, Tegra K1 SOC is selected.
The main functions of drive CX include: 1) natural language processing, address query and contact call through voice recognition; 2) 3D navigation and information entertainment, providing high-resolution and high frame rate graphic display for many applications; 3) Full digital instrument cluster, providing rich graphic display through instrument cluster or head up display HUD; 4) Surround vision: using complex motion recovery structure technology and advanced splicing technology, improve the image rendering of fish eye lens, reduce ghost phenomenon, and render a virtual car in high-precision model to achieve realistic surround vision effect; 5) Docking with Android auto, drivers with Android smartphones or iPhones can easily access their mobile devices and interact with applications such as maps, search and music.
NVIDIA autonomous vehicle platform drive PX: combines deep learning, sensor fusion and surround vision to strive to change the driving experience. The main functions of drive PX include: 1) sensor fusion, which can fuse data from 12 cameras, laser radar, millimeter wave radar and ultrasonic sensor; 2) Computer vision and deep neural network, suitable for running DNN (deep neural network) model, can realize intelligent detection and tracking; 3) End to end high-definition mapping, which can quickly create and constantly update high-definition maps; 4) The software development kit driveworks contains reference applications, tools and library modules.
2. Intel: actively merge and acquire, enter into the special chip for automatic driving
Intel's traditional business growth is weak, and it has entered the automobile field to create a new growth point. Intel was once the world's largest semiconductor chip manufacturer.
According to passmark's statistics, Intel's market share in the global CPU industry in Q1 2017 was 80%. In recent years, with the rise of smart phones and the decline of the personal computer market, the growth rate of the main business income of chips has significantly decreased, and the company's operating income has been surpassed by Samsung Electronics. The company tried to produce a mobile phone processor but failed in the end and had to disband the Department in charge of the business.
In recent years, Intel has actively deployed new fields such as driverless, Internet of things, artificial intelligence and VR through a large number of acquisitions to create new growth points of performance and strive to transform from a traditional chip manufacturer to a diversified solution provider. Intel acquires Mobileye: global leader in visual ADAS. Mobileye is one of the leaders in the global visual ADAS market. It holds 80% of the ADAS market and has rich visual ADAS products. Mobileye's proprietary software algorithm and eyeq chip can analyze visual information in detail and predict possible collision with other vehicles, pedestrians, bicycles or other obstacles, and also detect road signs, traffic signs and traffic signals.
By the end of 2017, Mobileye's products had been used in 313 models of 27 vehicle manufacturers, with a shipment of 8.7 million in that year. In March 2017, Intel purchased Mobileye with us $15.3 billion to build Intel's fleet. The fleet will include various automobile brands and models to demonstrate their versatility and adaptability. L4 class vehicles will be deployed in the United States, Israel and Europe for testing.
Intel acquires Altera: FPGA chips for automatic driving have been mass produced. At present, the global FPGA market is mainly divided by Xilinx and Altera, accounting for nearly 90% of the market share and more than 6000 patents.
Altera's FPGA products have four major series, namely, the Stratix series (nearly US $10000), the aria series (US $2000-5000), the low-cost cyclone series (US $10-20), and the max series CPLD. In 2015, Intel announced the completion of the acquisition of Altera to help the fast-growing data center and IOT business.
3. Qualcomm: from infotainment to Internet of vehicles with communication advantages
Qualcomm's traditional business income has declined, and it is actively engaged in the layout of emerging industries. Qualcomm is the global leader in smart phone SOC.
In the automotive field, the solutions provided by Qualcomm include: 1) in vehicle information system, cellular network solutions optimized for automobiles; 2) The driving data platform intelligently collects and analyzes data from different vehicle sensors to enable the vehicle to achieve accurate positioning, monitor and learn the driving mode, sense the surrounding environment, and accurately share the information of this platform with the outside world; 3) Infotainment, providing 3D navigation, online media playback and parking assistance support, as well as voice, face and terminal recognition functions; 4) Electric vehicle wireless charging, launched Qualcomm halo wevc wireless charging solution.
Qualcomm launches in vehicle infotainment system solutions. Snapdragon automotive platform information and entertainment system is now divided into select, high and premium solutions.
The minimalist solution can support three displays, including infotainment system, instrument and head up display (HUD); The high-end level can support up to 4 display screens. The front passenger or rear seat entertainment can have a separate screen. At the same time, it also supports top-level audio, low delay wireless transmission of high-definition video, surround view processing, and deep learning and computer vision processing to distinguish obstacles and pedestrians nearby; The top-level solution can support up to 6 screens, including instrument, infotainment system, HUD, copilot and rear seat (two different screens).
At CES 2017, Maserati's hardware was equipped with customized snapdragon automotive solutions, including snapdragon automotive processor, gobi3g / 4G LTE wireless modem, Wi Fi and Bluetooth modules. Another participating car, Chrysler portal, is equipped with Panasonic on-board entertainment concept system, which will be based on the latest version of Android cars and Qualcomm snapdragon chips.
Qualcomm launched the Internet of vehicles chipset, which supports LTE and DSRC. Snapdragon X5 LTE supports LTE vehicle networking, with a speed of up to class 4, a downlink speed of 150Mbps and an uplink speed of 50Mbps. Snapdragon X12 LTE supports 10 categories of high speed, and supports downlink rates up to 60 MHz 3x Ca (450mbps) to 40MHz 2x Ca (100Mbps) in the network uplink.
Snapdragon x16 LTE modem supports a peak download speed of up to 1Gbps, which helps meet the connection needs and use cases of the next generation of intelligent connected vehicles, including high-definition map updates, real-time traffic and road information connection navigation, software upgrades, Wi Fi hotspots and multimedia streams.
In addition, in September 2017, Qualcomm launched the world's first cellular car to car (c-v2x) commercial solution based on the third generation partnership project (3GPP) version 14 specification, the Qualcomm 9150 c-v2x chipset. The chipset includes an application processor running the v2x stack of the intelligent transportation system (ITS) and a hardware security module (HSM). It is expected to be launched in the second half of 2018, and will be mass produced and supplied to the vehicle manufacturer as early as 2019. C-v2x supports both DSRC and LTE communication and provides