Will x86 architecture chips dominate the smart cockpit again?
Latest update time：2024-08-29
Reads：
You and I both know that the world is dominated by two major instruction set architectures: x86 and Arm. The former dominates the PC field, and the latter dominates the mobile phone and portable field.
In the field of smart cockpits in cars, early intelligent upgrades were all made by Qualcomm or MediaTek from chips based on mobile Arm architecture. Therefore, it is natural that they dominate the cockpit field, forming the golden combination of "Arm+Android".
It is said that cars are "computers on four wheels", so is it also an idea to move x86 computers into the cockpit? In fact, in recent times, x86 has been very active in the field of smart cockpits. After Tesla used x86-based chips, x86 became popular in the cockpit field.
Fu Bin | Author
Automotive Development Circle | Produced
Two major manufacturers began to exert their strength
AMD and Intel, the two major manufacturers leading the x86 market, have worked particularly hard in the past two years, especially in recent times, frequently announcing progress related to automotive chips.
First of all, AMD not only launched automotive chips to compete with Qualcomm and Nvidia, but also continued to cooperate with car companies and software manufacturers:
On December 22, 2022, GAC announced that the ADiGO SPACE smart cockpit host uses AMD V1000 and W6600 series chips to create the highest computing power car cockpit in China, comparable to high-performance PCs, and can play 3A game masterpieces without any pressure, marking that AMD has won the first smart cockpit of a domestic car company;
In March 2023, as the world's first Tier 1 partner of the AMD V2000A series, Ecarx launched the Ecarx Makalu computing platform, equipped with an x86-based AMD V2000A processor and an AMD Radeon RX 6600 series discrete graphics card (optional). Through its self-developed underlying virtualization technology, it supports different operating systems, such as instrument OS (Linux), Android, and Game OS (3A games).
At CES 2024, AMD launched two in-vehicle AI chips, the Ryzen Embedded V2000A for smart cockpits and the Versal AI Edge XA adaptive chip for ADAS. The former is built on a 7nm process, a "Zen 2" core, and a high-performance AMD Radeon Vega 7 graphics card;
On August 6, 2024, AMD and Thundersoft reached a strategic cooperation. The main content of the cooperation between the two parties is the smart cockpit. AMD will provide the new generation of automotive APU hardware, and Thundersoft will be mainly responsible for the software platform;
On August 6, 2024, Lynk & Co. said that its Lynk X Lynk & Co. Z10 smart cockpit hardware is equipped with the high-performance Makalu computing platform jointly launched by Ecarx Technology and AMD, which can support Unreal Engine and full-scene 3D immersive cockpit HMI experience.
Next is Intel, which began to focus on x86-based SoCs and dGPUs as AI big models emerged:
At CES 2024, Intel will launch the first software-defined automotive SoC chip, which is also the world's first automotive-grade chip using Chiplet. Intel uses the more mature Intel 7 process technology. It is expected that Intel Intel 4 and Intel 3 processes will be integrated into it in the future to achieve iteration. Geely's Zeekr has made it clear that it will use this software-defined automotive SoC chip;
Later, Neusoft Group announced that it would launch Neusoft C4 4.0 intelligent cockpit domain controller based on Intel Malibou Lake;
On August 8, 2024, Intel released its first sharp car-mounted independent graphics card (dGPU) Arc A760. It is based on the x86 architecture and equipped with Xe graphics engine and xmx computing engine. With the powerful computing power of 229TOPS, it will help automobile manufacturers enter the forefront of technology and bring the ultimate generative AI experience to the smart cockpit.
On August 8, 2024, Thundersoft, a partner of ThunderSoft, demonstrated the "Dripping OS Cockpit" developed based on Intel's solution, which supports 7 high-definition screens for 3D image rendering, 6 on-board cameras and interactive functions. Thanks to the rendering performance of Intel's discrete graphics card, the reflections of vehicles and grass in the rain on the car interface have light and shadow effects close to 3A games.
However, despite its efforts, x86 still cannot beat Arm in the overall smart cockpit market. Currently, the standard configuration of the main control chip for the smart cockpit of the car is SoC system-level chip + MCU. Among them, SoC chip is an important growth point in the market, Qualcomm occupies a dominant position, and the mainstream framework adopted is Arm + Android.
According to Gasgoo's "Smart Cockpit Supplier Installed Volume Ranking from January to June 2024", Qualcomm is far ahead with more than 1.55 million installed units, accounting for 66.7% of the market share, while AMD and Intel in the x86 camp account for 12.1% and 2.2% respectively, which means that x86 only has nearly 14.3% of the market share.
Why do these two manufacturers work so hard? Why hasn't x86 become the mainstream in car computers? Will x86 eventually beat Arm? This starts with the history of car computers.
The history of vehicle computer development
Speaking of the past of smart cockpits, perhaps in line with the familiar phrase "software defines cars", Android was ported to cars earlier, driving Arm to enter car computers earlier.
The earliest car computers can be traced back to the end of the last century. At that time, the entertainment information system in the car was nothing more than radio, tape, CD player, DVD player, MP3, car navigation, insert SIM card to make calls, but it was still relatively traditional and had not yet formed a system-level concept. In order to make the car more competitive, manufacturers at that time added equipment without restraint. The result of crazy addition was that the car cockpit was full of buttons and instruments. At that time, the driver's operation was no less cumbersome than that of the pilot. To subtract from the center console, there must be a screen or system to cover all the software and functions.
In the late 1990s, the concept of car computers gradually took shape, and the world's first UNIX-like, POSIX-compliant microkernel hard real-time operating system QNX was born. Since its appearance, an average of 20 million smart cars equipped with BlackBerry QNX basic software have entered the global market each year. At the same time, Microsoft also developed the Win CE system for embedded systems, which was later upgraded to Windows embedded automotive.
In 2000, some were happy while others were sad. QNX was at its peak while Microsoft withdrew from the market. However, history began to turn around. With the emergence of car dashboards and central control screens, Linux systems began to attract widespread attention. In 2007, the iPhone became popular, which further promoted the application of Android. Android successfully entered the car with its open source characteristics and rich ecological construction. As the most suitable chip architecture, Arm won the market in one fell swoop.
In 2012, Tesla released the Model 5, which put a 17-inch screen on the car, shocking the world that a smart cockpit could be built this way. The system of the Model 5 was rewritten by Ubuntu, a Linux distribution, and Tesla continued to launch new products to drive the large-screen car center console.
Until today, QNX, Linux and Android have become the three major systems for car computers. Although they can run on chips of any architecture, they still have biases. The dashboards of electric vehicles focus on real-time performance, and most of them are built based on QNX. QNX can be used in combination with Linux and Android, so they can be simply divided into x86+Linux and Arm+Android.
This software shift has also led to a huge change in the chip industry. In the past, traditional Tier 1 companies such as Bosch, Continental, and Delphi were the mainstays of the cockpit. Later, professional semiconductor companies such as NXP, TI (Texas Instruments), and Renesas Electronics brought more powerful computing capabilities to the cockpit, and the chips of these companies also underwent strict automotive certification. Now, the times have changed, and changes in software have made car computers consumer electronics. Nvidia, AMD, Intel, and Qualcomm have started to attack from a lower dimension. Their smart cockpit chips are often directly derived from mobile phone SoCs or computer chips.
Another factor that has led to this change in the world is centralization. As the electrification process deepens, the automotive E/E architecture has gradually changed from traditional distributed to domain control and zonal control. In this case, automotive chips have gradually formed domain or regional controller chips. Therefore, before 2015, the cockpit was mainly controlled by MCU or low-computing power SoC. As the functions gradually become richer, high-computing power smart cockpit SoC has become mainstream.
Smart cockpit infrastructure
x86 has relatively strong computing power at this stage
Today's cars are no longer the same as they were in the past. In the past, cars may have been tools, but now they are more like consumer electronics or toys. In addition to the core three electrics, smart cockpits have become the decisive battleground for car companies, and naturally they are more "competitive".
3D HMI, 3D face recognition, AI large models, integrated HUD or augmented reality system, streaming central rearview mirror, Internet of Vehicles module, SRV surround view... Today's car companies have "played around" with the cockpit. They not only want you to work, watch audio and video, and entertain in the car, but also want you to play 3A games in it.
The reason why x86 is becoming popular again in the cockpit field may be due to the rise of large AI models. At present, the tasks that SoC chips have to undertake are too heavy, and the simple and crude computing power makes x86 start to compete directly with Arm.
Experts mentioned that computing power is an unavoidable factor in the current cockpit. If the 6B-7B large model is compressed to a 30-40TOPS NPU, the response time of the first token will basically be more than 3 seconds, which is unacceptable to users.
Let's first look at the Arm camp. Qualcomm's Snapdragon 8155 chip is sought after by almost all car companies, and Qualcomm's 8295 chip has also been supported by many car companies such as Leapmotor, NIO, Zeekr, Xiaomi and Mercedes-Benz. The AI computing power of Qualcomm's most powerful smart cockpit chip Snapdragon 8295 is only 30TOPS; NVIDIA's DRIVE Orin has an AI computing power of 254TOPS; the most powerful Horizon Journey 6 series, the Journey 6P, has a computing power of 560TOPS (but it is composed of 4 128TOPS BPUs, and is mainly aimed at intelligent driving); the latest C1200 series launched by Black Sesame Intelligent has an AI computing power of less than 100TOPS; TI TDA4 VH is a benchmark. Although its AI computing power is only 32 TOPS, it is very "playable" through flexible chip combination and algorithm adaptation.
Looking at the x86 camp, Intel released its first sharp car-mounted independent graphics card (dGPU) Arc A760 with a powerful computing power of 229TOPS; AMD's V1000 series CPU computing power reaches 210kDMIPS, GPU computing power reaches 1.5+10TFLOPS, and the CPU performance of V2000A is about 360~370kDMIPS, which is 88% higher than the previous generation V1000 series and better than Qualcomm's SA8295P.
x86 also has its own advantages
Of course, computing power is only one side of x86. Compared with Arm, it also has its own characteristics.
There are many discussions about x86 and Arm, which are nothing more than the difference between CISC (complex instruction set) and RISC (reduced instruction set). It is obvious that Arm's reduced instruction set will have certain advantages in terms of multi-core number and performance-power ratio. Then the entry of x86 into the cockpit will inevitably increase power consumption and heat dissipation. So, apart from computing power, where are the advantages of x86? After all, Arm can increase computing power by stacking the number of cores?
In comparison, x86 instructions are relatively complex and more numerous, and can handle relatively special and complex tasks. At the same time, compared with the Arm architecture, it has enhanced out-of-order execution capabilities. When users use the car computer, the operation is random and unpredictable, which makes the instructions unpredictable, so out-of-order execution capabilities are needed.
In addition, the X86 architecture uses a "bridge" method to connect with expansion devices, and has rich interfaces, so the performance of the x86 architecture car machine can be easily expanded, such as adding memory, hard disk, etc. The Arm architecture CPU connects to the data storage device through a dedicated data interface, so the Arm architecture's storage, memory and other performance expansions are difficult. Coupled with the multi-threaded feature, that is, the performance ceiling of the x86 architecture will be higher.
More importantly, the x86 ecosystem has always been stronger than Arm. In this case, the cockpit is not just a matter of computing power, but a matter of the entire ecosystem. In other words, the computing power must be strong and able to run all kinds of software and all kinds of programs. Perhaps this is the advantage of x86.
Copying Tesla's old path
If we talk about the most representative x86 car machine, it must be Tesla. Most car companies that want to adopt the x86 architecture may be copying Tesla's path.
Judging from Tesla's smart cockpit chip, in 2021, MCU 1.0 used Nvidia Drive CX; MCU 2.0 was launched in 2017, switching from Nvidia to Intel's Atom E3950; the 2021 Model S Plaid's MCU 3.0 uses AMD Ryzen+Navi 23 GPU.
After Tesla updated its AMD chip, the performance and fluency of its in-car system are comparable to or even better than that of a laptop computer. The AMD Ryzen processor chip is very powerful in comprehensive control. It uses a 12nm process technology, has a 4-core 8-thread CPU and an independent GPU. Many industry insiders have commented that the Tesla after the "chip" change is simply the most powerful car machine.
Ever since Tesla switched to AMD chips, AMD has become popular, so there is the current market. At present, all Tesla Model series are equipped with AMD Ryzen V1000. According to data from Zoss Automotive Research, AMD Ryzen V1000 has begun to take shape in the market, accounting for 6.1% of the market share of L1 and above smart cockpit SoCs.
But looking into the future, Nvidia's 1000TOPS Drive Atlan will be released in 2025, and the computing power of Arm SoC will expand rapidly. If we rely solely on computing power, this path may not work. However, judging from AMD's recent actions, it is not only moving fast in CPU and GPU, but also putting FPGA and adaptive SoC into car computers to perform tasks such as DPU or in-car communication.
Perhaps this is also the advantage of x86, that is, to put everything that has been done in the data center and PC in the car computer. Or perhaps, only by copying the past success in the car computer can x86 really win.
References
[1] Zos Automotive Research: Cockpit Domain Control Research: Facing x86 AI PC, multi-domain computing, and domestic substitution, how does cockpit domain control differentiate itself from the competition? .2024.7.17.https://mp.weixin.qq.com/s/9q-QDusCFfHW6_Ld3FN1Bg
[2] Gasgoo Automotive Community: Ranking of installed capacity of smart cockpit suppliers from January to June 2024: Technology is sinking, and domestic manufacturers are accelerating their production. 2024.8.10. https://mp.weixin.qq.com/s/ertqUPifwbW0dREDxf1lEA
[3] Hardware Product Manager Agent Wang: When it comes to car computer development, I only admire Tesla. 2024.4.25. https://mp.weixin.qq.com/s/7BkR9U_w5opToT6jA8twQQ
[4] Cyber Auto: In a tough fight with Qualcomm, what kind of cockpit products does Intel have? .2024.8.13.https://mp.weixin.qq.com/s/v0coLAkwpffcOIRMzZ9dhg
[5] Automotive Frontline Observation: x86 Chips, Counterattack in Smart Cockpits. 2022.12.15. https://mp.weixin.qq.com/s/Xfmr9xFQmsMYMs01UYAjOw
· END ·
What else do you want to see about automotive electronics? Please leave a message below to communicate!
You are welcome to set us as “
Star
” so that you can receive push messages as soon as possible.
Scan the QR code to follow:
Automotive Development Circle
, reply "
BMS
"
Get
the Battery Management System BMS Learning Collection!
Scan the QR code to add the assistant and reply "join the group"
Exchange experiences with automotive developers face to face
Two major manufacturers began to exert their strength
The history of vehicle computer development
Smart cockpit infrastructure
x86 has relatively strong computing power at this stage
x86 also has its own advantages
Copying Tesla's old path
京公网安备 11010802033920号