In 1885, Karl Benz invented the first automobile, which ushered in a new era of human transportation. He never imagined that more than 130 years later, the phrase "a supercomputer on four wheels" would echo in the sky of the times repeatedly, and no one knew about it.
Every time you hear a similar statement, the car chip that underpins the car to become a supercomputer will be honored.
01
Who can give the optimal solution to the three roadmaps of automotive chips?
Creating a selling point is a key point in the automotive marketing process.
In the past two years, car companies have understood one thing: the intelligent cockpit is not omnipotent, but without it, it is absolutely impossible. Judging from the data, the market is also feedback, and the intelligent cockpit installation rate of more than 500,000 market of automobile manufacturers has reached more than 80%.
This raises the question, what kind of cockpit chip to choose?
From the perspective of basic requirements, at least a high-computing power on-board chip is needed. Under the requirements of "one core and multiple screens", the computing power requirements of cockpit chips have caught up with or even exceeded those of mobile phone chips. Volume computing power has become an unavoidable hurdle for automotive chips.
From the perspective of future development, AI large models will be overwhelmingly moved from the cloud to the device side, and the landing terminals are not only mobile phones, computers, but also cars. Chips that integrate generative AI technology are naturally a rigid demand for car companies.
The next question is, who can provide such a cockpit chip?
At present, there are three different types of routes for automotive chips.
The first category is traditional car core manufacturers, such as NXP, Renesas, etc.; the second category is PC chip manufacturers, such as Intel, AMD, etc.; and the third category is mobile chip manufacturers, similar to MediaTek, Qualcomm, Samsung and other manufacturers that cut into mobile phone AP chips.
The market for automotive chips has already begun to change. More and more chip manufacturers are pouring into the automotive track, but the ultimate winner must be the one who has the ability to manufacture high-computing power chips and the experience of integrating generative AI technology and ecology into the terminal.
Who is the manufacturer that can meet the two?
The answer is not difficult to answer: MediaTek.
Traditional manufacturers struggle with the design and manufacture of high-computing power chips, so we won't go into details. PC manufacturers have inherent advantages in computing and processing power, and are currently mainly attacking autonomous driving chips.
Mobile chip manufacturers such as MediaTek not only have successful experience in mobile fields such as smartphones, but also have significant advantages in low power consumption, high performance, integration, application ecology, etc., which are more in line with the needs of the automotive cockpit chip market.
02
MediaTek, accumulating growth
In the past two years, Qualcomm has been in the limelight in the field of intelligent cockpit chips. According to the statistics of Gasgoo Automotive Research Institute, in 2023, the installed capacity of Qualcomm cockpit domain control chips will exceed 2.26 million, accounting for nearly 60% of the market.
In the field of mobile chips, MediaTek has always been a strong competitor to Qualcomm, and this competition has also continued to the automobile.
In fact, MediaTek has long been an "old player" in the automotive field. MediaTek has been deeply involved in the automotive field for more than ten years based on its flagship market experience, and last year released an integrated automotive solution, the Dimensity Automotive Platform, covering the four business directions of Dimensity Automotive Cockpit Platform, Dimensity Automotive Connectivity Platform, Dimensity Automotive Driving Platform, and Dimensity Automotive Key Components, providing a portfolio of automotive solutions with high computing power, high intelligence, energy saving, and reliability as its core advantages.
Judging from the latest business situation announced by MediaTek, the Dimensity automotive platform is growing explosively. Among them, the Dimensity automotive cockpit platform has shipped more than 20 million units worldwide, and the 3nm flagship and 4nm sub-flagship cockpit chips have been adopted by more than 6 domestic leading car manufacturers.
03
There is an inequality in automotive chips
Those who know change with the times, and those who know make things as they go.
Wanting to combine AI and high computing power at the same time, MediaTek has found a strong partner.
In 2023, when MediaTek announced that it would cooperate with NVIDIA to provide a complete AI intelligent cockpit solution for software-defined vehicles, it caused a thousand waves with one stone.
Judging from the cooperation plan announced by MediaTek and NVIDIA, the two sides will fully enter the fields of intelligent cockpit chips and automotive SoCs, which is equivalent to directly entering the hinterland of Qualcomm.
One is the super giant of the mobile chip market, and the other is the absolute king of the AI chip market, and the combination of the two hegemons has begun to shake the automotive chip market.
In the future intelligent cockpit, a single SoC can support digital cockpit, ADAS and autonomous driving functions, which requires the computing power of the chip to not only meet the data calculation, but also the graphics computing and artificial intelligence computing, and can achieve the stability and safety of the vehicle specification level.
Judging from the current officially announced cooperation, the cooperation between MediaTek and Nvidia has brought three impacts to the automotive market.
First, volume computing power.
As mentioned above, intelligent cockpit chips require high computing power, and high-process process technology is a necessary condition to support high computing power.
At present, mainstream intelligent cockpit SoC chips have basically implemented sub-10nm processes, including Samsung V9 with 8nm process, Qualcomm 8155, Huawei Kirin 90A, and SiEngine SE1000 at 7nm.
Even automotive chip companies that have the ability to invest in automotive cockpit systems, such as Texas Instruments, NXP Semiconductors, and Renesas Electronics, can only use the most advanced process at the 5nm node at this stage.
On April 26 this year, MediaTek officially launched the 3nm process Dimensity automotive cockpit chip CT-X1, as well as the 4nm process CT-Y1 and CT-Y0. MediaTek is the first in the industry to use the 3nm process in the field of intelligent cockpit SoC chips, and the advanced process will greatly release the computing power of the chip and make the performance of the automotive cockpit leap.
As early as March this year, MediaTek has launched Dimensity automotive cockpit chips that combine AI technology: C-X1, C-Y1, C-M1 and C-V1, all of which support NVIDIA DRIVE OS software, covering the automotive segment from luxury to entry-level. It is worth noting that these Dimensity car cockpit platforms also integrate NVIDIA RTX GPUs, which can support ray tracing technology.
Ray tracing technology is needless to say about the improvement of game graphics, but it can be used in automotive chips, which means that the in-cabin equipment of the car has a major breakthrough in graphics processing capabilities. In the past, due to hardware limitations, the graphics processing of cockpit chips was often not comparable to that of PCs or game consoles, but this combination allows users to enjoy a more realistic visual experience in the cockpit of the car, which is indeed surprising.
Second, volume AI.
MediaTek and Nvidia both agree that AI will shape the future. For example, Renjie You, Senior Vice President of MediaTek and General Manager of the Metaverse Business Group of Computing Unicom, said: "Just as the personalized and intuitive computing revolution has been brought about in the mobile communications market, generative AI is also transforming the automotive industry. ”
Ali Kani, vice president of NVIDIA's automotive business unit, also said, "Generative AI and accelerated computing are reshaping the automotive industry. ”
In fact, MediaTek's combination of NVIDIA technology with Dimensity automotive cockpit chips (SoCs): C-X1, C-Y1, C-M1 and C-V1 is unique in that it embodies AI technology in all aspects. On the one hand, these cockpit chips integrate NVIDIA's next-generation GPU-accelerated AI computing and NVIDIA RTX graphics processing technology, and on the other hand, they can support the running of large language models (LLMs) in the car, which makes it possible for large AI models to be put on the car and support advanced AI safety and entertainment applications such as in-vehicle voice assistants, multi-screen displays, and driver alertness monitoring. These applications run on the automotive end-side to improve safety and enjoy the benefits of high responsiveness and low latency.
Third, roll the future.
In the context of the integration of the two sectors of intelligent driving and intelligent cabin, the "integration of cabin and cabin" has also been put on the agenda.
Cockpit convergence refers to the integration of the cockpit domain and the intelligent driving domain in a high-performance computing unit, which supports both intelligent driving and intelligent cockpit functions.
In 2023, there will be 6,346,700 new passenger cars in the Chinese market (excluding import and export) equipped with cabin driving intelligence (L2 and above intelligent driving + intelligent cockpit) as standard, a year-on-year increase of 66.37%, and the penetration rate will exceed the 30% mark for the first time. According to the forecast, this number will most likely exceed the 50% mark this year.
But it is not a simple matter to achieve "cabin and driver integration".
From the perspective of the supply chain, the "cabin and driver integration" puts forward higher requirements for the participants. On the one hand, it needs to have the mass production and delivery capabilities and large-scale effect of the cabin and the driver at the same time, and on the other hand, it needs to have cross-domain capabilities for basic elements such as computing platforms and software.
And these two capabilities are a perfect match for MediaTek and Nvidia.
Behind the cockpit integration, it is also an important consideration for the cooperation between MediaTek and NVIDIA. In terms of intelligent cockpit, MediaTek has already accumulated deep experience in the field of automotive chips. In terms of intelligent driving systems, NVIDIA has also launched platforms such as DRIVE AGX Orin.
Good computing power, smarter AI, and a broader future all show that the joint force formed by the combination of MediaTek and Nvidia seems to be greater than that of Qualcomm.
Inequalities in the field of automotive chips are starting to emerge.
04
MediaTek set off a new trend
During the Beijing Auto Show on April 26, MediaTek released a new product of the Dimensity automotive cockpit platform.
Among them, the Dimensity automotive cockpit platform CT-X1 uses a 3nm process, and the CT-Y1 and CT-Y0 use a 4nm process. The three chips are based on the Arm v9 architecture, with a built-in HDR ISP image processor, and support multiple intelligent image optimization technologies such as AI noise reduction and AI 3A, as well as advanced device-side generative AI technologies.
Specifically, MediaTek's new cockpit computing chip has excellent performance.
In terms of AI large models, CT-X1 can support up to 13 billion parameters of Al large language models, and CT-Y1 and CT-Y0 can support 7 billion parameters of AI human language models. You must know that the large models that can run on cars in the industry can support up to 1 billion parameters. In addition, the three chips can also run a variety of popular large language models (LLMs) and Al drawing functions (StableDiffusion) in the car.
The Antutu car machine version of the 4nm process sub-flagship CT-Y1 chip has a running score of more than 1.07 million, which has equaled the Snapdragon 8295 and reached the top flagship level. The flagship CT-X1 built on the 3nm process was unveiled at the same time, with 30% better performance than the Snapdragon 8295 (which is a very conservative number from the architectural point of view), and 4-5 times stronger AI computing power.
In fact, whether it is the cockpit scenario or the autonomous driving scenario in the future, there are extremely high requirements for computing power, and the final route is to go to the advanced manufacturing process.
MediaTek's core technology is computing, which is the computing power of CPU, GPU and APU, and the support behind high computing power is the mass production capacity of advanced processes. In the medium to long term, we will find that IC design companies with advanced process capabilities will be able to gain a foothold in the future generative AI-defined smart car market.
The Xiangjiang River brings the sea, and the wind is long. With its successful experience in mobile chips, MediaTek deeply combines AI and computing technology to drive the development of smart cars and reshape the future of automobiles.
In 2024, the market structure of the automotive market is still uncertain.