On July 27th, at the 2024 NIO Innovation Technology Day, NIO Automobile's chairman Li Bin announced that the world's first automotive-grade 5nm intelligent driving chip, the Shengyuan NX9031, has successfully taped out, with both the chip and underlying software achieving independent design.
According to the introduction, this chip adopts a 32-core CPU architecture, built-in LPDDR5x, 8533Mbps rate RAM, has a pixel processing capability of 6.5GPixel/s, and the processing delay is less than 5ms.
Li Bin said that the Shengyuan NX9031 has more than 50 billion transistors, and whether it is comprehensive ability or execution efficiency, a self-developed chip can achieve the performance of four industry-leading chips.
This time, Li Bin's speech has caused some controversy in the semiconductor industry, because he claimed that the Shengyuan NX9031 is the world's first automotive-grade 5nm intelligent driving chip. However, before this, there have been 5nm chips that can be used for intelligent driving systems, with typical representatives being NXP's S32N55 processor and Ambarella's CV3 series domain controllers.
Advertisement
01
Where is the point of contention?
Li Bin said that the Shengyuan NX9031 is the world's first 5nm intelligent driving chip, which is debatable.
First, let's take a look at the 5nm automotive chips launched by NXP and Ambarella.It can be said that NXP was the first company in the industry to announce the production of automotive chips using the 5nm process technology. As early as June 2020, the company released this news, with TSMC as its wafer foundry partner.
The S32N55 processor, which uses the 5nm process, integrates 16 Arm Cortex-R52 real-time processor cores running at a frequency of 1.2 GHz, capable of meeting the high computational requirements of software-defined vehicles. The Cortex-R52 cores of the S32N55 can operate in either separated or lockstep mode, supporting the ASIL ISO 26262 functional safety level. Two pairs of auxiliary lockstep Cortex-M7 cores support system and communication management.
As the central vehicle controller solution for the S32 CoreRide platform, the S32N55 processor integrates advanced networking technology, with a Time-Sensitive Networking (TSN) 2.5 Gbit/s Ethernet switch interface, a CAN hub for efficient internal routing of 24 CAN FD buses, 4 CAN XL interfaces, and a PCI Express Gen 4 interface, enabling efficient communication and collaboration between various systems in the vehicle. In addition, the "core-to-pin" hardware isolation and virtualization technology of the S32N55 allows its resources to be dynamically partitioned to adapt to the ever-changing vehicle functional requirements.
In early 2022, Ambarella released the 5nm process CV3 series chips, capable of supporting the development of ADAS and L2+ to L4 systems. This series of chips is based on the scalable, high-energy efficiency ratio CVflow architecture, which can achieve 500 eTOPS computing power, 42 times higher than Ambarella's previous generation of automotive-grade CV2 series.
The "e" in eTOPS refers to equivalent. Since CVflow is not equivalent to any GPU, this leads to the counting unit of the AI computing power of the CV3 chip being different from the commonly used GPU's TOPS. Here, "e" is added to indicate that compared to the general chip architecture, it can achieve equivalent performance. NVIDIA's Orin chip has a computing power of 254 TOPS, and NIO ET7 achieves 1016 TOPS computing power through the cascading of 4 Orin chips. If 4 CV3 chips are cascaded, 2000 eTOPS computing power can be achieved.
In February 2023, Ambarella announced the production of the CV3-AD685 using Samsung's 5nm process technology.
Following NXP and Ambarella, Qualcomm's automotive chips also began to adopt the 5nm process. At this point, it must be said that NVIDIA and Intel's Mobileye, the intelligent driving chips of these two companies mostly use the 7nm process, while Tesla's HardWare 3 chip uses Samsung's 14nm process. Not long ago, the supply chain news reported that Tesla's new HW4.0 chip will switch to TSMC's 4nm/5nm process.
It can be seen that before NIO, NXP, Ambarella, and Qualcomm have all taped out 5nm process automotive chips. However, the types and applications of chips from these companies are still somewhat different. From the introduction above, it can be seen that NXP's S32N55 belongs to the control chip, while Ambarella, NVIDIA, Tesla, and NIO are computing chips, and Qualcomm's is an intelligent cockpit chip, which is more control-oriented.
Here is a brief introduction to the types of automotive chips, which can be divided into computing, control, analog, power, communication, sensor, power, and storage. Among them, computing and control belong to digital chips, with the highest process requirements. With the rise of intelligent driving, the requirements for chip computing power are increasing day by day. At this time, the computing power of computing chips has become a very key indicator, followed by control.
In summary, the industry's first to use the 5nm process technology to manufacture intelligent driving chips should be NXP or Ambarella. NIO is the first company in China to use the 5nm process to manufacture intelligent driving chips.So, why does NIO want to develop such a high-end chip on its own? It starts with NVIDIA.
At present, the most widely used flagship intelligent driving chip in the industry is NVIDIA's Orin-x, with a single chip computing power of 508 TOPS. In addition, NVIDIA also released a DRIVE Thor chip, with a single chip computing power of 2000 TOPS, which will only be mass-produced in 2025.
In 2023, NIO purchased a large number of NVIDIA intelligent driving chips, accounting for 46% of NVIDIA's shipments, with a total amount of 300 million US dollars. This is a huge expense. For NIO, which has been incurring losses in research and development, it is essential to reduce costs and increase efficiency in the increasingly competitive Chinese automotive market. Based on this, it is natural to develop its own intelligent driving chip. A single Shenji NX9031 can replace four NVIDIA Orin X chips, which can save a lot of chip expenses.
02
The value of making intelligent driving chips with 5nm process
Traditionally, automotive chips do not have high requirements for process technology (mostly above 20nm process), but have high requirements for the stability and reliability of the chips. That is to say, cars need to use automotive-grade chips (Automotive Grade Chip).
Automotive-grade chips refer to those chips that are designed and manufactured specifically for automotive applications and meet the stringent automotive industry standards. These chips need to maintain stable and reliable performance in harsh environments such as extreme temperature ranges, high vibration, high pressure, high humidity, EMI, etc., and usually need to pass automotive industry quality standards such as AEC-Q series certification.
Based on the extremely high requirements for safety and reliability of automotive applications, any chip failure can lead to serious safety accidents. Therefore, compared with consumer-grade or industrial-grade chips, automotive-grade chips have higher quality requirements. These chips are widely used in engine control, braking systems, safety systems, in-vehicle entertainment information systems, ADAS, and other in-vehicle subsystems.
Although advanced processes (16nm and below) can improve chip performance and reduce power consumption, they also bring some challenges. For example, the smaller the process node, the higher the production cost of the chip. In addition, small feature size chips require more precise production equipment and technology, which also increases costs.
Therefore, automotive chip manufacturers and automotive manufacturers need to find a balance between chip performance, cost, and reliability. They need to choose the appropriate process technology according to the vehicle's usage, performance requirements, and cost budget. For some high-end models, manufacturers may adopt more advanced processes to improve vehicle performance. For some economical models, manufacturers will choose more cost-effective process technologies to reduce production costs.In general, the mainstream manufacturing process for automotive chips ranges between 40nm and 16nm. However, with the popularization of intelligent driving, the traditional manufacturing framework for automotive chips has been disrupted, as computing power begins to dominate automotive applications.
Practitioners are well aware that the stacking of computing power will inevitably lead to waste, but compared to the invisible software algorithms, tangible computing power metrics can be easily judged. The pursuit of hardware capabilities is vividly reflected in mobile electronic products, and now the same situation has continued to intelligent vehicles.
Based on this, various marketing rhetoric has emerged in the market. For example, some media have compared the computing power level of chips to "room rate," using the differences between dense computing power and sparse computing power to calculate completely different computing power conclusions. Nowadays, the volume of computing power has become an insurmountable hurdle for car manufacturers and related chip companies, and more and more new intelligent driving chips have proven that increasing computing power is the most effective way to improve market evaluation levels.
At present, the computing power of many new force SUVs priced above 300,000 yuan has already exceeded 100 TOPS, and some brand cars have even exceeded 1,000 TOPS. Even with a large redundancy, it seems that no one would refuse higher computing power.
As chip computing power easily breaks through 500 TOPS, or even 1,000 TOPS, other indicators of the chip will inevitably attract public attention, such as manufacturing processes. Although there is no extreme pursuit of the process for automotive-grade chips, in the fields of intelligent driving and intelligent cockpits, the chip process has obviously begun to advance towards 5nm, or even smaller process nodes. Compared with 7nm, TSMC's 5nm process has increased the processing speed by 20% and reduced power consumption by 40%. Transitioning to 5nm will help car manufacturers bring differentiated advantages to their cars by enhancing functions, simplifying the increasingly complex architecture challenges of cars, and easily deploying powerful computing systems.
Therefore, the value of 5nm process for automotive chips is highlighted.
Under such market demand, TSMC has also started "scarcity marketing." In July 2023, TSMC's European General Manager Paul de Bot said at the "27th Automotive Electronics Conference" held in Germany that the automotive industry has long been considered a technological laggard, focusing only on mature processes. However, some automotive chip suppliers have been using 5nm process technology since 2022, just two years after 5nm officially entered mass production. Due to Samsung's average 5nm process yield, TSMC has almost become the only wafer foundry that can mass-produce 5nm process chips. As a result, the company's production capacity is in short supply, and TSMC stated that it is impossible to reserve idle production capacity for the automotive industry, and automotive chips need to accelerate the transition to advanced processes. Paul de Bot believes that it is absolutely necessary for car manufacturers to plan and control the order quantity in a forward-looking manner, and some automotive chips transitioning from the original mature process nodes to advanced processes are also an important means to ensure supply.
Compared with consumer electronics and server applications, the automotive chip wafer foundry market share is relatively small (less than 10%), but the unit price is higher, which is an extremely profitable market. From TSMC's perspective, during the COVID-19 pandemic, TSMC's automotive chip business has grown by about 40% each year. The wafer foundry leader hopes to retain and expand this customer base in the future, especially for advanced processes.The intelligent cockpit chip also needs 5nm
The chips introduced above are all intelligent driving chips, which are more computational, while another major category is the intelligent cockpit chip, which is more control-oriented, and the demand for advanced process technology is becoming more urgent.
The intelligent cockpit has multiple functional blocks, mainly including high-definition display, instrument panel, active safety alarm, real-time navigation, online information entertainment, emergency rescue, vehicle networking, and human-computer interaction system (voice recognition, gesture recognition), etc. Its main function is to improve the experience of drivers and passengers by changing the way of human-computer interaction. At this time, the importance of artificial intelligence (AI) technology is highlighted, and the performance requirements for related chips are also increased.
The typical representative of the intelligent cockpit chip is the Qualcomm Snapdragon 8155 that we often hear about, as well as the subsequent 8295 chip.
At the end of 2021, Qualcomm released the Snapdragon 8295, which adopts the 5nm process. Compared with the previous generation 8155 (7nm process) with a computing power of 8TOPS, the computing power of 8255 reaches 30TOPS, the 3D rendering capability is increased by 3 times, and functions such as integrated electronic rearview mirror, machine learning, passenger monitoring, and information security are added. A single chip can drive 11 screens.
In addition to the cockpit chip, the core SoC of Qualcomm's Snapdragon Ride intelligent driving platform is also based on the 5nm process and integrates core components such as high-performance CPU, GPU, and AI engine, with a maximum computing power of 700TOPS. However, compared with other major intelligent driving chip manufacturers (NVIDIA, Intel's Mobileye, Tesla), Qualcomm's intelligent driving chip presence is relatively weak.
In addition to major companies in the industry such as Qualcomm, Chinese local SoC companies are also moving towards advanced process intelligent cockpit chips, which have now reached 7nm. If there were not so many international trade restrictions, there would definitely be 5nm process chips. At present, Horizon Robotics, Black Sesame Intelligent, Xinchi Technology, and Xinjing Technology have all released relevant products. Among them, Xinchi Technology's self-developed "Dragon Eagle No.1" as the first domestic automotive-grade 7nm chip has been on the road, Black Sesame Intelligent has launched the first self-developed 7nm chip Wudang C1200, and Horizon's Journey 6 series chips also adopt the 7nm process technology, with the flagship chip's single computing power reaching 560TOPS, mass production in 2024, and it is expected to further improve the process technology when the Journey 7 or Journey 8 is mass-produced.
04
Automotive chips are waving to the 3nm process
With the improvement of the level of automotive intelligence, related chips are also advancing towards more advanced process technologies.NVIDIA's latest smart driving chip, DRIVE Thor, boasts a computing power of 2000 TOPS with the Hopper architecture; it will use a 4nm manufacturing process and is scheduled to go into production in 2025. NVIDIA has stated that Chinese automotive brands such as BYD, Aion, XPeng, Li Auto, and Zeekr will adopt the DRIVE Thor.
Tesla is even more aggressive, already preparing to initiate a 3nm manufacturing process chip foundry plan, further enhancing speed and power performance based on TSMC's N3E, with plans to start production in 2024. However, there are still doubts about whether they can secure production capacity.
Seeing Qualcomm's success in the smart cockpit application field, MediaTek can no longer remain idle and has also begun to enter the automotive chip market, especially in the smart cockpit. They plan to launch the "Tianji Automotive Platform," which will be built using a 3nm manufacturing process.
Leave a Reply