In the development of smart cars, autonomous driving systems are one of the most challenging technologies today. Today, each electric vehicle uses approximately more than 1,000 semiconductors, of which SoC is an essential semiconductor for automotive autonomous driving technology and multimedia systems, and they often require the most advanced semiconductor technology to achieve advanced computing capabilities. However, as computing power increases exponentially, the cost of autonomous driving chips will inevitably increase significantly. This poses a major challenge to the popularization of autonomous driving technology. After all, a car is still an end product in the final analysis, and there are strict considerations on cost and power consumption.
In order to solve this problem, using Chiplet design to build high-performance automotive chips has become a feasible solution. Chiplets are a modular approach based on heterogeneous integration that allows the number of transistors and other components to be scaled up without touching the physical limitations of a single chip. It is being implemented in various supercomputing applications, and cars cannot be left behind.
The reasons why autonomous driving chips are suitable for chiplet design are mainly based on the following aspects:
1) Performance requirements: Autonomous driving systems require extremely high computing power to process large amounts of data from various sensors (such as cameras, radar, lidar, etc.). Chiplet design allows for a more flexible combination of different processing units (such as CPU, GPU, NPU) to meet these high-performance needs.
2) Energy efficiency: Autonomous driving systems require efficient energy management to extend the cruising range of electric vehicles. Chiplet design can manage energy consumption more effectively and improve overall energy efficiency by integrating different functions on multiple smaller chips than on a single large chip.
3) Cost-effectiveness: Manufacturing large, complex systems on a chip (SoC) is expensive, and the defect rate during production may be higher. Chiplet design can reduce production costs and defect rates by using multiple smaller chip components with different processes, thereby reducing overall costs. Although lower production costs will be partially offset by higher packaging costs, overall, the use of chiplets is expected to save up to 40% compared to traditional monolithic designs.
4) Customization and scalability: Intelligent driving technology is constantly evolving towards the L2/L3/L4 level, and different car models have different requirements for chips. Do they need to use multiple chips? In this regard, chiplet design allows for more flexible customization and expansion of autonomous driving systems. Depending on the needs of different models and autonomous driving levels, different chiplet combinations can be selected to provide the best performance and functionality.
5) Adaptability to technological progress: In the rapidly developing field of autonomous driving technology, chiplet design provides the ability to adapt to new technologies faster. For example, a specific processing unit can be upgraded individually without replacing the entire system.
6) Car manufacturers participate in chip definition: Nowadays, in order to control their own destiny, many car manufacturers have stopped making chips, and Chiplet gives car manufacturers the opportunity to participate in chip definition and design, and even the opportunity to dominate key chips.
In terms of research and development of automobile chiplet technology, Japan is very powerful. On December 1, 2023, 12 leading companies in Japan formed a group called "Automotive Advanced SoC Research" (ASRA) and will begin installing SoCs in mass-produced vehicles from 2030. These chips will focus on the development of artificial intelligence accelerators, graphics engines and enhanced computing capabilities, and are planned to achieve mass production in 2030. The group is chaired by Toyota and includes well-known automakers such as Nissan, Honda, Mazda and Subaru. In addition, chip suppliers such as Renesas Electronics, Mirise Technologies and Socionext, as well as first-tier suppliers such as Denso and Panasonic Automotive Systems, also participated and jointly served as executive directors. At the same time, companies such as Cadence Design Systems and Synopsys are also involved, providing the necessary EDA (electronic design automation) development tools for Chiplet design. It is particularly worth mentioning that Renesas Electronics has adopted a small chip architecture in its fifth-generation R-Car X5 high-performance automotive SoC.
Wishlist (0 Items) 