Intel moves to chiplets for automotive AI
Intel has taken a key step in using chiplets for its second generation software-defined vehicle system-on-chip (SoC) device.
The SoC, launched at the Shanghai 2025 Auto Show overnight, combines chiplets built on different process technologies to provide large language model AI support and more graphics support for user interfaces as well as 12 camera channels for ADAS driver assistance systems.
Intel has given some relative performance figures compared to a Core i7 processor without numbers, but we will update with more technical details of the chiplets in the device as they become available.
- Intel shows whole vehicle platform
- Intel launches its first dedicated graphics chip for automotive
- Silicon Mobility deal shows chiplet push
This puts Intel and its customers significantly ahead of other chip suppliers. imec is just setting up a lab in Germany to help car makers define and develop chiplets, while Japan is aiming to have chiplet-based automotive chips in production by 2028.
The advantage of a chiplet approach is that different chiplets can be produced on different technologies, for example with a 3nm digital processor for AI or self driving capabilities, and an optimised high speed analog process for camera interfaces.
In January, Intel showed an Adaptive Control Unit (ACU) that is an alternative to a microcontroller that avoids dataflow bottlenecks in real time, safety critical designs in autotmoive designs. This integrates a flexible logic area that offloads real time control algorithms from the CPU core.
At the same time it has developed a second generation of its standalone B-series Arc GPU for Automotive which is set for production by the end 2025.
Intel also signed strategic software deals with ModelBest in Beijing and Black Sesame Technologies in Wuhan to run on the SDV SoC.
Black Sesame is developing a central compute platform that fuses ADAS and immersive cockpit with its autonomous driving technology . This combines the SDV chip with Intel’s discrete Arc graphics chip. The Black Sesame HuaShan-2 A1000 chip is being used by many Chinese car makers for autonomous driving tasks,
A GUI Intelligent Agent from ModelBest allows on-device LLM operation with the SDV SoC and the Arc graphics chip. This enables offline, AI-enhanced voice control and personalized interaction without network connectivity. The agent enhances voice interaction by accurately understanding natural language in complex situations, ensuring an intuitive cockpit experience. This collaboration uses ModelBest’s experience with Intel’s AI PC acceleration programme to optimize AI for a seamless out-of-the-box experience on Intel’s automotive platform.
“Intel is redefining automotive compute with our second-generation SDV SoC, combining the flexibility of chiplet technology with our proven whole-vehicle approach. Together with our partners, we’re solving real industry challenges — from energy efficiency to AI-driven experiences — to make the software-defined vehicle revolution a reality for all,” said Jack Weast, Intel Fellow, vice president and general manager of Intel Automotive (above).
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
