Intel unveiled its first discrete graphics processing unit (dGPU) at its AI Cockpit Innovation Experience event today in Shenzhen, China. The Arc Graphics for Automotive is meant to address the growing demand for compute power in increasingly sophisticated vehicle cockpits, unlocking AI-driven cockpit experiences and enhanced personalization for manufacturers and drivers. The product will be commercially deployed in vehicles as soon as 2025.

Intel’s AI Cockpit Experience showcased a future where in-vehicle AI creates intuitive and personalized interactions between passengers and vehicles. The software-defined vehicle (SDV) system-on-chip (SoC) is said to offer automakers an open, flexible, and scalable platform solution so automakers can now design a single vehicle platform that scales across trim levels. The approach streamlines development as software is fully compatible across integrated GPUs (iGPUs) and dGPUs.

“Intel’s strategy is to bring the power of AI into devices of every size and shape, and we’re thrilled to bring that expertise and our vast open AI ecosystem to the automotive industry,” said Jack Weast, Vice President and General Manager of Intel Automotive. “China’s rapid electric vehicle development cycles and advanced technological adoption make it an ideal testing ground for our next-generation technologies.”

Automakers can choose an Intel SDV SoC and later add the discrete GPU for managing larger compute workloads and expanding the AI features thanks to a unified instruction set that simplifies the development of voice, camera, and gesture recognition.

In Shenzhen, a ThunderSoft demo showcased a new generation cockpit user interface (UI) that transforms vehicles into immersive mobile hubs by supporting seven high-definition screens rendering 3D graphics and six in-vehicle cameras and interactive features. The new UI showed its capacity to run high-demand AAA gaming titles while simultaneously running AI PC-rich applications for smart mobile working.

With powerful AI algorithms learning driver preferences, automakers can offer a highly personalized experience, adapting cockpit settings without voice commands. Zhipu AI showed an AI car assistant powered by local LLMs (large language models) running on Intel’s compute platform. The demo highlighted the ability to execute complex vehicle control commands through natural language processing, answer vehicle-related questions accurately, and even engage in leisurely chats with users.

Among the key features of its new offering are an open-source Linux-based automotive operating system and a graphics driver for both dGPUs and iGPUs. GPU single root input/output virtualization (SR-IOV) provides up to a 40% performance boost with a virtual machine, enhanced security, isolation, and robustness.

Intel optimizations integrated into industry-standard LLM frameworks are said to offer significant improvement in efficiency, customization, and customer experiences. An XeDisplay Engine supports an additional four display outputs and up to 4K resolution for additional in-vehicle infotainment use cases.

OpenGL and Vulkan open standard graphics and next-gen graphics APIs feature ray tracing for graphics realism. The high-performance AAA gaming is enabled by gaming-optimized XeHPG microarchitecture.

Intel aims to be the go-to partner for automakers with its XPU strategy, robust AI accelerator software ecosystem, and optimization capabilities aimed at providing a strong foundation for innovation in the automotive space. Its AI PC Accelerator Program brings together more than 100 independent software vendor partners that have created 500+ features and AI apps in an ecosystem that can be leveraged within vehicles.