News

Intel Arcs into automotive with a dGPU

Company launches its discrete GPU for vehicle AI enhancements.

Jon Peddie

Intel unveiled its first automotive dGPU, Intel Arc Graphics, to empower automakers with AI-driven cockpit experiences and personalization capabilities. Commercial deployment is planned for 2025, offering a flexible and scalable platform solution for high-fidelity vehicle experiences, simplified development, and personalized rides.

Intel Auto
(Source: Intel)

Analysts and reporters following the automotive industry have been wondering about two issues: what’s going on in China, and what’s Intel doing in automotive. Intel sought to answer both in what it labeled a strategic move to empower automakers with groundbreaking opportunities. At its AI Cockpit Innovation Experience event in Shenzhen, China, the company unveiled its first dGPU, the Intel Arc Graphics for Automotive.

Intel plans to commercially deploy its automotive product in vehicles by 2025, hoping it will drive advancements in automobile technology and enable new AI-driven cockpit experiences and personalization capabilities for manufacturers and drivers.

Intel says discrete GPUs for automotive address the growing need for compute power in vehicle cockpits. By combining Intel Arc graphics for automotive with its AI-enhanced software-defined vehicle (SDV) system-on-chips, the company says it is offering automakers a flexible and scalable platform solution. Intel believes this will enable the creation of high-fidelity experiences in vehicles.

The company says that with its system, automakers can design a single-vehicle platform that scales across different trim levels. The Intel SDV SoC can be used in entry and mid-level models, while higher-end vehicles can utilize the discrete GPU for additional premium features. This approach simplifies development, as software is compatible across both integrated and discrete GPUs.

Intel envisions a human-centric automotive future where in-vehicle AI creates intuitive and personalized interactions among drivers, passengers, and vehicles. This future enables personalized rides and optimized drives.

The key features of the Intel automotive dGPU are:

  • Automotive software and optimized drivers: Open-source Linux-based automotive operating system and graphics driver for both dGPU and iGPU.
  • Intel GPU single-root input/output virtualization (SR-IOV) advantage: It provides up to a 40% performance boost with a virtual machine and enhanced security, isolation, and robustness.
  • Large language model (LLM) frameworks designed for automotive: Intel optimizations are integrated into industry-standard LLM frameworks, offering significant improvements in efficiency, customization, and customer experiences.
  • Xe display engine: This engine, says Intel, 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 API featuring ray tracing for graphics realism.
Intel Arc Graphics for AutomotiveArc A760A
Xe cores28
Ray-Tracing Units28
Matrix  Engines for AI  (XMX)448
Vector Engines448
Memory Size/Interface16GB/256bit
GDDR6 Capacity (GB)16
PCI Express  ConfigurationsGen4 x16
SR-IOVYes
TFLOPS FP32Up to 14
Peak TOPsUp to 229
TBP (watts)225
No. of Display Ports4
Media DecodeAVC, HEVC, VP9, AV1
Media EncodeAVC, HEVC, VP9, AV1
O/S, hypervisor, orchestration ACRN Hypervisor, Linux (Yocto), Android, Linux-in-Container within Android virtual machine
Supported Technologies Vulkan, OpenGL, OpenCL, OpenVINO, OneAPI, Proton, virGL, Venus, SR-IOV, virtIO display
Auto TemperatureExtended temp support     (-40°C to 105°C)
Auto AvailabilityQ1 2025

Intel has reference design that it says enables the local operation of large language models. According to Intel, this means:

  • Lightning-fast performance: Enjoy low-latency responses and seamless interactions, even offline.   
  • Unwavering privacy: All data processing happens within the vehicle, eliminating the risk of information leakage and ensuring a rider’s privacy is always protected.
  • Seamless interaction and intuitive understanding: Always-on voice dialog, allows natural, continuous conversations with the AI assistant for effortless control.
  • Precise intent recognition: The system intuitively understands the driver’s and riders’ needs, and delivers relevant services efficiently.
  • Immersive and engaging experience: High-quality voice interaction allows for crystal-clear communication with the AI assistant.
  • Immersive 3D HMI display: Experience a cinema-grade visual interface that enhances every interaction.

The company has also signed up a collection of partners to contribute to the automotive environment.

ThunderSoft. Founded in 2008 and headquartered in Beijing, ThunderSoft has utilized Intel Automotive Solutions to develop an in-vehicle cockpit platform. The company’s SenseAuto Cockpit Multimodal Interaction platform features capabilities such as screen and camera expansion, deployment of computationally demanding applications, and support for AI-powered applications.

  • Screen and camera expansion: It supports up to seven high-definition screens with 3D rendering and six in-vehicle cameras for driver monitoring, panoramic exterior perception, and more.
  • High-computational-demand applications: Enables AAA gaming and deployment of large local AI models for enhanced productivity and creativity.
  • Mobile office: Introduces light office applications and a rich app ecosystem to support smart mobile working within the vehicle.

Zhipu AI Car Assistant. Founded in 2016 in Shanghai, Zhipu’s AI Car Assistant is powered by Zhipu large language models and optimized for Intel’s  high-compute platform, including its powerful discrete graphics processing unit (dGPU). This seamless integration delivers a smoother, richer, and more enjoyable driving experience. Highlights include:

  • Effortless vehicle control: Goes beyond simple commands. The AI understands complex and nuanced instructions, like “It’s too hot in here,” to automatically adjust the car’s settings for optimal comfort.
  • Interactive car manual: Say goodbye to cumbersome paper manuals. Users can receive accurate real-time answers by asking questions directly while driving. Snap a photo of a malfunction, ask a question, and receive instant, accurate answers and solutions directly via the AI assistant.
  • On-the-go companion: Enjoy engaging conversations, role-playing games, and continued story chats. The AI adapts to riders’ interests and uses multimodal capabilities to incorporate real-time information like vehicle recognition and illustrative drawings for a truly immersive experience.

Baichuan NPC AI Companion. Baichuan Information Technology Co., Ltd., also known as Baichuan NPC, was founded in 2015 in Hangzhou, China. The Baichuan NPC AI Companion is a specialized conversational chat non-player character (NPC) engineered for AI cockpits. It incorporates advanced natural language processing technology, enabling smooth, natural, and lifelike dialog experiences.   

Users can engage with the NPC using voice commands to enjoy an emotionally responsive    companion. The NPC is designed to adapt its responses based on the user’s tone and context,    ensuring seamless and coherent conversations.

Setting the Baichuan NPC AI Companion apart is its highly personalized customization features. With just a single phrase, users can create unique characters. This showcases not only the    system’s flexibility but also opens a world of possibilities for future interaction methods in AI cockpits.

Intel believes a state-of-the-art AI cockpit should incorporate three fundamental capabilities, such as cockpits featuring 4K high-refresh-rate displays and fully-3D user interfaces to deliver an immersive experience for users. They also should have enhanced sensing: privacy and safety prioritized in cockpit designs that can capture panoramic views outside the vehicle while maintaining strict privacy inside. And, such cockpits should be equipped with multi-modal interaction, LLMs, and cockpit AI perception, providing them with cognitive abilities.

Of course, that could be achieved with an Intel dGPU.