Discrete graphics processing units (dGPUs) are pieces of computer hardware with their own dedicated memory separate from the central processing unit (CPU). While more power and heat intensive than integrated graphics, dGPUs can provide enhanced image rendering for systems and applications with advanced processing needs. In the tech world, they are generally found in high-spec desktop computers and gaming laptops.
In the automotive industry, dGPUs provide an opportunity to facilitate computationally intensive tasks and next-generation user experiences as OEMs develop their software-defined vehicle (SDV) concepts. Immediate enhancements could include more immersive infotainment systems with high-resolution displays to create richer in-cabin experiences, but the future also promises breakthroughs in the deployment of artificial intelligence (AI).
In August 2024, Intel announced that it would bring its first dGPU specifically for automotive—Intel Arc A760A—to commercially deployed vehicles from Q1 2025. “Our current family of SDV SoCs already offers significant scalability, but some automakers just want more,” Jack Weast, Intel Fellow and Vice President of Intel Automotive, tells Automotive World. “We see dGPUs as a crucial component in the development of SDVs.”
Pushing the envelope
The Intel Arc A760A dGPU unit spec features 28 Xe-cores with 16GB of 256-bit memory and 225W TBP, making its performance equivalent to or better than some of the company’s desktop computer GPUs. Weast explains that it can support up to four in-cabin display screens with 4K resolution, sophisticated 3D human-machine interfaces (HMIs), and smooth AAA gaming.
He proposes that infotainment systems powered by dGPUs could soon rival home theatres in terms of quality and immersion. However, differentiation is key in a highly competitive market. Beyond improving familiar aspects of the in-cabin experience today, Weast adds that Intel Arc provides automakers with scope to “push the envelope” of HMI performance in the SDV era.
Although GPUs are generally used for image and video creation, they are also well suited for deployment in AI applications. This is because, unlike the sequential functionality of CPUs, they employ parallel processing to handle multiple calculations simultaneously. As such, GPUs can train, run, and iterate AI models in a fast and energy efficient manner. This is highly advantageous in the automotive space, where power usage is limited and automated functions often require split-second decision making. Intel states the Arc A760A can deliver up to 229 TOPS for AI inferencing.
The possibilities could prove game-changing. Through its specialised large language model (LLM) frameworks, Intel can enable interactive 3D navigation systems responsive to voice commands and gestures, as well as AI-powered co-pilots that provide occupants with personalised recommendations and assistance. The same technology can also perform predictive vehicle maintenance and monitor driver drowsiness and distraction in real time.
A powerful, scalable platform
As brands shape their SDV transition, they will need to determine how best to incorporate new and enhanced features across their product portfolios, and Intel Arc was conceived with flexibility in mind. “It’s designed to work alongside our AI-enabled SDV SoCs, creating a powerful and scalable platform,” says Weast. “This allows automakers to choose the optimal configuration for different vehicle models and price points.” For example, entry-level volume models might only feature basic AI, while luxury vehicles could have the highly personalised cockpits that are often integral to the segment.
Intel is “actively engaging” with automakers and partners in key markets to realise dGPU-driven innovation for a wider audience of customers. However, Weast highlights that China has been an important developer market for AI cockpit solutions in particular, citing its “rapid adoption of new technology, robust automotive ecosystem, and government support for intelligent vehicles.” Subsequently, Intel has managed to produce several compelling use cases for advanced graphics processing in vehicles.
Most prominently, Weast highlights tech company Zhipu, which optimised proprietary LLMs to run on the Intel Arc A760A and produce its AI Car Assistant. As a result, the feature can understand nuanced voice prompts and act according to predetermined user preferences—for example, simply saying, ‘It’s too hot in here,’ would cool the cabin down to an appropriate temperature. The company claims its responses are sophisticated enough to become an “on-the-go companion” capable of maintaining conversations, playing games, and using multimodal capabilities to maximise engagement. Zhipu AI Car Assistant can also function as a personal engineer, answering specific and complex user questions about the vehicle, meaning the driver doesn’t have to consult the owner manual.
dGPU-driven differentiation
While Intel is not the only company exploring AI and accelerated computing solutions in an SDV context—Nvidia is a notable competitor—Weast believes his company’s product gives it an advantage over other players. “Our approach simplifies development with a single software stack for both integrated and dGPUs, reducing costs and time-to-market.” He adds that Arc’s capacity to facilitate a high level of processing power in the vehicle without the added latency of cloud-based analytics will enable the development of SDV features that previous generations of hardware made either impossible or impractical to realise.
Although industry consensus has yet to solidify around what ‘software-defined’ comprehensively means, many commentators believe it is ultimately predicated on updateability. As SDV concepts meet the reality of a varied global market, system agility will likely become table stakes. “As technology advances and consumer demands grow, OEMs have the flexibility and assurance that the Intel system is built on an open platform that can easily be upgraded to meet these needs without requiring a complete overhaul,” says Weast. In a virtuous cycle, the availability of powerful and scalable hardware will encourage software developers to create innovative applications and in-cabin experiences, nurturing sufficient customer interest in SDVs to generate a lively partner ecosystem dedicated to meeting demand.
“With Intel’s comprehensive open AI solutions, automakers can unlock the full potential of AI to create compelling, differentiated experiences that set their vehicles apart in the market,” Weast concludes. By shaping the future of in-vehicle experiences and delivering unprecedented levels of productivity, customisation and value to drivers and passengers alike, the importance of dGPUs to the future of automotive could be inestimable.
