Today, Imagination has introduced the DXS GPU, their latest innovation in automotive GPU IP designed for in-vehicle intelligence and interaction. The DXS GPU offers a scalable and adaptable solution for managing graphics and compute tasks across cockpit displays, infotainment systems, and advanced driver assistance systems (ADAS). This cutting-edge GPU integrates new innovations in distributed safety, removing the traditional overhead associated with achieving ASIL-B functional safety on Imagination processors. The DXS GPU has already been licensed for the automotive sector.
The DXS GPU is capable of supporting the graphics and compute needs of an entire vehicle lineup, with single-core configurations starting at 0.25 TFLOPS and scaling up to 1.5 TFLOPS—an impressive 50% increase in peak performance over Imagination’s previous generation automotive GPU IP. Featuring Imagination’s unique multi-core technology, the DXS supports configurations with two, three, and four cores. Its low bandwidth bus between cores and support for isolation make it ideal for use in chiplets. The GPU incorporates advancements from the PowerVR D-Series architecture, such as Pipelined Data Masters and 2D Dual-Rate Texturing, which enhance performance efficiency by 20% compared to previous generations and outperform competitive cores in real-world benchmarks.
Anshel Sag, Principal Analyst at Moor Insights & Strategy, comments, “Imagination is a leading provider of GPU IP for automotive applications. The growing demand for pixel processing in vehicles, driven by digital cockpits, expansive dashboard displays, and rear entertainment screens, underscores the need for high-performance graphics solutions. Imagination’s blend of exceptional 3D graphics performance, flexible multitasking capabilities, and robust safety features has significantly increased its popularity.”
The DXS GPU is optimized for next-generation vehicle intelligence tasks, including computer vision and data processing. With the addition of an FP16 pipeline and new compute libraries and AI toolkits, the DXS GPU delivers up to ten times higher performance for compute workloads compared to previous models.
James Chapman, Chief Product Officer at Imagination, explains, “Today’s automotive GPUs are not just about graphics. As vehicles become more software-defined, designers need flexible and programmable processors that can handle current AI workloads and adapt to future demands. Our automotive GPUs are increasingly valued for their compute capabilities as much as for their rendering performance.”
Distributed Safety Mechanisms
Imagination’s new “Distributed Safety Mechanisms” technology allows the DXS GPU to achieve ASIL-B functional safety with minimal impact on performance, power, or area (PPA). Unlike traditional methods, such as dual-core lockstep (which doubles silicon area) or workload repetition (which halves performance), this innovative approach leverages processor parallelism and idle thread periods to conduct safety tests. This results in nearly zero performance impact and a 10% area cost. Detailed information about this technology can be found in Imagination’s whitepaper, “Innovations in Distributed Functional Safety.”
Enhanced Compute Libraries and AI Toolkits
Accompanying the hardware are new compute libraries (imgBLAS, imgNN, imgFFT) designed to help developers achieve up to 80% GPU utilization. These libraries accelerate common compute tasks like computer vision and ADAS data processing. They work in tandem with new reference toolkits (oneAPI and TVM), creating a robust software stack based on open standards for easy porting and performance optimization.
Rod Burns, Chair of the UXL Foundation Steering Committee, says, “Software portability and the ability to leverage heterogeneous hardware are crucial for the automotive industry’s future. The UXL Foundation aims to simplify this through open-source development of the oneAPI standard. Imagination’s new compute solutions further support this vision.”
Zhang Xianyi, CEO of PerfXLab Technologies, adds, “We’re using Imagination’s compute software solutions to enhance AI performance, including our LLM inference engine, PerfXLM. We’ve achieved performance gains of up to 100% compared to CPUs with minimal porting time.”
Software and Support
Imagination’s automotive GPUs support OpenGL® ES, Vulkan®, OpenGL®, and OpenCL™, and are compatible with popular automotive operating systems such as QNX, Green Hills Software’s INTEGRITY RTOS, Linux, and Android. Hardware-based virtualization allows the GPUs to run up to eight operating systems simultaneously with full memory isolation for secure multitasking. Imagination collaborates with CoreAVI, a leader in safety-critical graphics and compute software, to develop a safety-critical driver supporting OpenGL SC and Vulkan SC.
Dan Joncas, Chief Marketing and Sales Officer & Deputy CEO of CoreAVI, states, “As vehicles become increasingly autonomous, the ability to manage and handle faults effectively is critical. Our collaboration with Imagination provides a comprehensive safety solution that addresses both hardware and software needs.”
For more details on the Imagination DXS GPU, visit the Imagination website.
