Imagination Technologies has introduced its latest E-Series graphics processing units (GPUs) designed for graphics and AI processing at the edge.
Imagination’s new E-Series GPUs aim to redefine edge AI and graphics system design by utilizing a highly efficient parallel processing architecture. These GPUs offer exceptional graphics performance and can scale from two to 200 TOPS INT8/FP8 for AI workloads.
According to Kristof Beets, vice president of product management at Imagination, the E-Series GPUs feature a unique blend of AI processing components within the chip. This innovative design allows the GPUs to efficiently handle various processing tasks, making them stand out in the market and enabling Imagination to target specialized markets such as automotive.
The E-Series GPU family provides a versatile and programmable solution for future edge applications, including graphics, desktop applications, natural language processing on smartphones, industrial computer vision, and vehicle autonomy.
Two key technologies underpin the E-Series GPUs’ potential to transform edge system design:
Neural Cores: These cores can scale up to 200 TOPS (INT8/FP8) and deliver significant acceleration for AI and compute workloads.
Burst Processors: This innovative solution offers a 35% improvement in average power efficiency for edge applications.
The E-Series GPUs continue to offer advanced graphics capabilities, including support for ray tracing, while integrating acceleration for power-efficient, low precision AI operations into every GPU core. This integration results in compute-dense Neural Cores that can scale up to 200 TOPS INT8, delivering up to 400% more AI performance than the previous D-Series.
The Neural Cores support a wide range of AI number formats, allowing developers to design networks that meet various performance, accuracy, and power demands. Additionally, the GPUs feature an AI-friendly memory architecture that prioritizes local memory for compute, reducing the power and performance costs of accessing external memory.
These GPUs are programmable processors that can future-proof devices against the evolving landscape of AI, compute, and graphics workloads. The E-Series Neural Cores align with the broader GPU and heterogeneous computing software ecosystem, enabling developers to unlock their capabilities using popular APIs like OpenCL.
Imagination’s E-Series GPUs also boast enhanced multitasking capabilities, with support for multiple graphics and AI workloads simultaneously. The GPUs feature hardware-backed virtual machines and sophisticated QoS support to ensure smooth operation across various tasks.
The first E-Series GPU IP is set to be available in Autumn 2025, with automotive, consumer, desktop, and mobile variations currently in development. Imagination Technologies aims to position the E-Series GPUs at the forefront of both graphics and edge AI systems, offering a versatile and cost-effective solution for system designers.
