Imagination Technologies Launches E-Series GPUs for Advanced Edge AI and Graphics

Imagination Technologies has unveiled its groundbreaking E-Series graphics processing units (GPUs), designed to redefine edge AI and graphics system design. These GPUs leverage a highly efficient parallel processing architecture to deliver exceptional graphics performance while scaling AI workloads from two to 200 TOPS (tera operations per second) in INT8 and FP8 precision formats.

A key differentiator of the E-Series is its clever integration of AI processing components within the GPU chip, enabling efficient handling of diverse processing tasks. This unique design positions Imagination to target specialty markets such as automotive, industrial computer vision, and mobile devices, where both graphics and AI workloads are critical.

Innovative Technologies Powering the E-Series

Two core innovations underpin the E-Series’ transformative potential:

• Neural Cores: Scalable up to 200 TOPS in INT8/FP8 formats, these cores accelerate AI and compute workloads with high efficiency.
• Burst Processors: An innovative architecture that improves average power efficiency by 35%, optimizing performance for edge AI, gaming, and user interfaces.

These advancements address the critical challenge of balancing performance, efficiency, and flexibility in on-device AI, making the E-Series a compelling choice for edge AI system designers.

High-Performance AI Acceleration with Power Efficiency

The E-Series GPUs maintain the advanced graphics capabilities of previous Imagination GPUs, including ray tracing support, while integrating deeply embedded acceleration for low-precision AI operations. This integration results in compute-dense Neural Cores capable of delivering up to 400% more AI performance compared to the prior D-Series generation.

The Neural Cores support multiple AI numeric formats, allowing developers to tailor networks for diverse performance, accuracy, and power requirements. Additionally, an AI-friendly memory architecture prioritizes local memory access, significantly reducing power consumption and latency by minimizing external memory usage.

Programmable AI for Future-Proof Edge Systems

Imagination emphasizes that GPUs are inherently programmable processors, offering future-proofing against evolving AI, compute, and graphics workloads. The E-Series Neural Cores integrate AI acceleration seamlessly within the GPU architecture, compatible with popular APIs such as OpenCL and supporting open standards and tools like oneAPI, Apache TVM, and LiteRT.

This programmability enables developers to easily migrate AI workloads onto the Neural Cores, maximizing efficiency through Imagination’s compute libraries and optimized graph compiler. Such integration is critical for unlocking the full potential of on-device intelligence across diverse applications.

Sustained Performance and Multitasking for Complex Edge Devices

The E-Series builds on Imagination’s renowned PowerVR GPU architecture, known for energy efficiency in power-constrained devices over two decades. The Burst Processors technology further enhances power efficiency by 35% by reducing pipeline depth and minimizing data movement within the GPU.

Modern edge devices demand simultaneous support for multiple graphics and AI workloads. The E-Series addresses this by doubling hardware-backed, zero-overhead virtual machines to sixteen, with sophisticated quality of service (QoS) support. Multicore variants offer additional cores for enhanced performance or flexibility, capable of handling multiple graphics and AI workloads concurrently.

This versatility eliminates the need for separate vector-based or fixed-function AI solutions, reducing system design complexity and costs while future-proofing edge AI and graphics systems.

The first E-Series GPU intellectual property (IP) will be available in Autumn 2025, with licenses already secured. Variants targeting automotive, consumer, desktop, and mobile markets are under development, signaling broad applicability across edge AI and graphics domains.