The GPU Inside 12Bn+ Devices

Imagination has spent more than thirty years optimising our GPU architecture for advanced gaming, rich 3D user interfaces and compute-centric workloads. We license the architecture as a soft GPU IP, our licensees build it into their silicon, and that silicon ships in billions of cars, phones, wearables and PCs worldwide.

WHY LEADING SoC ARCHITECTS CHOOSE IMAGINATION

Imagination GPUs go beyond graphics, offering an ideal platform for AI acceleration in power-constrained edge devices. Their high parallelism, originally engineered for complex 3D rendering, is perfectly suited for executing the massive compute workloads of modern AI models with efficiency and speed.

EFFICIENCY

Our GPU architecture is designed intelligently to deliver superior performance efficiency. Tile-based deferred rendering minimises data movement and removes redundant operations early in the pipeline.

FLEXIBILITY

When a SoC is overloaded, having the ability to offload compute-heavy tasks onto an AI-capable GPU relieves pressure. Imagination GPUs can communicate directly with other accelerators to form an efficient edge AI subsystem.

FUTURE-PROOF

AI models and the operators inside them are changing month by month – but silicon can be in the market for a decade. General-purpose accelerators, like Imagination GPUs, help a chip maintain relevance in a fast-moving market.

PROGRAMMABILITY

Software teams find it easy to get their code running on Imagination IP thanks to the familiar GPU programming model (based on industry-standard APIs like Vulkan and OpenCL) plus an extensive software ecosystem with optimised backends to popular frameworks.

FEATURE SET

SoC Architects looking to tailor their hardware to a particular market can choose from an impressive array of advanced graphics and computing technologies including powerful ray tracing, DirectX support and ASIL-B safety coverage.

CONCURRENCY

Our multi-threaded shading clusters feature an ALU architecture with high SIMD efficiency. Different jobs can be submitted to the GPU concurrently, with scheduling managed by the GPU’s firmware processor.

Explore our IP

WHAT ARE YOU INTERESTED IN?

OUR LATEST INNOVATIONS

Up to 32 TOPS INT8 AI acceleration per core, deeply integrated within the shading cluster for maximum performance efficiency with support for low precision number formats including FP8 and MXFP4.

A hyper-efficient upscaling solution leveraging the AI acceleration of the E-Series Matrix Accelerator. NSR upscales a typical 540p frame to 1080 in 2.3ms with half the memory bandwidth consumption, delivering a significant boost to frame rates.

The new, shorter pipeline introduced with E-Series delivers up to 39% power efficiency improvements by cutting the unnecessary data traffic caused by interrupted operations.

Recent news and updates

IMAGINATION ANNOUNCES E-SERIES: A NEW ERA OF ON-DEVICE AI AND GRAPHICS

Massive gains in Imagination E-Series establishes the GPU as the principal accelerator for both graphics and AI at the edge.

Read more
IMAGINATION TECHNOLOGIES BRINGS GPU EXPERTISE TO CHASSIS’ NEW CHIPLET PROJECT 

Imagination contributes  its latest functionally safe GPU IP to the CHASSIS program’s new Automotive Base Die.

Read more
IMAGINATION TECHNOLOGIES APPOINTS MARKUS MOSEN AS CHIEF EXECUTIVE OFFICER

Proven semiconductor business leader to provide solid leadership, operational excellence and outstanding experiences for Imagination’s customers.

Read more
IMAGINATION DXTP GPU REDEFINES GRAPHICS AND PERFORMANCE EFFICIENCY

Imagination DXTP redefines graphics and compute acceleration for smartphones and other power-constrained devices. Learn more below.

Read more