Imagination Technologies Redefines Edge AI and Graphics System Design
Imagination Technologies is changing the game with the launch of Imagination E-Series GPU IP, redefining edge AI and graphics system design. The E-Series GPU leverages a highly efficient parallel processing architecture to deliver exceptional graphics performance, scaling from 2 to 200 TOPS INT8/FP8 for AI workloads. This versatile and programmable solution is ideal for future edge applications such as graphics, desktop applications, natural language processing on smartphones, industrial computer vision, and vehicle autonomy.
Key Technologies
Two key technologies underpin the potential of the E-Series GPU:
- Neural Cores: Scaling up to 200 TOPS (INT8/FP8), these cores provide significant acceleration for AI and compute workloads.
- Burst Processors: A highly innovative solution that improves average power efficiency by 35% for edge applications.
High Performance Acceleration for Low Power AI
The E-Series GPU combines advanced graphics capabilities with deeply integrated acceleration for power-efficient, low precision AI operations. The E-Series Neural Cores scale up to 200 TOPS INT8, offering up to 400% more AI performance than the previous D-Series. These Neural Cores support a wide range of AI number formats, enabling developers to design networks that meet various performance, accuracy, and power demands.
Programmable AI for Future-Proof System Design
The E-Series GPU is a programmable processor that future-proofs devices against evolving AI, compute, and graphics workloads. The Neural Cores align with the broader GPU and heterogeneous computing software ecosystem, allowing developers to easily move their workloads onto the Neural Cores using popular APIs like OpenCL.
Efficient Processing for Sustained Edge AI and Graphics Performance
The E-Series GPU architecture enhances power efficiency by 35% for AI workloads, games, and user interfaces. This improvement is achieved by reducing pipeline depth and minimizing data movement within the GPU.
The GPU That Does More
The E-Series GPU enhances multitasking capabilities by doubling the number of hardware-backed virtual machines supported by Imagination GPUs to sixteen, with sophisticated QoS support. Multicore variants of E-Series GPUs can handle multiple graphics and AI workloads simultaneously, providing a versatile solution for system designers.
The first E-Series GPU IP will be available in Autumn 2025 and has already been licensed. Automotive, consumer, desktop, and mobile variations are currently in development.
