Imagination unveils E-Series GPUs for graphics and AI at the edge

Imagination technologies E-Series opens graphics and AI processing on the edge of graphics and AI edge.

Imaginative Products, when using the INT8 / FP8 from two to 200 to 200 for AI to workload, redeem the design of the Edge AI and graphical system to ensure a highly effective parallel processing architecture, redefine Exchine AI and graphical system design.

In a briefing, the Vice President of Product Management, Christopel Beet, said that GPU is a smart mix of AI processing components in the chip, which is the most effective management. This helps to distinguish from other GPUs in the market and allows you to target the target markets as a car.

According to him, the GPU family offers a versatile and programmable solution for future extraneous applications, including graphics, desktop applications, smartphones, industrial computer vision and natural language processing.

E-Series ‘e-Series’ Light System Design Transformation potential:

• Neuron Cores: Up to 200 above (INT8 / FP8), these cores are significantly accelerating significant acceleration and calculation for AI.
• Burst processors: A highly innovative solution that conveys 35% of the average power efficiency for Edge applications.

“The AI ​​is developing at the moment, but Edge AI system designers are still facing difficulties with performance and efficiency effectiveness,” he said. “Imagination has used a long-term experience that develops the effective GPU, and combines a push of a push on the GPU’s wheel to create a compelling solution for the device’s system designers in the device mode.”

High performance acceleration for low power AI

The e-Series continues to offer advanced graphics capabilities of the previous generations of GPU, including Ray Tracking Support. To do this, each GPU core adds deep integrated acceleration for efficient, low-precision AI transactions with power. This creates computational-intense e-series si-series cores by which the previous person, which lowered up to 200 and up to 200 peaks and up to 400% previous person’s AI performance Derzord.

Neuron nuclei supports widespread popular AI number formats, developers allow you to design networks that meet extensive performance, accuracy and power requirements. One of the many performance efficiency measures is a Memory architecture of AI-friendly memory, which is prioritizing the power of foreign memory and reduce performance costs.

Programmed AI for continuous system design in the future

GPUs are processors of programmeal processors, AI, computing and graphical workloads Programming processors in future continuous devices. E-Series Neural Cores, compatible with a more extensive GPU and heterogeneous computing program by combining AI acceleration in GPU with an ecosystem.

Their capabilities can unlock as an OpenCL, as the popular APIs and can easily transfer workloads to nerve nucleus using open standards and tools such as Developers, OneApi, Apache TVM or Liter. Imagination increases the efficiency of the computing libraries and highly optimized graphic compiler GPU.

“EDGE AI device and software integration is very important to unlock intelligence potential on the device,” he said. “E-Series allows developers to place AI algorithms on many applications and devices on the latest devices.”

Effective operation for continuous Edge AI and graphic performance

Imagination PowerVR GPU architecture is famous for energy efficiency and has been placed on electricity restricted devices for about twenty years. Eserea’s new explosion processor technology increases 35% power efficiency for AI workload, games and user interfaces. This improvement is obtained by reducing the depth of the pipeline and minimize the data movement within the GPU.

GPU that makes more

Modern devices are increasingly complex and processors are required to support multiple graphics and AI workloads at the same time. Providing a clear priority throughout the quality of service (GOS) and this workload is important for the user experience.

The e-Series increases numerous opportunities of numerous generations of previous generations, increasing the number of zero-surface virtual machines, which is supported by gentle GPUs with subtle QoS support. Numerous options of E-Series can use additional cores for additional performance or advanced flexibility in GPUs. This GPUS is able to deal with a very graphic workload, a combination of a combination of many AI workloads or simultaneously.

“E-Series will place GPU at the center of both graphics and the EDGE AI system,” said Tim Mamtora, Innovation and Engineering, in a statement in the statement. “Both graphics workload and workload, an e-series GPU, when saving the general system design costs, the addition of additional vector-based or stable functionality is a versatile solution.”

The first e-Series GPU IP is available in the fall of 2025 and has already been licensed. Cars, consumers, desktops and mobile changes are developing.