Imagination Unveils D Series GPU, Teases E Series

Imagination Technologies has introduced the final member of its D series graphic processing unit IP, marking a significant advancement in bandwidth and performance tailored for AI applications as the company sets its sights on the upcoming E Series.

According to Kristof Beets, Vice President of Product Management at Imagination, "This is the last entry in the D family. It’s always a little artificial, what’s an iteration or a new architecture. The next step is bigger and more fundamental in performance increase."

The newly launched DXTP core is specifically optimized for data flow in AI applications, featuring an additional port and a revised ratio of resources within the core structure. This enhancement introduces a new data path and fixed function block, emphasizing the synergy between graphics and compute.

Beets highlighted, "There’s enormous amounts of synergy between graphics and compute. We are making it easier to implement, and this is why we keep saying graphics becomes a form of compute."

The company is now positioning graphics as a primary use case, with compute taking precedence. Developers in the graphics market are increasingly bypassing fixed function units in favor of programmable compute, enabling them to utilize their own techniques to enhance speed and efficiency.

D-Series articles:

• Ray tracing GPU aims for mainstream mobile gaming
• Safety architecture boosts automotive GPU for chiplets
• GPU IP extends battery life for mobiles

Imagination Technologies has evolved its architecture to cater to the changing demands of the market, transitioning from graphics-centric to AI-focused designs. The shift towards AI frameworks, which are highly network intensive, necessitates a greater emphasis on data flow to ensure optimal performance and efficiency.

Beets explained, "For every compute, there is set up work to prepare the registers, and that is 16x faster by tuning the fixed function logic from one setup per clock to go massively parallel. Some of the processing jobs were relatively small, and in those cases, we were set up limited, and that makes a big difference."

The company has also embraced in-memory compute, leveraging its GPUs' storage capabilities to enable tile-based computing and efficient data management on-chip. With half a megabyte per unit and a significant focus on local register space, Imagination's GPU IP offers flexibility and performance optimization.

While Imagination's GPU IP is commonly associated with mobile applications, it has found applications beyond the mobile realm. Self-driving startup Zelos, for instance, is utilizing the DGXT core as a compute engine for its operations.

Beets elaborated, "This is a massive growing area for us. We developed the computational libraries for them focused on neural networks, FFTs, and linear algebra for ADAS. We pick relevant networks, but there are odd ones they can look at the optimized libraries and map it to their flow."

Looking ahead, Imagination is focused on compute and AI, with plans to develop software and tools for the forthcoming E-series. Beets emphasized the company's commitment to aligning its IP with evolving customer needs and industry trends, particularly in the realm of AI and computational technologies.