Imagination's E-Series GPU IP, announced in 2025, is positioned as a step toward converged acceleration for edge systems. The architecture unifies graphics, compute, and AI on a single flexible design backed by one programmable software stack, aiming to simplify system design while supporting evolving AI models.
Performance-wise, the E-Series offers up to 32 TOPS Int8 per core at 1GHz, which the company says is four times the raw compute performance of the previous D-Series. It also supports concurrent execution of graphics and AI workloads, and can act as a flexible co-processor to offload dedicated AI processors during peak demand.
According to the source, the E-Series provides advanced graphics acceleration for games and user interfaces, a general-purpose accelerator for future AI operators, and edge-level power efficiency. First silicon is expected by the end of 2026, with real-world performance data covering gaming, neural rendering, and language models included in the announcement.