Qualcomm unveils next-generation Adreno GPU with integrated AI

Qualcomm has unveiled technical details for its next-generation Adreno GPU, designed for upcoming flagship Snapdragon mobile platforms. The new architecture features an increased clock speed of 1.45 GHz per slice, a 250 MHz rise over the previous generation.

A key addition is the 18MB Adreno High Performance Memory (HPM), an on-chip high-speed storage space intended to reduce frequent system memory access. This design is expected to improve power efficiency by approximately 12% by minimizing data latency and power consumption during rendering.

The architecture introduces Qualcomm Matrix Cores and Adreno Neural Fusion, which integrate AI computations directly into the graphics pipeline. This allows for tasks such as AI-driven super-resolution, frame generation, and neural rendering to be processed within the GPU rather than requiring data migration between different processing units. Qualcomm reports that this new architecture can provide up to a 40% increase in performance and a 40% reduction in power consumption compared to the previous generation. The technology is already supported by major game engines, including Unity and Unreal Engine, with commercial applications expected later this year.