Qualcomm to integrate AI directly into next-gen Adreno GPUs

Qualcomm is introducing significant architectural upgrades to its next-generation flagship mobile platforms, specifically focusing on integrating AI directly into the GPU pipeline via Adreno Neural Fusion.

Unlike previous methods where AI upscaling was a post-processing step that often resulted in visual artifacts during movement, the new Adreno Neural Fusion technology integrates neural network processing, AI super-resolution, and frame generation into a single pipeline. This allows AI to participate directly in the rendering process, resulting in fewer artifacts, more stable frame rates, and clearer details in high-demand mobile games.

To support this, Qualcomm is implementing Adreno Matrix Cores, which enable the GPU to run AI models locally without the latency and power consumption associated with transferring data to the NPU. The architecture also utilizes an optimized 18MB High-Performance Memory (HPM) system, acting as a dedicated high-speed cache to reduce data retrieval distances, thereby lowering heat and power consumption while improving loading speeds and frame stability.