[Maui, United States = DigitalToday reporter Suk Dae-geon] AI computing for mobile game graphics processing is being combined directly inside the GPU. Qualcomm uses dedicated AI cores to raise resolution and fill in intermediate frames so it can deliver higher image quality with less rendering work. It also unveiled technologies to save power through on-chip memory and package thermal design. Some assess that the benchmark for game performance is shifting from peak instantaneous frame rates to how long that performance can be sustained.

Qualcomm on Sept. 22 local time unveiled the Adreno Matrix Core and a rendering technology using it, called Adreno Neural Fusion, at Snapdragon Summit 2026. The structure processes AI computations within the graphics pipeline and changes the package thermal design. Reka Motiwala (레카 모티왈라), senior director for Qualcomm mobile handsets, said, "Power efficiency of the new Adreno GPU improved by 40 percent."

Mobile games have so far struggled to simultaneously deliver console-grade graphics, stable frame rates and battery life that lasts throughout a gaming session. That is because games are the most demanding workload on mobile platforms. To address this, Motiwala said, "We newly added the Adreno Matrix Core, a dedicated AI core, to the Adreno GPU."

The Matrix Core runs AI models directly within the graphics pipeline. It does so without moving data outside the graphics pipeline to handle neural workloads. As a result, Matrix acceleration has expanded beyond the NPU and CPU to the GPU. Motiwala said, "Performance of the new GPU increased by 44 percent," adding, "The next era of gaming will be defined by smarter rendering, more efficient performance and rich experiences enabled by AI."

Qualcomm named the technology that uses the Matrix Core Adreno Neural Fusion. It combines AI super resolution, frame generation and neural processing into a single graphics pipeline. Super resolution uses AI to boost a screen rendered at a low resolution to a higher resolution, while frame generation creates intermediate frames with AI between frames that were actually rendered.

Qualcomm said this reduces rendering workload and power consumption, while also reducing screen flicker and ghosting that had appeared in existing mobile upscaling. Qualcomm also stressed that Neural Fusion is a technology it designed in-house from the start.

Design around the GPU also shifted toward reducing power use and heat. Motiwala said, "We improved Adreno High Performance Memory (HPM) to a second generation to reduce DDR memory access." HPM is a graphics-dedicated workspace that keeps intermediate data such as render targets and depth and color buffers on the chip.

A new thermal design called offset PoP (package on package) was also applied. It changes the placement of memory and silicon within the package to create a path to draw heat out of the processor, and Qualcomm said it can sustain peak performance more than twice as long in high-load tasks. Some assess that the GPU core, memory, package and rendering technology have all been aligned toward reducing power use and heat.

Adreno Neural Fusion is supported in more than 20 game titles. Qualcomm said it is the first mobile AI super resolution technology supported by major game engines. It is currently integrated into Unity and Unreal Engine, and Qualcomm said game studios can use the image quality and performance improvements without developing separate functions themselves or waiting for game engine roadmaps.

Chris Patrick (크리스 패트릭), senior vice president at Qualcomm, said, "The key to mobile gaming is not peak momentary performance, but the ability to sustain high frame rates and graphics quality at low power for long periods," adding, "Delivering console-grade lighting and graphics on mobile without compromise will become the new reality of mobile games."