Arm Launches Neural Super Sampling for Mobile Gaming and Graphics

Arm has released Neural Super Sampling (NSS), an AI-powered upscaling solution designed for graphics and gaming developers building for future mobile devices.

How It Works#

NSS is a parameter prediction model optimized for real-time execution on Neural Accelerators (NX) in mobile GPUs. It reconstructs high-quality output frames from low-resolution temporal inputs, enabling high-resolution rendering at significantly lower computational cost. In Arm's Enchanted Castle demo, the approach reduced GPU workload by 50 percent, rendering at 540p and upscaling to 1080p in just 4ms.

The technology is particularly suited for mobile gaming, extended reality (XR), and other power-constrained graphics applications—areas where battery life and thermal management are critical concerns.

Training and Tools#

Arm has published the Neural Graphics Dataset, a collection of reference images, sequences, motion data, depth information, and other assets needed to train and validate neural super sampling algorithms. The dataset currently contains a limited set for demonstration purposes. Arm plans to expand this through the Neural Graphics Model Gym, which will provide tools for developers to capture and convert content for model training and retraining.

Getting Started#

The technology has been integrated into Unreal Engine through two plugins: the NSS Plugin for Unreal Engine and the Unreal NNE Plugin for ML extensions for Vulkan. Developers can follow quickstart guides for both Vulkan and Unreal implementation.

For technical details, see Arm's technical blog on how Neural Super Sampling works and the white paper on the Unreal plugin and ML extensions.