ARM’s new AI-powered image upscaling technology for mobile devices, Neural Super Sampling (NSS), offers an alternative to desktop solutions like DLSS and FSR. Building on the company’s Accuracy Super Resolution (ASR) technology, NSS is introduced as a second-generation solution aimed at improving mobile graphics performance.
ARM Launches New NSS Technology
NSS uses AI to upscale low-resolution images to higher resolutions, completing the process in a remarkably short time. The new technology can upscale images rendered at 540p to 1080p in just 4 milliseconds. According to ARM, this upscaling process is executed directly through its AI cores.
This distinguishes NSS from traditional upscaling techniques. The NSS system reduces the workload on the GPU by up to 50% and will be usable in conjunction with NPU (Neural Processing Unit)-class accelerators, which will be integrated into ARM GPUs starting in 2026.
NSS offers significant improvements, particularly when compared to ASR technology. Compared to ASR, used in games like Fortnite and Infinity Nikki, NSS delivers clearer and more consistent results in sharpness, stability, and shadow detail. It also offers greater accuracy in rendering visual details such as fine geometry and particle effects.
NSS utilizes techniques such as INT8 data compression, learned accumulation, and targeted pixel filtering. This allows it to operate more efficiently without the need for traditional upscaling methods like reactive masking.
Developer tools offered by NSS have also been announced. The Neural Graphics Development Kit includes the Unreal Engine plugin, Vulkan emulation, performance profiling tools, and machine learning extensions to ARM’s Vulkan API. Major gaming and software companies such as Epic Games, Tencent, NetEase, and Sumo Digital have already begun working on NSS.
ARM also stated that NSS will not be limited to resolution upscaling in the future. The company plans to incorporate AI-powered technologies into its system in areas such as frame generation in the future.
