Chip designer Arm has unveiled an AI-based technology to provide desktop-quality graphics on mobile phones. Announced at SIGGRAPH 2025, the method involves creating sharper visuals at lower power by computing fewer pixels per image—and using AI to add the finer details.
The company touts the advance—scheduled for 2026 release—as “Arm neural technology,” a set of features for future Mali GPUs that adds dedicated on-chip neural accelerators to speed AI-assisted rendering. Arm’s Neural Super Sampling (NSS), an AI upscaler tuned for phones, can transform a lowly 540p image into a sharper 1080p graphic in about 4 milliseconds per frame. And it’s energy efficient in the manner that Arm has always been known for: the neural accelerator lowers the burden on the processor up to 50% when compared with drawing frames at higher resolution.
The NSS technology is a step forward from Arm’s Accuracy Super Resolution (ASR), unveiled in March, which is also an upscaling technology developed for mobile devices. But ASR uses temporal accumulation to improve image quality, meaning it leverages data points from previous video frames to present higher-res images. NSS, in contrast, uses an AI-boosted on-chip neural accelerator to enhance image quality. In any case, that Arm would announce another image upscaler just five months after ASR is a testament to today’s hyper-fast growth in AI technology.
While impressive, Arm’s innovation builds on earlier desktop technology, which it has re-architected for the constraints of mobile. Similar technology is provided for the desktop by AMD’s FidelityFX Super Resolution (FSR), a spatial upscaling technology that boosts gaming performance by rendering at a lower resolution and enhancing to a higher resolution. Also similar is NVIDIA’s Deep Learning Super Sampling (DLSS), an AI-based technology that boosts gaming graphics. AMD’s image upscaler is open source, while NVIDIA’s is limited to the company’s own graphics cards.
Arm’s model will be fully open, with developer tools available on HuggingFace and GitHub. And clearly the company is courting developers with this release: it has debuted what it calls the industry’s first open neural-graphics development kit. Developers get an Unreal Engine plugin for drop-in NSS experiments, PC-based emulation of Arm’s Vulkan ML extensions, and updated profiling tools.
Arm is stressing that this release isn’t just for gamers, noting that use cases for its new on-device AI technology include productivity tools and intelligent cameras. At the same time, the company lists a long list of gamer partners, including Epic Games, NetEase Games, Tencent Games, Sumo Digital, all of which have signaled support, giving gaming studios a lengthy runway before hardware arrives.
“Sumo Digital believes that neural graphics and AI-based upscaling will revolutionize mobile gaming—delivering console-quality visuals and deeply immersive experiences without draining battery life,” said Scott Kirkland, Group Technology Director, Sumo Digital.
Arm also looked forward at SIGGRAPH to preview related aspects of its roadmap. In 2026, the company will offer Neural Frame Rate Upscaling (NFRU), which uses AI to create twice the frame rate without doubling the processor load. To support this, the processor will add hardware for faster, cheaper motion-vector generation. Also slated for 2026, Arm’s Neural Super Sampling and Denoising (NSSD) drives real-time path tracing on mobile. The system can deploy only a handful of rays per pixel and let AI reconstruct the missing lighting detail using information from neighboring pixels and prior frames.
