At the 2025 Siggraph Conference, ARM introduced Super Super Sampling (NSS) – its first extension technology based on artificial intelligence, developed for mobile graphics processors.
Mobile chip manufacturer claims that nerve sampling is the first technology in the industry to add a specialized neurological accelerator to the integrated graphics processor of the chipsets. Another bold statement is the ability to provide PC graphics on mobile devices.
According to ARM, nerve supermodel can provide double resolution during the processing time of 4 ms. NSS, for example, divide the image ratio with a resolution of 540p to 1080p with almost indistinguishable quality with the original, maintain details, reality of light and the clarity of motion. Developers can use released computing resources to increase the frequency of personnel, improve visual components or reduce energy consumption.
In order to simplify the implementation of ARM technology, I released the first instruments in the world for neurological graphics, including:
- The plugin for the engine is not real;
- support for the emulation of Vulkan;
- Advanced tools for performance analysis;
- The open models are completely placed on GitHub and hugging the face;
- ARM ML extension for vulkan with new graph pipes, optimized for neurological calculations.
Epic Games, NetEase, Sumo Digital, Tencent Games, long -term games and Traverse research, confirmed Super Super Supling. Currently, this technology is used in Fortnite and Infinity Nikki, improving image sharpness, graphic stability, ball processing, as well as details of small objects and particle effects.
RM plans to expand the proposed solution and exist in 2026, increasing the speed of the nerve frame will be available to double the frequency of personnel without additional loading on the graphics processor, as well as the nerve supermodel and noise reduction, allowing real -time monitoring on mobile devices with smaller rays. All these opportunities will be available to developers before new hardware solutions.
