ARM recently announced its new Neural Super Sampling (NSS), it’s latest AI-powered upscaling algorithm. The algorithm was announced at SIGGRAPH, and is designed to work with low-power ARM devices.
The mobile chipmaker claims Neural Super Sampling to be an industry first, and adds dedicated neural accelerators to the integrated GPUs of its chipsets. Another bold claim is that it is capable of delivering PC quality graphics to mobile devices.
Performance-wise, the chipmaker says that Neural Super Sampling has the potential to deliver up to double the resolution and at 4ms per frame.
“NSS delivers the potential for upscaling from 540p resolution to 1080p at a cost of 4ms per frame, while delivering near-native quality. Developers can save up to 50% of the GPU workload compared with rendering the full frame using traditional methods, and either bank that saving to reduce the overall power consumption of their game, spend it on delivering a higher frame rate or increasing the quality of the visuals. With NSS, developers can use AI to preserve surface detail, lighting, and motion clarity, giving them the flexibility to balance visual fidelity with energy efficiency depending on their game’s needs.”
The kit for its Neural Super Sampling kit to developers. The kit includes an Unreal Engine Plugin, PC-based Vulkan emulation, updated profiling tools, fully open models available via GitHub and Hugging Face, for which we’ve provided a link in the source; and ARM ML extension for Vulkan.
(Source: ARM, Hugging Face)
