Arm is bringing Android its own version of DLSS – game with Unreal MegaLights will show it off

Source Domain: www.techspot.com

Something to look forward to: Advanced rendering features such as ray tracing, Unreal MegaLights, and neural supersampling are primarily associated with Nvidia’s PC graphics cards. But Arm recently revealed that these technologies are coming to new Android mobile devices later this year. A short game from Sumo Digital demonstrates how the latest phones and tablets can handle ray tracing at playable framerates with AI-based upscaling.

Sumo Digital’s upcoming Neural Dawn aims to demonstrate that mobile graphics chips can achieve high-quality upscaling, denoising, frame generation, and ray tracing with large numbers of dynamic lights. The tech demo is set to launch in the fourth quarter of this year exclusively for devices equipped with an upcoming successor to Arm’s Mali G1 GPU.

Neural Dawn will be the first mobile game to support Unreal Engine 5’s MegaLights. Introduced in 2024, MegaLights enables games to render scenes with a massive number of dynamic light sources at a lower performance cost compared to earlier methods. It achieves this by estimating the directions of most lights based on a heuristically selected set of primary lights.

Unreal MegaLights has thus far been exclusively associated with high-end console and PC games, such as Microsoft’s upcoming Gears of War: E-Day. Furthermore, adding ray tracing to the pipeline increases the number of visible lights, thus raising the cost of MegaLights, but Arm’s Mali G1 is up to the task.

Arm introduced Mali G1 as part of its Lumex platform late last year. Mobile processors featuring G1 have seen significantly improved ray tracing performance, but the next generation introduces support for neural upscaling features that closely resemble Nvidia DLSS. Mali G1’s Neural Super Sampling and Denoising (NSSD) and Neural Frame Rate Upscaling (NFRU) sound a lot like Nvidia’s DLSS, Ray Reconstruction, and Frame Generation technologies, and appear to target similar…

Source