Arm explains its mobile-first AI-reconstruction technology, which takes a different tack from the "black box" approach of DLSS 5

Arm, the British semiconductor and software design company, recently unveiled its latest technology at the Arm Everywhere China event in Shanghai. The focus of this presentation was on the company's AI-native computing approach, which aims to provide a full ecosystem for AI-powered devices.

Central to Arm's offering is the new AI-native compute platform, CSS for Mobile 2, alongside the Arm Mali G2-Ultra NX GPU. The company claims that this GPU delivers superior graphics performance at a lower cost in terms of heat and power consumption compared to Nvidia's DLSS 5 technology. While Arm did not disclose which specific devices will utilize the Mali G2-Ultra NX GPU, it has previously incorporated the Mali G1-Ultra chip into devices from Xiaomi, Vivo, and OnePlus.

Deyan Lazarov, part of Arm's Edge AI business unit, explained that the company has been working on the Mali G2-Ultra NX for two and a half years. He noted a growing adoption of ray tracing in mobile titles, which the new GPU aims to support. The Mali G2-Ultra NX enables mobile developers to take advantage of Unreal Engine technologies like MegaLights, as demonstrated by Arm's collaboration with Sumo Digital to create the game Neural Dawn.

Lazarov emphasized that Arm's technology is distinct from Nvidia's DLSS 5, as the company's approach does not involve generative networks that can deviate from the developers' original intentions. Instead, Arm's networks are convolutional, meaning they do not create or imagine anything but rather know what things should look like. These networks are trained on real content and are publicly available on GitHub and Hugging Face, allowing developers to customize the AI models for their specific needs.