Arm Unveils Mobile-First AI Graphics Technology as Alternative to DLSS 5

Arm has unveiled its mobile-first approach to AI-powered graphics reconstruction, positioning it as a more transparent and developer-friendly alternative to Nvidia's DLSS 5 technology. The British semiconductor company announced the Mali G2-Ultra NX GPU and a suite of AI-native computing tools at its Arm Everywhere China event in Shanghai, designed to help mobile developers achieve console-quality graphics while managing power and thermal constraints.

The core difference between Arm's approach and DLSS 5 lies in architecture and transparency. Arm's technology uses non-generative convolutional neural networks that are trained on real content and published openly on GitHub and Hugging Face. "Our networks are not generative, so they're not from the same nature as the latest DLSS model," said Deyan Lazarov, who leads Arm's GPU product team in the Edge AI business unit. "Our networks are convolution type networks, which means that they don't imagine anything, they don't create anything, they just kind of know what things need to look like." In contrast, Lazarov characterized DLSS 5 as "proprietary, it's closed, it's a black box, you don't know what's happening, you don't know what it's trained on."

Arm's solution integrates three key technologies: Neural Super Sampling (NSS), Neural Super Sampling and Denoising (NSSD), and Neural Frame Rate Upscaling (NFRU). These enable upscaling of images, additional ray casting, and AI-reconstructed frames. In one example, Lazarov described rendering where the GPU renders only a quarter of an image with NSS reconstructing the remaining three quarters on neural accelerators, and in subsequent frames, all pixels are AI-reconstructed with zero traditional GPU rendering.

The Mali G2-Ultra NX GPU was developed over two and a half years and targets mobile titles seeking desktop-quality graphics. It supports technologies like Unreal Engine's MegaLights and enables ray tracing on mobile devices. Lazarov noted growing adoption of ray tracing in mobile titles and cited the benchmark title Neural Dawn, developed with Sumo Digital, which showcases the GPU's capabilities. The chip achieved 60 fps in Neural Dawn compared to only 40 fps on the previous Mali G1-Ultra GPU at maximum settings, or 15 fps under normal conditions.

Games set to leverage the Mali G2-Ultra NX include Arena Breakout Infinite from Tencent, Where Winds Meet from NetEase, and Infinity Nikki from Infold Games. Earlier Mali G1-Ultra chips appear in devices from Xiaomi, Vivo, and OnePlus, though Arm has not confirmed which devices will use the new GPU.

A key advantage of Arm's open approach is developer flexibility. Developers can retrain Arm's networks on their own game data, allowing the AI to preserve unique art styles and visual identity. This contrasts with DLSS 5's closed nature, where developers must accept results as-is. Lazarov emphasized that this developer control eliminates risks of IP loss while ensuring upscaling and frame generation match a game's intended appearance.

Power efficiency remains central to Arm's design philosophy. Mobile devices face stricter thermal and power budgets than consoles or PCs and operate on batteries, requiring different optimization strategies. The Mali G2-Ultra NX achieves a 70% reduction in DRAM traffic, one of the largest power consumers on mobile hardware. By reducing GPU workload and power consumption, the technology gives developers the flexibility to reinvest efficiency gains into visual fidelity, performance, battery life, or a combination of these.