Arm, the British semiconductor and software design powerhouse, staged its Arm Everywhere China showcase in Shanghai to announce fresh advances in AI‑driven computing. The company’s vision is to deliver a complete ecosystem for AI‑native workloads, highlighted by the new CSS for Mobile 2 compute platform and the Mali G2‑Ultra NX GPU.
This GPU is built around neural processing, promising higher graphics performance while keeping heat and power draw to a minimum. Arm positions its on‑device AI solution as a more transparent alternative to Nvidia’s controversial DLSS 5, aiming to align more closely with developers’ original creative intent. While Arm did not name specific products that will ship with the Mali G2‑Ultra NX, its predecessor, the Mali G1‑Ultra, already powers devices from Xiaomi, Vivo and OnePlus. Deyan Lazarov, a member of Arm’s GPU product team within the Edge AI division, explained that the Mali G2‑Ultra NX has been in development for roughly two and a half years.
Its core purpose is to give mobile developers access to compute‑intensive features such as ray tracing, which has historically been slow to appear on phones. "We’re seeing more mobile titles interested in ray tracing now," Lazarov said. The new GPU enables developers to tap into Unreal Engine capabilities like MegaLights, and Arm partnered with Sumo Digital to produce the demo title Neural Dawn, which showcases the chip’s capabilities.
Lazarov described the target market as "mobile‑first games that want desktop‑level visual quality" as well as cross‑platform titles that should deliver a consistent experience whether played on a console at home or on a train abroad. Games slated to benefit from the Mali G2‑Ultra NX include Tencent’s Arena Breakout Infinite, NetEase’s Where Winds Meet, and Infinity Nikki from Infold Games. Lazarov highlighted the unique thermal and power constraints of phones compared with consoles or PCs: mobile devices cannot dissipate heat as aggressively and must run continuously on battery power.
Arm’s answer is to blend neural processing with the traditional rendering pipeline. By offloading part of the workload to dedicated neuro‑accelerators, the GPU’s power consumption drops, which in turn improves overall system efficiency. "You gain new efficiency that eases the trade‑offs developers normally face," Lazarov explained.
This efficiency lets developers decide whether to reinvest saved resources into higher visual fidelity, smoother performance, longer battery life, or a mix of all three. To foster adoption, Arm is taking an open‑software stance, offering a suite of tools and publicly available neural networks. Over the past two years the company has worked closely with the ecosystem to make AI‑enhanced graphics developer‑friendly. The key techniques include Neural Super Sampling (NSS), Neural Super Sampling and Denoising (NSSD), and Neural Frame Rate Upscaling (NFRU).
These methods can upscale images, add extra rays, and generate additional frames through AI reconstruction. Lazarov illustrated the process with two example frames. In the first, the GPU renders only a quarter of the image while the neuro‑accelerators reconstruct the remaining three‑quarters using NSS. In the second frame, the GPU does no traditional rendering at all; every pixel is produced by the neuro‑accelerators, resulting in a fully AI‑reconstructed frame.
Nvidia’s DLSS 5 also uses AI to enhance image detail, but it has drawn criticism for potentially diverging from developers’ artistic intent. Arm’s approach differs fundamentally: its networks are convolutional rather than generative, meaning they do not invent new content but rather learn how existing content should look. All of Arm’s resolution and frame‑rate networks are open‑source on GitHub and Hugging Face, allowing developers to fine‑tune them with custom data sets that reflect a game’s unique art style.
This openness eliminates worries about IP loss or black‑box behavior. Lazarov contrasted this with DLSS’s closed‑box nature, describing Nvidia’s solution as a proprietary “black box” that cannot be retrained by developers. In contrast, Arm’s networks can be adapted, giving studios full control over how upscaling and frame generation affect their visual identity.
Another major distinction is the mobile‑first design philosophy. Arm’s networks are engineered to respect mobile constraints such as power, bandwidth and memory traffic. Lazarov noted that DLSS would simply exceed a phone’s budgets, making it impractical for handheld devices. Power efficiency is a cornerstone of the Mali G2‑Ultra NX.
Arm reports up to four‑times performance and efficiency gains, plus a 70 % reduction in DRAM traffic—a major source of power draw on mobiles. Less memory access translates directly into lower energy consumption. In practical terms, the new GPU can run Neural Dawn at a smooth 60 fps, whereas the previous generation struggled to reach 40 fps under ideal conditions and dropped to 15 fps in typical use. Development tools are already available as Unreal Engine plug‑ins, with Unity support on the roadmap and a recent partnership with Unity China announced.
For studios using custom engines, Arm provides an SDK that streamlines integration without requiring engine‑specific plug‑ins. Lazarov emphasized that existing games can be upgraded quickly: integrating Arm’s AI‑native graphics into a shipped title can take as little as three months. He cited Neural Dawn’s development—a team of fewer than 20 people built the demo in under 18 months—as proof that Arm’s technology enables rapid iteration and high‑quality results. Overall, Arm’s mobile‑first AI reconstruction strategy seeks to give developers transparent, efficient tools that enhance visual quality without sacrificing power or creative control, offering a clear alternative to the more opaque DLSS 5 approach.