Arm, the British chip and software design firm, staged its Arm Everywhere China showcase in Shanghai to announce a suite of fresh AI‑computing innovations aimed squarely at mobile devices. The centerpiece of the reveal is a new AI‑native compute platform, CSS for Mobile 2, and the Mali G2‑Ultra NX graphics processor – a GPU built around neural‑accelerator cores that promises higher visual fidelity while keeping heat and power draw low.
Arm positions its on‑device AI graphics pipeline as a more transparent alternative to Nvidia’s controversial DLSS 5, aligning more closely with developers’ original artistic intent. While Arm did not name specific handsets that will ship with the Mali G2‑Ultra NX, its predecessor, the Mali G1‑Ultra, already powers phones from Xiaomi, Vivo and OnePlus, suggesting a similar market trajectory. Deyan Lazarov, a member of Arm’s GPU product team within the Edge AI division, explained that the new chip has been in development for roughly two and a half years.
Its goal is to give mobile developers access to compute‑heavy features such as real‑time ray tracing, which has historically been limited on smartphones. "We’re seeing a growing uptake of ray tracing in mobile titles," Lazarov said. "It was slow to start, but more games are now looking to enable it." The Mali G2‑Ultra NX is designed to work hand‑in‑hand with Unreal Engine tools like MegaLights, and Arm partnered with Sumo Digital to produce the demo game Neural Dawn, which serves as a showcase of the chip’s capabilities. Lazarov emphasized that the GPU targets three kinds of experiences: mobile‑first games that still aspire to desktop‑level visuals; cross‑platform titles that need a consistent look across consoles, PCs and phones; and on‑the‑go gaming where a player might switch from a home console to a train seat without a noticeable drop in quality.
Games slated to benefit from the new hardware include Tencent’s Arena Breakout Infinite, NetEase’s Where Winds Meet, and Infold Games’ Infinity Nikki. "Mobile devices have very different thermal envelopes and power budgets than consoles or PCs," Lazarov noted.
"They can’t dissipate heat the way a desktop can, and they run off batteries all the time." To address these constraints, Arm blends neural processing with the traditional rendering pipeline. By offloading portions of the workload to dedicated neuro‑accelerators, the GPU can cut both compute and power consumption, which in turn lifts system‑level efficiency. "That extra efficiency removes many of the compromises developers normally have to make," Lazarov explained. "They can reinvest the saved budget into higher visual fidelity, better performance, longer battery life, or any combination thereof." Arm backs this promise with an open‑source software stack and a toolbox of developer utilities.
Over the past two years the company has worked closely with the ecosystem to make AI‑driven graphics easy to adopt. The core techniques include Neural Super Sampling (NSS), Neural Super Sampling and Denoising (NSSD), and Neural Frame Rate Upscaling (NFRU).
Together they allow the system to upscale images, generate additional rays, and synthesize new frames through AI reconstruction. Lazarov illustrated the process with a two‑frame example.
In the first frame the GPU renders only a quarter of the pixels; the remaining three‑quarters are reconstructed by the neuro‑accelerators using NSS. In the second frame, the GPU does no traditional rendering at all—the entire image is produced by the AI engines, demonstrating a fully neural‑reconstructed frame. Nvidia’s DLSS 5 also uses AI to add detail, but it has drawn criticism for being a closed, generative system that can stray from a developer’s artistic vision.
Arm’s approach is fundamentally different. "Our networks are not generative; they are convolutional," Lazarov said. "They don’t imagine new content—they simply know how existing content should look. We train them on real data and then publish the models on GitHub and Hugging Face, so anyone can inspect, fine‑tune, or retrain them for a particular art style." This openness means a studio can capture data from its own game, retrain Arm’s public model, and ensure the upscaling respects its unique visual language, preserving intellectual‑property control.
By contrast, DLSS is a proprietary black box: developers cannot see the training data, cannot retrain it, and must accept whatever output the Nvidia model produces. Another distinction is the mobile‑first design philosophy.
Arm’s networks are built with strict limits on power, memory bandwidth and thermal headroom, making them viable on smartphones. "DLSS would simply blow past a phone’s budget even if you could get it to run," Lazarov warned. Performance numbers reinforce the claim. The Mali G2‑Ultra NX delivers roughly four times the performance and efficiency of its predecessor, while cutting DRAM traffic by about 70 %—a major source of power draw on mobile SoCs.
In a head‑to‑head test, the older GPU could only push Neural Dawn to 40 fps at best, dropping to 15 fps under typical conditions, whereas the new chip comfortably sustains 60 fps. On the tooling side, Arm currently offers plugins for Unreal Engine and is working toward Unity support, having recently announced a partnership with Unity China. For studios using custom engines, Arm provides an SDK that streamlines integration without needing a dedicated plugin.
Lazarov highlighted that existing shipped titles can be upgraded with Arm’s AI graphics in as little as three months, giving developers a rapid path to enhanced visuals. He cited Neural Dawn’s development timeline—under 20 engineers built the demo in less than 18 months—as evidence of how Arm’s technology can accelerate production while reducing the workload required to achieve high‑end graphics on mobile hardware. In summary, Arm’s Mali G2‑Ultra NX and accompanying AI‑native graphics suite aim to give mobile developers a transparent, open, and power‑efficient alternative to Nvidia’s DLSS 5, enabling richer visuals, smoother frame rates and longer battery life without sacrificing artistic control.