Arm, the British chip and software design firm, staged its Arm Everywhere China showcase in Shanghai this week to announce fresh advances in AI‑driven computing. The centerpiece of the presentation was a vision for an end‑to‑end AI‑native ecosystem that includes the CSS for Mobile 2 compute platform and a brand‑new graphics processor, the Mali G2‑Ultra NX GPU.
This GPU is built around neural‑accelerator blocks and promises higher visual fidelity while keeping thermal output and power draw far lower than traditional mobile graphics solutions. While Arm did not name specific products that will ship with the Mali G2‑Ultra NX, the earlier Mali G1‑Ultra has already found its way into smartphones from Xiaomi, Vivo and OnePlus, suggesting a similar market path for the new chip. Deyan Lazarov, a member of Arm’s GPU team within the Edge AI division, explained that the G2‑Ultra NX has been under development for roughly two and a half years. Its primary goal is to give mobile developers a practical way to employ compute‑heavy techniques such as ray tracing without exhausting the limited battery and thermal envelopes of phones, tablets and handheld consoles.
"We’re seeing a steady rise in mobile titles that want to use ray tracing," Lazarov said. "It was slow to start, but now more games are asking for it, and our hardware is ready to support those ambitions." The new GPU also integrates tightly with Unreal Engine features like MegaLights. Arm collaborated with Sumo Digital to produce a demo titled *Neural Dawn*, which showcases how the Mali G2‑Ultra NX can render complex lighting and geometry on a handheld device.
Lazarov emphasized that the chip targets "mobile‑first" games that still aspire to look like desktop experiences, as well as cross‑platform titles that need a consistent visual quality whether they run on a console at home or on a train in London. Upcoming games slated to use the processor include Tencent's *Arena Breakout Infinite*, NetEase's *Where Winds Meet* and Infold Games' *Infinity Nikki*. Lazarov noted that mobile devices face distinct constraints: they cannot dissipate heat like a PC or console and must operate continuously on battery power.
To address this, Arm blends neural processing with the conventional rendering pipeline. By offloading portions of the image generation to dedicated neuro‑accelerators, the GPU workload shrinks, leading to lower power consumption and higher overall system efficiency. "That extra efficiency removes many of the compromises developers usually have to make," Lazarov explained.
"They can choose to reinvest the saved budget into sharper visuals, smoother frame rates, longer battery life, or any combination thereof." Arm is backing this vision with an open‑source software stack. Over the past two years the company has worked with ecosystem partners to make its AI tools developer‑friendly. The core technologies—Neural Super Sampling (NSS), Neural Super Sampling and Denoising (NSSD) and Neural Frame Rate Upscaling (NFRU)—are all released publicly on GitHub and Hugging Face. Unlike Nvidia's DLSS 5, which relies on proprietary, generative networks that act as a "black box," Arm's models are convolutional, trained on real‑world content, and can be fine‑tuned by developers on their own art assets.
Lazarov gave a concrete illustration: in one scenario the GPU renders only 25 % of a frame, while the remaining three‑quarters are reconstructed by the neuro‑accelerators using NSS. In a more extreme case, the entire frame is generated by the AI engines with no traditional rasterisation at all. This approach yields up to four times the performance and efficiency of previous mobile GPUs, and cuts DRAM traffic by roughly 70 %, a major source of power draw on handheld devices. When comparing to DLSS 5, Lazarov stressed that Arm's networks do not "imagine" new details; they simply infer the most likely appearance based on learned patterns, preserving the creator's intent.
Because the models are open, developers can retrain them with custom datasets that reflect a game's unique visual style, ensuring that upscaled images and inserted frames stay true to the original art direction and protecting intellectual‑property rights. The openness also contrasts with Nvidia's closed‑box model, where developers cannot see the training data or modify the algorithm. Arm's philosophy is "take it and make it work for you," rather than a binary "take it or leave it" approach.
Another key distinction is mobile readiness. DLSS, even if technically ported, would exceed power and bandwidth budgets on a phone, whereas Arm's solution is engineered from the ground up to respect those limits. Lazarov highlighted a real‑world benchmark: the previous‑generation Mali GPU could only push *Neural Dawn* to 40 fps at its limit and dropped to 15 fps under normal conditions, while the new G2‑Ultra NX easily sustains 60 fps.
On the tooling side, Arm currently offers plug‑ins for Unreal Engine and plans to expand to Unity and other engines. The company recently announced a partnership with Unity China, signaling broader support. For studios using proprietary engines, Arm provides an SDK that streamlines integration without needing a dedicated plug‑in. Lazarov added that retrofitting existing games with Arm's AI‑native graphics can be fast: a three‑month development cycle is enough to ship an updated version with upscaling, denoising and frame‑rate enhancement features.
The *Neural Dawn* demo itself was built by fewer than 20 engineers in under 18 months, demonstrating how Arm's technology can accelerate development and reduce the effort required to bring high‑quality visuals to mobile platforms. In summary, Arm's mobile‑first AI reconstruction strategy offers an open, efficient, and developer‑centric alternative to Nvidia's DLSS 5, promising higher performance, lower power use, and greater creative control for the next generation of handheld gaming experiences.