Arm, the British chip and software design powerhouse, staged its Arm Everywhere China showcase in Shanghai to announce a suite of fresh AI‑computing advances aimed squarely at mobile devices. The centerpiece of the announcement is a new AI‑native compute platform called CSS for Mobile 2, together with the Mali G2‑Ultra NX GPU – a graphics processor built around neural‑accelerator blocks that promise higher visual fidelity while keeping heat and power draw modest. Arm positions its on‑device AI‑enhanced graphics pipeline as an open, developer‑friendly alternative to Nvidia’s controversial DLSS 5, arguing that its method stays truer to a developer’s original artistic intent. While Arm did not disclose specific product partners, the predecessor Mali G1‑Ultra has already shipped in smartphones from Xiaomi, Vivo and OnePlus, suggesting the upcoming G2‑Ultra NX will soon appear in a similar tier of Android handsets.
Deyan Lazarov, a senior engineer on Arm’s GPU team within the Edge‑AI division, explained that the chip has been in development for roughly two and a half years. Its primary goal is to give mobile creators the ability to harness compute‑heavy techniques such as ray tracing without sacrificing battery life or thermal headroom. "We’ve seen a steady rise in ray‑tracing adoption on phones," Lazarov noted.
"It was slow to get started, but more titles are now experimenting with it, and the G2‑Ultra NX is designed to make that feasible." In partnership with Sumo Digital, Arm produced a demo title called Neural Dawn that showcases the GPU’s capabilities, including support for Unreal Engine features like MegaLights. Other upcoming games slated to run on the new silicon include Tencent’s Arena Breakout Infinite, NetEase’s Where Winds Meet, and Infold Games’ Infinity Nikki. Lazarov emphasized the unique constraints of mobile hardware: unlike consoles or PCs, phones cannot dissipate large amounts of heat and must operate continuously on battery power.
To address this, Arm’s solution blends neural processing with the traditional rendering pipeline. By offloading portions of the image synthesis to dedicated neuro‑accelerators, the GPU workload drops dramatically, leading to lower power consumption and, consequently, a more efficient overall system.
This efficiency, Lazarov argues, frees developers from making hard trade‑offs between visual quality, frame rate, and battery life. Arm is backing this vision with an open‑source software stack. Over the past two years the company has worked closely with the broader ecosystem to deliver tools that let developers plug AI‑enhanced graphics into their workflows without a steep learning curve.
"We wanted an AI approach that is developer‑friendly and gives clear, measurable efficiency gains," Lazarov said. "Those gains can be reinvested in higher resolution, smoother performance, longer battery life, or any combination the studio prefers." The core techniques Arm is promoting are Neural Super Sampling (NSS), Neural Super Sampling and Denoising (NSSD), and Neural Frame‑Rate Upscaling (NFRU). In practice, NSS can render only a quarter of a frame on the GPU and then reconstruct the remaining three‑quarters using the neuro‑accelerators.
In an even more extreme mode, the entire frame can be generated by the AI engine, eliminating traditional rasterisation for that frame altogether. This contrasts with Nvidia’s DLSS 5, which relies on generative networks that can introduce content that diverges from a developer’s original vision. "Our networks are convolutional, not generative," Lazarov explained. "They don’t hallucinate new details; they simply learn how existing content should look.
Because we train them on real‑world data and publish the models on GitHub and Hugging Face, developers can fine‑tune them with their own art assets. That guarantees the upscaled output respects the game’s unique style and protects intellectual‑property rights." By contrast, DLSS is a closed‑source, black‑box solution.
Developers cannot inspect the training data or retrain the model for a specific aesthetic, leaving them with a take‑it‑or‑leave‑it proposition. Arm’s open approach also means the AI models are engineered with mobile constraints in mind—power, bandwidth, and memory budgets are baked into the design. Lazarov warned that trying to run a DLSS‑style system on a phone would instantly blow those limits.
Power efficiency is a headline claim for the Mali G2‑Ultra NX. Arm reports up to a four‑fold increase in performance and efficiency, alongside a 70 % reduction in DRAM traffic—a major source of energy consumption on smartphones. In a side‑by‑side test, the previous‑generation GPU could only push Neural Dawn to 40 fps at its limit and fell to 15 fps under normal conditions, whereas the new chip sustains a smooth 60 fps.
From a developer tooling perspective, Arm currently offers plug‑ins for Unreal Engine and plans to add Unity support later, having already announced a partnership with Unity China. For studios using custom engines, Arm provides a lightweight SDK that streamlines integration without requiring deep knowledge of the underlying AI pipelines.
Lazarov claims that retrofitting an existing shipped title with Arm’s AI graphics can be accomplished in as little as three months, delivering noticeable visual upgrades with minimal effort. The rapid development cycle was illustrated by Neural Dawn itself, which a team of fewer than 20 built in under 18 months.
Arm’s technology, Lazarov says, enables developers to quickly enhance legacy games or launch groundbreaking new experiences without the usual overhead of extensive re‑engineering. In summary, Arm’s mobile‑first AI reconstruction strategy seeks to give game creators a transparent, adaptable, and power‑aware alternative to Nvidia’s DLSS 5.
By open‑sourcing its convolutional networks, integrating neural accelerators directly into the rendering path, and delivering a robust SDK and engine plug‑ins, Arm aims to democratize high‑quality graphics on phones, tablets, laptops and other consumer devices, while keeping thermal and battery constraints firmly in check.