Arm, the British semiconductor and software design powerhouse, recently staged its Arm Everywhere China showcase in Shanghai to announce a suite of fresh AI‑computing initiatives aimed at the mobile market. The centerpiece of the event was the unveiling of an end‑to‑end AI‑native ecosystem that includes the new CSS for Mobile 2 compute platform and the Mali G2‑Ultra NX graphics processor. Unlike conventional GPUs, the Mali G2‑Ultra NX is built around neural‑accelerator cores that promise higher graphics throughput while keeping power draw and heat generation to a minimum. Arm positions this on‑device AI solution as a more transparent alternative to Nvidia’s controversial DLSS 5, arguing that it aligns better with developers’ original creative intent.
While Arm did not disclose the exact product line‑ups that will ship the Mali G2‑Ultra NX, the predecessor Mali G1‑Ultra has already found its way into smartphones from brands such as Xiaomi, Vivo and OnePlus, suggesting a similar market trajectory for the new chip. Deyan Lazarov, a senior member of Arm’s GPU team within the Edge AI division, explained that development on the G2‑Ultra NX has spanned roughly two and a half years.
The design goal was to give mobile developers access to compute‑intensive features—most notably real‑time ray tracing—without the prohibitive power budgets that have traditionally limited such capabilities on handheld devices. "We’re seeing a steady rise in ray‑tracing adoption on mobile titles," Lazarov noted. "It was a slow burn at first, but now more games are actively looking to enable it." This trend is reinforced by recent reports from GamesIndustry.biz that the new GPU will support Unreal Engine technologies like MegaLights.
Arm has already partnered with Sumo Digital to produce a demo title called Neural Dawn, which showcases the G2‑Ultra NX’s ability to render high‑fidelity scenes on a smartphone. Lazarov described the chip’s target audience as developers who create "mobile‑first" experiences but still want their games to look as polished as desktop releases. He also highlighted cross‑platform titles that aim to deliver a consistent visual experience whether the player is on a console at home or a train in London. "The challenge with phones is their thermal envelope and power budget are far tighter than a console or PC," he explained.
"They can’t dissipate heat the same way, and they’re always running off a battery." Arm’s answer is to weave neural processing directly into the traditional rendering pipeline. By offloading portions of the workload to dedicated neuro‑accelerators, the GPU can render fewer native pixels and let the AI reconstruct the remainder. This reduces both computational load and energy consumption, which in turn lifts overall system efficiency. "In practice that means developers face fewer trade‑offs," Lazarov said.
"They can reclaim budget that would otherwise be spent on raw rendering and reinvest it in visual fidelity, performance, battery life, or any combination thereof." To make this vision accessible, Arm is taking an open‑software stance. Over the past two years the company has collaborated with ecosystem partners to build a toolkit that developers can use without licensing hurdles. The core of the offering includes Neural Super Sampling (NSS), Neural Super Sampling and Denoising (NSSD), and Neural Frame‑Rate Upscaling (NFRU). These technologies enable image upscaling, additional ray generation, and AI‑driven frame synthesis.
Lazarov illustrated the process with a two‑frame example: in the first frame the GPU renders only 25 % of the pixels while the neuro‑accelerators reconstruct the remaining three‑quarters via NSS; in the second frame the GPU does no traditional rendering at all, and the entire image is produced by the AI system. Nvidia’s DLSS 5 also relies on AI to enhance image detail, but Arm argues that its approach is fundamentally different. "Our networks are not generative; they’re convolutional and trained on real content," Lazarov emphasized.
"They don’t hallucinate new details—they simply know how the image should look based on the data they’ve seen. Moreover, all of our models are publicly available on GitHub and Hugging Face." This openness lets developers fine‑tune the models for their own artistic styles.
If a game has a distinctive visual language, a studio can capture representative frames, retrain the network on that dataset, and ensure that upscaled or interpolated frames preserve the unique look. "Because you control the retraining, there’s no risk of losing IP or having the AI impose an unwanted aesthetic," Lazarov added. By contrast, DLSS is a closed‑source black box: developers cannot see the training data, nor can they modify the model, leaving them with a take‑it‑or‑leave‑it proposition.
Arm also stresses that its solution was engineered with mobile constraints in mind—power, memory bandwidth, and thermal limits are baked into the network design. "If you tried to run DLSS on a phone, you’d instantly exceed the power envelope," Lazarov warned. The Mali G2‑Ultra NX, however, delivers up to four times the performance and efficiency of its predecessor while cutting DRAM traffic by roughly 70 %, a major source of power draw on smartphones.
Performance numbers illustrate the leap. When pushed to its limits, the previous generation GPU could only sustain Neural Dawn at about 40 fps under ideal conditions and dropped to 15 fps in typical use. The new G2‑Ultra NX comfortably hits 60 fps on the same title, delivering smoother gameplay without overheating.
On the developer side, Arm currently provides Unreal Engine plug‑ins and plans to extend support to Unity and other engines. A recent partnership with Unity China signals that Unity integration is on the roadmap, though Unreal remains the immediate priority. For studios using custom engines, Arm offers an SDK that streamlines integration without requiring engine‑specific plug‑ins. Lazarov highlighted that retrofitting an existing shipped game with Arm’s AI‑native graphics can be accomplished in as little as three months, allowing publishers to roll out visual upgrades without a full remake.
The Neural Dawn demo itself was built by a sub‑20‑person team in under 18 months, showcasing how the technology can accelerate development cycles while delivering high‑quality results. In summary, Arm’s mobile‑first AI reconstruction strategy seeks to give developers a transparent, customizable, and power‑efficient alternative to Nvidia’s DLSS 5. By open‑sourcing its convolutional networks, integrating neural accelerators into the rendering pipeline, and providing robust tooling for major game engines, Arm aims to empower creators to push the visual envelope on smartphones without sacrificing battery life or thermal stability.