Arm Details Its Mobile‑First AI Reconstruction Approach, Diverging from the "Black Box" Model of DLSS 5

Arm, the British chip and software design powerhouse, recently staged its Arm Everywhere China showcase in Shanghai, unveiling a suite of fresh AI‑driven technologies aimed at reshaping mobile graphics. The centerpiece of the announcement is a comprehensive AI‑native computing ecosystem that includes the CSS for Mobile 2 compute platform and the brand‑new Mali G2‑Ultra NX GPU. This graphics processor is built around neural‑computing blocks, promising higher visual fidelity while keeping power draw and heat generation to a minimum. Arm positions its on‑device AI solution as a more transparent alternative to Nvidia’s controversial DLSS 5, aligning more closely with developers’ original creative intent.

While Arm did not name specific products that will ship with the Mali G2‑Ultra NX, the predecessor Mali G1‑Ultra already powers smartphones from brands such as Xiaomi, Vivo and OnePlus, suggesting a similar market trajectory. Deyan Lazarov, a senior 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 design goal is to give mobile developers access to heavyweight rendering techniques—most notably ray tracing—without the prohibitive energy costs traditionally associated with those features. "We’re seeing a real uptick in interest for ray tracing on phones," Lazarov noted. "It used to be a niche, but more and more titles are asking for that level of realism." The Mali G2‑Ultra NX is already being leveraged in demos such as Sumo Digital’s "Neural Dawn," which showcases how Unreal Engine’s MegaLights and other advanced features can run smoothly on a handheld device. Other upcoming titles slated to benefit from the chip 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, smartphones operate with limited thermal envelopes and rely on battery power.

"A phone can’t dump heat the way a desktop can, and it’s always on a battery," he said. To address this, Arm fuses neural processing units (NPUs) directly into the rendering pipeline. By offloading portions of the image synthesis to the NPU, the GPU’s workload—and consequently its power consumption—drops dramatically.

This efficiency gain frees developers from the usual trade‑offs between visual quality, frame rate, and battery life. Arm’s software strategy is deliberately open. Over the past two years the company has collaborated with ecosystem partners to create a developer‑friendly AI stack. Lazarov explained that the goal was to build AI tools that are both powerful and accessible, allowing developers to reallocate saved resources toward higher resolution textures, smoother frame rates, longer battery endurance, or any combination thereof.

Key components of the stack include Neural Super Sampling (NSS), Neural Super Sampling and Denoising (NSSD), and Neural Frame Rate Upscaling (NFRU). These technologies can upscale lower‑resolution frames, generate additional rays for improved lighting, and synthesize entirely new frames using AI‑based reconstruction. In practice, a typical frame might be rendered by the GPU at only 25 % of the final resolution, with the remaining 75 % reconstructed by the on‑chip neural accelerators. In some cases, the GPU does virtually no traditional rendering at all; the NPU produces the entire image, delivering a fully AI‑reconstructed output.

Nvidia’s DLSS 5 follows a similar high‑level concept—using AI to enhance image detail—but it has drawn criticism for its "black‑box" nature. Lazarov drew a clear line between the two approaches: "Our networks are convolutional, not generative. They don’t hallucinate new content; they simply learn how existing content should look and apply that knowledge.

All of our models are open‑source on GitHub and Hugging Face, so developers can inspect, modify, and even retrain them on their own art assets. This eliminates any risk of losing artistic intent or intellectual‑property control." By contrast, DLSS is a proprietary system whose inner workings are hidden from developers.

"You get a take‑it‑or‑leave‑it solution," Lazarov said. "With Arm, you can take the model, fine‑tune it on your own data, and guarantee that the upscaled frames respect your unique visual style." Another distinction lies in the mobile‑first design philosophy. Arm’s neural networks are engineered to respect the strict power, bandwidth, and memory budgets of smartphones. Attempting to run DLSS on a phone would instantly exceed those limits, even if the software were technically compatible.

Arm’s solution, however, delivers up to four‑times performance and efficiency improvements while slashing DRAM traffic by roughly 70 %, a major source of power draw on mobile devices. Performance figures illustrate the impact. 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 Mali G2‑Ultra NX, by contrast, runs the same demo smoothly at 60 fps, demonstrating both higher throughput and lower energy consumption. From a tooling perspective, Arm currently offers plug‑ins for Unreal Engine and plans to extend support to Unity and other engines. The company recently announced a partnership with Unity China, signaling a broader intent to cover the most popular development environments.

For studios that rely on custom engines, Arm provides a comprehensive SDK that streamlines integration without requiring engine‑specific plug‑ins. Lazarov highlighted that retrofitting an existing shipped game with Arm’s AI graphics can be accomplished in as little as three months, enabling rapid iteration and post‑launch enhancements. He cited the development of "Neural Dawn"—built by a team of fewer than 20 people in under 18 months—as proof that Arm’s technology can accelerate both new and legacy projects. In summary, Arm’s mobile‑first AI reconstruction platform offers a transparent, open, and power‑efficient alternative to Nvidia’s closed‑source DLSS 5.

By embedding neural accelerators into the rendering pipeline, providing open‑source convolutional models, and delivering robust development tools, Arm aims to empower developers to create console‑quality visuals on handheld devices without sacrificing battery life or thermal headroom. The Mali G2‑Ultra NX GPU, together with the broader AI‑native ecosystem, signals a significant step toward a future where high‑fidelity, AI‑enhanced graphics become commonplace across the entire spectrum of mobile gaming hardware.