Arm is bringing AI-powered graphics technology closer to mainstream Android gaming with the Mali G2-Ultra NX, its first mobile GPU featuring dedicated neural accelerators. Announced on September 8, 2026, the processor introduces neural upscaling, frame generation, and more advanced ray tracing, promising better visuals and smoother gameplay without a proportional increase in power consumption.
The technology debuts in Xiaomi’s Xring O3 chipset, used in the Xiaomi 18 Fold. Arm says the new GPU delivers up to 14% higher conventional gaming performance than its predecessor, while supported neural-rendering techniques can produce substantially greater improvements under specific conditions.
The development brings mobile graphics closer to technologies already used in PC gaming, including Nvidia’s DLSS. However, its practical benefits will depend on individual game support, device cooling and how effectively smartphone manufacturers integrate the hardware.
Arm Introduces Its First AI-Native Mali GPU
The Mali G2-Ultra NX represents a significant architectural change for Arm’s mobile graphics business. Previous Mali generations focused primarily on improving conventional rendering performance and efficiency, while the new design integrates dedicated neural-processing hardware directly into the graphics pipeline.
These neural accelerators operate alongside traditional shader cores, sharing memory resources and processing infrastructure. The arrangement is intended to reduce the overhead of transferring graphics workloads between separate processing units.
According to Arm’s official announcement, the Mali G2-Ultra NX is part of its CSS for Mobile 2 platform, which also introduces an updated CPU architecture and additional on-device AI capabilities.
Arm claims up to 24% higher performance in selected graphics benchmarks and a 14% improvement in existing games compared with the previous GPU generation. Those figures are company-reported results, not independently verified performance across commercial smartphones.
Neural Super Sampling Brings AI Upscaling to Mobile Games
One of the GPU’s most important features is Neural Super Sampling, or NSS. The technology uses trained neural networks to reconstruct a higher-resolution image from graphics rendered internally at a lower resolution.
Traditional high-resolution rendering requires substantial processing power, especially in visually demanding games. Neural upscaling offers an alternative by reducing the initial rendering workload and using AI to reconstruct fine visual details.
The underlying idea resembles Nvidia’s DLSS and other reconstruction technologies available on PCs and consoles. However, Arm’s implementation is designed specifically for the memory, power and thermal limitations of mobile devices.
For players, the potential benefit is clearer images without requiring the GPU to render every pixel conventionally. Developers may also use the available processing headroom to introduce more detailed environments, richer lighting or improved performance.
Results will depend on implementation quality. Neural reconstruction can introduce artefacts around moving objects, fine textures and rapidly changing scenes, so developer integration and image stability remain important considerations.
AI Frame Generation Could Make Games Look Smoother
Arm is also developing neural frame generation through a technology called Neural Frame Rate Upscaling, or NFRU. Instead of rendering every displayed frame conventionally, the system uses AI to create intermediate frames between the game engine’s frames.
This approach can increase the number of frames shown onscreen while reducing the workload required to render each one traditionally. Arm aims to deliver visually smoother gameplay within a smartphone’s limited power budget.
On its official Mali G2-Ultra NX product page, Arm claims that supported neural graphics workloads can achieve up to four times the displayed frame rate at a sustained power budget compared with native rendering.
That figure requires careful interpretation. Four times as many displayed frames does not mean the GPU becomes four times faster at rendering conventional game frames, nor does it guarantee a corresponding improvement in control responsiveness.
Frame generation introduces synthetic intermediate images, while player input and game simulation continue to run through the underlying rendering pipeline. The resulting experience depends on the original frame rate, latency management and the quality of generated frames.
Arm has demonstrated neural graphics at up to 120 frames per second, but the company has not established that every supported smartphone or game will reach that figure.
Third-Generation Ray Tracing Improves Lighting and Reflections
The Mali G2-Ultra NX also introduces Arm’s third-generation ray tracing hardware, designed to handle more complex lighting, reflections and shadows efficiently.
Ray tracing models how light interacts with objects and surfaces, allowing games to produce more realistic reflections, illumination and shadow behaviour. Although the technology is established on high-end PCs and consoles, its computational demands have limited its adoption on smartphones.
Arm says the new architecture reduces memory traffic by up to 13% in selected ray tracing benchmarks. Lower memory bandwidth requirements can improve efficiency, especially in games with complex environments and detailed geometry.
The GPU also supports hardware-accelerated Opacity Micromaps, a technique that makes ray tracing more efficient for objects with complex transparent or partially transparent surfaces.
Such improvements could benefit detailed vegetation, layered materials and other complex environments. However, actual image quality and performance will depend on how individual game engines use the hardware.
Xiaomi Xring O3 Introduces the Technology in Smartphones
Xiaomi is among the first manufacturers to implement the Mali G2-Ultra NX, integrating it into the company’s Xring O3 system-on-chip for the Xiaomi 18 Fold.
The implementation lets Xiaomi combine its processor design with ARM’s neural graphics architecture. It also demonstrates how smartphone manufacturers can adopt these capabilities without developing every component of the graphics pipeline independently.
Initial availability is centred on Xiaomi’s Chinese-market device. Broader adoption will depend on other chip designers and smartphone manufacturers licensing and integrating the new GPU architecture.
Arm’s technology is not exclusive to Xiaomi. The Mali architecture is licensed across the mobile industry, giving the company a potential route to introduce neural graphics into additional Android phones, tablets and other devices.
However, that doesn’t determine which future commercial products will use the Mali G2-Ultra NX. Manufacturers must announce their own processor implementations, device specifications and release schedules.
Where Winds Meet, Infinity Nikki and Arena Breakout Are Early Partners
Hardware support alone cannot deliver meaningful neural-rendering improvements. Game developers must integrate the relevant features into their graphics pipelines, making software adoption essential to the technology’s success.
Arm has identified several early projects involving established game developers. NetEase plans to add Neural Super Sampling support to Where Winds Meet in 2026, while Infold Games is integrating the technology into Infinity Nikki.
Arm is also working with Tencent Games on a neural super sampling and denoising demonstration using Arena Breakout: Infinite.
These partnerships show growing developer interest, but they don’t confirm that all three games already have every neural-rendering feature.
The company has also collaborated with Sumo Digital on Neural Dawn, a technical demonstration designed to showcase sophisticated mobile graphics using neural processing and modern rendering techniques.
Arm supports adoption through its Neural Graphics Development Kit, which gives developers tools to experiment with AI-assisted rendering.
What AI Graphics Mean for Smartphone Battery Life
Smartphone gaming performance depends on more than processing power. Unlike most desktop gaming PCs, smartphones must operate within strict temperature limits while preserving battery life and keeping the device comfortable to the touch.
Conventional rendering of complex graphics can quickly increase power consumption. As a device heats up, its processor may reduce speeds to stay within safe thermal limits, hurting sustained gaming performance.
Neural rendering aims to address part of that problem by delivering comparable visual results with less traditional GPU work. Arm reports up to four times greater performance per watt for supported neural graphics workloads.
That improvement could let compatible games maintain higher displayed frame rates within the same power budget. Alternatively, developers might use the efficiency gains to improve image quality without increasing power consumption as much as conventional rendering would require.
These are potential benefits, not guarantees, of longer battery life. Overall energy consumption also depends on the display, CPU, memory, networking activity, game engine and thermal design of the finished device.
Can Arm Challenge Nvidia DLSS in Mobile Gaming?
Arm’s approach shares several ideas with Nvidia’s DLSS, particularly the use of neural networks for image reconstruction and frame generation. Both technologies attempt to deliver better perceived graphics performance without requiring the hardware to render every displayed pixel or frame conventionally.
The important difference is their intended environment. Nvidia’s DLSS primarily runs on supported GeForce RTX hardware, while Arm is developing neural rendering for mobile GPUs in compact, power-constrained devices.
Arm also benefits from relationships with smartphone chip designers and game engine developers. If enough manufacturers and developers adopt it, its architecture could make AI-assisted rendering available across a broad range of Android devices.
Nevertheless, matching the maturity of established PC graphics technologies will require more than promising demonstrations. Developers will need reliable tools, consistent visual quality, predictable performance and enough compatible devices to justify integration.
The availability of supported commercial games will ultimately determine whether neural rendering becomes a standard mobile graphics feature or remains limited to selected flagship devices.
What the Mali G2-Ultra NX Means for Android Gaming
The Mali G2-Ultra NX points to a future in which smartphone gaming performance increasingly depends on a combination of traditional graphics hardware and dedicated AI acceleration.
Its neural upscaling, frame-generation capabilities and improved ray tracing could help mobile games deliver richer environments while managing the power and thermal restrictions of handheld devices.
Arm’s reported performance improvements are substantial, but they remain conditional on software support and real-world implementation. Independent testing of finished devices will be necessary to determine how consistently those benefits appear during extended gameplay.
Xiaomi’s Xring O3 offers an early commercial example, while planned integrations with Where Winds Meet, Infinity Nikki, and Arena Breakout: Infinite show that developers are beginning to explore the technology.
For Android gaming, the significance extends beyond a single GPU launch. If neural rendering becomes widely supported, future smartphones could deliver more sophisticated graphics without relying exclusively on increasingly powerful and energy-intensive conventional rendering hardware.
