Arm has unveiled the Mali G2-Ultra NX, the first GPU in its Mali family built specifically for neural graphics, promising desktop-class visual quality on smartphones by putting AI acceleration at the heart of the graphics pipeline rather than bolting it on.
Announced Tuesday on Arm's newsroom, the new architecture integrates dedicated neural accelerators directly within the shader cores, allowing neural workloads to run alongside conventional graphics and compute tasks while sharing the GPU's memory system, coherent caches and control structures. Arm says that tight integration reduces data movement and improves efficiency — a critical consideration inside the strict power and thermal envelope of a mobile device. For more latest AI developments, follow AI Buzz Wire's ongoing coverage. latest AI developments
Three Neural Technologies for Mobile Gaming
The Mali G2-Ultra NX ships with three headline neural techniques:
Neural Super Sampling (NSS) reconstructs higher-resolution images from lower-resolution renders, combining temporal data with integrated anti-aliasing to cut the amount of conventional rendering work needed for a high-quality final frame.
Neural Frame Rate Upscaling (NFRU) generates intermediate frames using motion, depth and rendered-frame data, boosting frame rates for smoother gameplay — and, according to Arm, supporting longer mobile gaming sessions at up to 120 FPS.
Neural Super Sampling and Denoising (NSSD) pairs neural upscaling with denoising to improve image quality in demanding ray-traced scenes with complex lighting and shadows.
To demonstrate the pipeline in production, Arm and game developer Sumo Digital built a showcase called Neural Dawn. Arm says the demo achieves up to 4x higher performance efficiency and up to 70 percent lower external memory traffic compared with native rendering, and makes techniques such as Unreal Engine's MegaLights practical on mobile hardware.
The Biggest Mali ISA Upgrade in Seven Generations
Beyond the neural hardware, the G2-Ultra NX introduces a new execution engine that Arm describes as the biggest Mali instruction set architecture upgrade in seven generations. It provides up to twice the registers per warp, giving developers headroom for richer scenes and more complex effects while holding frame rates steady.
The architectural gains translate into measurable numbers, at least according to Arm's own benchmarks: up to 24 percent higher benchmark performance and 14 percent higher non-AI gaming performance compared with the previous generation.
Third-Generation Ray Tracing
Ray tracing is back for a third round. The G2-Ultra NX carries Arm's third-generation hardware ray tracing unit, which the company says supports more complex lighting, shadows, reflections and geometry while cutting DRAM traffic by up to 13 percent on leading ray tracing benchmarks.
New hardware support for Opacity Micromaps helps the GPU handle intricate transparent geometry — foliage, fabrics and layered surfaces — more efficiently. In one Arm demo, using Opacity Micromaps increased frame rates by 30 percent and reduced the ray tracing workload by up to 70 percent.
Why AI-Native Graphics Matter
The announcement reflects a broader industry conviction that the future of real-time graphics is hybrid: conventional rendering for what can be computed cheaply, neural reconstruction for everything else. Desktop GPUs from Nvidia, AMD and Intel have normalized AI upscaling and frame generation at the high end; console vendors use machine-learning techniques for similar gains. Mobile has lagged, largely because neural models compete for the same power budget as the game itself.
Arm's answer is architectural: rather than treating AI as a separate accelerator competing for memory bandwidth, the G2-Ultra NX embeds neural units inside the shader cores where the data already lives. The company argues this lets neural techniques move from tech demos to production games, within the power envelope of a phone.
That memory argument deserves emphasis. In mobile system-on-chips, external memory traffic is one of the largest consumers of both power and latency — moving data off the GPU and back costs more energy than computing on it. By reusing the GPU's coherent caches and control structures for neural workloads, Arm claims the design avoids much of that round trip. The 70 percent reduction in external memory traffic from the Neural Dawn demo is, in that sense, as much a battery-life claim as a performance one.
The company is also making an ecosystem play. Arm's announcement frames the G2-Ultra NX as a way for developers to bring neural graphics into production faster, suggesting that reference implementations like Neural Dawn are meant to be blueprints studios can adapt, rather than one-off marketing demos. Mobile users' expectations — richer visuals, smoother frame rates, more immersive scenes — continue to climb, and Arm is betting developers will adopt whichever pipeline meets those expectations within thermal limits.
The scale of the installed base gives the pitch weight. Arm says more than 14 billion Mali GPUs have shipped to date, and while the flagship G2-Ultra NX will appear first in premium devices, neural graphics techniques tend to trickle down across licensing partners' product lines in subsequent generations.
For game developers, the message is that mobile players increasingly expect the visual fidelity and smoothness associated with consoles — and that meeting that expectation within a phone's thermal limits now requires AI at every stage of the pipeline. Arm has effectively declared that the era of treating neural rendering as optional for mobile GPUs is over.
Stay Ahead of AI
Track the latest AI developments and artificial intelligence hardware news as on-device AI reshapes the chip industry.
Read more AI news →
