ARM's neural upscaler wants to bring AAA gaming to your phone

Playing high-end console ports on a smartphone still feels like a compromise. You usually get muddy textures, jagged edges, and a device hot enough to fry an egg. While PC gaming solved this bottleneck years ago through AI upscalers like DLSS and FSR, mobile chips have largely been left behind.

Now, ARM wants to bring that same magic directly into your pocket with a new architecture built specifically for phones.

540p in, 1080p out

ARM revealed details on Tuesday for its Neural Super Sampling tech, a dedicated AI upscaling solution designed for future mobile chip architectures. Instead of rendering a game at native resolution, the hardware runs the game internally at 540p. A trained neural model then predicts and reconstructs the missing pixels to output a crisp 1080p frame.

ARM claims the visual output looks remarkably close to native rendering, avoiding the soft edges and blurred textures that plagued earlier mobile upscaling attempts.

Last year, the chip designer introduced Accuracy Super Resolution, an upscaler relying on temporal data from multiple frames. ARM noted that traditional temporal upscalers struggle to scale properly on mobile form factors. Neural Super Sampling shifts the heavy lifting to dedicated AI models trained specifically for tight mobile constraints.

Cutting the native rendering workload directly addresses the biggest headache in mobile gaming: thermal throttling and battery drain. Titles like Honkai Star Rail chew through smartphone batteries at maximum resolution. Offloading pixel generation to neural silicon could keep phones cooler while significantly extending playtime.

The long road to actual devices

There is a big catch: you will not be playing with this anytime soon. Smartphone makers and silicon partners still need to integrate the neural hardware into upcoming chipsets. Reports indicate we are at least a year away from seeing this silicon in commercial devices.

Latency is another major hurdle. Running frames through an AI upscaler takes processing time. If next-generation GPUs are not configured properly by chipmakers, that extra overhead can introduce noticeable input lag during fast-paced gameplay.

Developers also need convincing. While ARM is rolling out a plugin for Unreal Engine 5, studios will only invest resources into optimization if there is a massive installed base of compatible phones. Mobile players have different habits than PC or console gamers, and publishers will wait to see if the demand for heavy AAA mobile gaming actually materializes.

Why it matters

For gamers, this could finally turn phones into viable handheld consoles without battery anxiety. If it works, you get sharper graphics and double the battery life on demanding titles without having to carry a dedicated handheld PC.

For the industry, ARM is tackling the biggest bottleneck holding mobile games back. Phones are already fast enough to outlive older consoles—HoYoverse recently dropped Genshin Impact from the PS4 citing hardware limitations, yet the game remains active on phones. Raw compute is no longer the issue; power budgets are. A standardized neural upscaler from the core architecture provider could set a unified baseline across Android devices.

We think ARM's technical approach is clever, but its business model remains the weak spot. Unlike Nvidia, ARM does not sell finished silicon. Success depends entirely on whether partners like Qualcomm or MediaTek choose to prioritize this neural hardware over other features. If integration is inconsistent or the latency penalties ruin gameplay, this tech risks becoming another demo-reel feature that gamers keep turned off.