Smartphones are getting Nvidia DLSS-style tech, and I can’t wait

There’s no escaping AI technologies these days, and when computer giant NVIDIA attempts to blend this hugely lucrative future with its gaming legacy, controversy soon follows. Its latest DLSS 5 technology is a prime example, not only building on its AI-infused frame generation and upscaling tech, but also generating its own “photoreal” pixels to replace original textures and lighting.

Where the PC and console industries lead, smartphones soon follow. It won’t be long before more and more smartphone games start using AI-assisted features to improve performance and image quality, like frame generation and upscaling. I’m skeptical but intrigued.

On that note, Arm has just unveiled its new Neural Dawn technological showcase. Developed by Arm and Sumo Digital, Neural Dawn leverages Arm Neural Technology to become the first mobile game to deliver Unreal Engine’s MegaLights, which blends real-time, dynamic lighting with ray-traced shadows for console-like quality. The lighting in the tech demo video looks undeniably impressive.

Peering deeper into the Arm Neural Technology layer, we find familiar DLSS-like ideas trimmed down for mobile. Arm’s Neural Super Sampling (NSS) can deliver 540p to 1080p upscaling at just 4ms per frame, freeing up GPU resources for enhanced lighting and other effects. Neural Super Sampling and Denoising (NSSD) removes noise generated by ray tracing, while Neural Frame Rate Upscaling (NFRU) generates intermediate frames, making a 30fps base rate look closer to 60fps.

The only caveat is that Arm notes this is a showcase for next-generation Mali GPUs that will launch later this year, meaning we’ll likely have to upgrade to a new flagship phone to enjoy the benefits.

But Arm isn’t the only mobile GPU player rolling out AI-enhanced performance features. Qualcomm’s upcoming AI Frame Fusion (AIFF) — an improved version of Snapdragon Game Super Resolution — supports both frame generation and resolution upscaling. What’s quite interesting about Qualcomm’s proposition is that it can run on either the GPU or the NPU, if available, potentially making it more performant.

Apple’s MetalFX Upscaling uses a lighter neural spatial upscaler and a non-AI temporal option, allowing iPhones and iPads to power games like Resident Evil and No Man’s Sky. Likewise, MediaTek’s HyperEngine offers AI-based Variable Rate Shading and ray-tracing. Even the popular PC emulator GameNative integrated Lossless Scaling Frame Generation (LSFG) into its latest builds, which uses Vulkan on a traditional GPU layer rather than requiring “neural” or AI capabilities.

While we’re still in the relatively early stages of smartphone development, the technology looks set to become a mainstay of the mobile games industry.

So what’s in it for mobile? A lot, actually. Lower native resolution means less GPU work, less heat, and less throttling. That alone is huge — sustained performance is arguably the biggest problem in mobile gaming. On top of that, frame-gen can deliver smooth 120fps, making the most of our high-refresh-rate displays without added heat or power draw. In some ways, mobile gaming has more to gain from this tech than PC ever did.

However, as promising as this sounds, mobile is fraught with its own technical challenges, making it easier said than done.

For starters, frame generation is not a magic cure-all for low performance. As anyone who has tried the technology in the PC space will attest, frame generation works best by smoothing an already robust frame rate, but still struggles with high latency and jankiness when the baseline frame rate is low.

For instance, frame generation works brilliantly at taking 60fps and interpolating it to your display’s 120fps refresh rate. That’s the ideal case. Using frame-gen to push from a 30fps baseline to 60fps is far riskier — the added latency makes the result feel sluggish, even if the frame counter says otherwise. Both Arm and Qualcomm have mentioned targeting a 30fps baseline, which is a little concerning.

Still, if implemented properly, frame generation should be an even more powerful tool on mobile than on PC gaming. Rendering at 60fps requires less GPU work, uses less battery, and produces less heat than fully rendering at 120fps. This is the reason so many mobile games prefer limited caps as the default setting.

Equally, AI-assisted resolution upscaling can’t turn a blurry mess into something that looks pristine. Arm and Qualcomm both give examples of turning low-res 540p images into pristine 1080p, which is perhaps a little fanciful if they’re insinuating indistinguishable AAA quality graphics.

However, smartphones certainly do benefit from small screens that don’t require high resolutions to look pin-sharp. 720p, for instance, looks crisp on a 6-inch panel, especially when in motion. Upscaling from 720p to 1080p, or the often-native 1440p resolution of a mobile panel, would look really impressive and even allow scaling to larger screens with negligible performance impact.

Mobile’s biggest hurdle here doesn’t appear to be the technology — at least if Arm and Qualcomm’s demos are anything to go by. The bigger problem is likely to be getting these technologies into our hands and supporting them in the games we love to play most.

There are two ends to the problem here: the first is that hardware supporting these new AI features is reserved for upcoming GPUs or those currently at the very top end of the market. That means the adoption market is starting from a low point, making it less likely that game companies will invest in the technology when the biggest profits come from mass-market titles. As such, it’s going to take time for any features to arrive in the mainstream.

Likewise, Arm’s and Qualcomm’s tools appear to be heavily integrated into Unreal Engine, which, while a very popular choice, isn’t the only option for game developers. Other development platforms may eventually support these vendor-specific SDKs, but that’s a workload that has to justify the cost. Mobile technology, particularly in the realm of AI, consistently suffers from per-vendor APIs and toolchains that make mass adoption of great technologies very difficult. Hopefully, the gaming benefits are far more tangible.

This isn’t to be overly pessimistic — adoption of any new technology takes time after all, and the benefits of frame generation and resolution upscaling are very promising for mobile devices, even more so than high-powered PCs and consoles. If the technology lives up to its potential, there’s no reason mobile gaming engines shouldn’t quickly adopt it.

Personally, I can quite happily live with some “fake” frames and upscaled image quality if it means my smartphone runs high-fidelity titles smoothly, with lower battery draw and less heat in my hand.