Embedded processing giant Arm has announced what it claims to be “an industry first:” dedicated artificial intelligence-accelerating neural coprocessor cores in its upcoming graphics processing IP, which aims to deliver “PC quality, AI-powered graphics” to mobile devices.
“On-device AI is transforming workloads everywhere, from mobile gaming to productivity tools to intelligent cameras. This is driving demand for stunning visuals, high frame rates and smarter features – without draining battery or adding friction,” claims Arm fellow Geraint North. “Announced today at SIGGRAPH, Arm neural technology is an industry first, bringing dedicated neural accelerators to Arm GPUs from 2026.”
“This,” North continues, “takes the performance of GPUs for graphics rendering to new heights, delivering up to 50 percent GPU workload reduction for today’s most intensive mobile content, starting with mobile gaming. And this is just the beginning – the availability of this new technology lays the foundations for the industry to deliver even more on-device AI innovation in the future.”
Arm is no stranger to AI acceleration, having its own Ethos IP family that is designed to be paired with its Cortex microcontroller and microprocessor cores. Today’s announcement is different, though: it’s not about a discrete accelerator but about adding neural coprocessors directly within a graphics processing unit, in the same way as NVIDIA has with its RTX family of desktop and laptop graphics cards and their Tensor cores.
The Arm neural accelerators, which sit in the GPU core alongside the existing execution engine, will, the company has said, initially focus on accelerating 3D rendering tasks — primarily in gaming, where they can be used to upscale a lower rendering resolution to the higher resolution of a modern smartphone or tablet display without the performance hit you’d expect by rendering directly at a higher resolution.
“NSS [Neural Super Sampling] delivers the potential for upscaling from 540p resolution to 1080p at a cost of 4ms per frame,” North claims, “while delivering near-native quality. Developers can save up to 50 percent of the GPU workload compared with rendering the full frame using traditional methods, and either bank that saving to reduce the overall power consumption of their game, [or] spend it on delivering a higher frame rate or increasing the quality of the visuals. With NSS, developers can use AI to preserve surface detail, lighting, and motion clarity, giving them the flexibility to balance visual fidelity with energy efficiency depending on their game’s needs.”
Arm’s plan will put a dedicated neural accelerator alongside the existing execution engine in GPU shader cores from 2026. (📷: Arm)
As with NVIDIA’s Tensor cores, Arm has more in mind for its new IP than simple upscaling: the company has confirmed plans to release Neural Frame Rate Upscaling (NFRU) which promises to double frame rates without doubling the corresponding load on the GPU, along with Neural Super Sampling and Denoising (NSSD) which provides real-time path-tracing on mobile devices — both of which will be ready for developers ahead of hardware availability, North claims.
Arm is providing developers with access to a software development kit which includes an Unreal Engine plugin, PC-based Vulkan emulation, profiling tools, fully-open upscaling models, and open Arm ML (Machine Learning) extensions for Vulkan. No release dates has yet been provided for the hardware itself, though, beyond the promise that the hardware IP will launch some time in 2026.
More information, along with links to download the development kit, can be found on the Arm community site.