Intel Arc GPUs successfully run NVIDIA DLSS 5 via AI-assisted mod

A developer known as Uzbekunknown has successfully implemented a port of NVIDIA’s DLSS 5 Neural Rendering technology on Intel Arc 140V integrated graphics. The project, titled ‘dlss-nr-on-intel’, does not use NVIDIA’s proprietary CUDA or NGX libraries. Instead, it is an independent reimplementation of the 71-block U-Net neural network pipeline designed to run on Intel’s Xe2 architecture via the Vulkan API.

The developer utilized AI coding assistants, such as Claude and OpenAI Astra, to build the project. The implementation works by extracting model weights from original NVIDIA files and converting them to FP16 to run on Intel’s XMX engines.

While technically significant, the project is currently an experimental proof-of-concept rather than a consumer-ready product. Performance is extremely low; tests on titles like Tekken 7 and Mortal Kombat 1 showed frame rates as low as 3 to 10 FPS at low resolutions. The developer noted that the current state of the software is better suited for capturing static images in photo modes rather than active gameplay.