Community projects that bring Nvidia's DLSS 5 Neural Rendering to AMD Radeon GPUs are becoming quicker to run and less tedious to install, according to VideoCardz. The work remains unofficial, but recent releases attack the two biggest obstacles to experimentation: performance and per-game setup.
Developer Daniel Blanco released DLSS-NR-on-AMD Alpha 0.3.3 with a claimed 23% speed improvement over the prior build, then followed it with Alpha 0.4.0 and another claimed 42% gain. VideoCardz says the project targets Radeon RX 7000 and RX 9000 cards in DirectX 12 games that use FSR. Those percentages come from the developers rather than independent benchmarking, so they show the pace of the project rather than a guaranteed uplift for every card or game.
The separate AMDNR Launcher simplifies installation. Instead of manually copying and replacing DLL files for each title, users can choose a game and runtime, install or remove the needed components, and update them from one interface. The launcher supports Blanco's runtime and Kien's open-source HIP runtime while keeping OptiScaler options for FSR, XeSS, frame generation, Ray Regeneration and Model Interleave.
Performance is still the practical constraint. The developers reported that one neural-network pass on an RX 7800 XT at 1440p with FSR Quality fell from roughly 73 milliseconds to 52 milliseconds. That is meaningful progress, but the cost of the neural pass can still make ordinary gameplay less attractive than a technical demo or photo-mode experiment.
Both projects are community efforts and are not affiliated with AMD or Nvidia. The launcher also lists experimental handheld support, including testing on the ROG Ally Z1 Extreme, but that should not be confused with official DLSS support for Radeon hardware.
