Developer danielblnc released version 0.8.0 of the DLSS-NR-on-AMD project, bringing NVIDIA proprietary rendering features to AMD Radeon RDNA 3 and RDNA 4 architecture graphics cards. The unofficial software mod introduces Super Resolution 4.5, frame generation, multi-frame generation, and ray reconstruction to competing hardware, TechNews 科技新報 reported.
Radeon RX 9070 XT Runs Path Tracing in Cyberpunk 2077
The update marks the first time NVIDIA ray reconstruction operates on an AMD graphics card. The neural network replaces traditional ray tracing denoising workflows to restore intricate scene details. Hardware enthusiasts tested the setup using an AMD Radeon RX 9070 XT paired with an AMD Ryzen 7 9700X processor. They ran Cyberpunk 2077 at 1440p resolution with maximum settings and path tracing enabled. TechSpace 鐵客空間 noted that testers observed significantly reduced noise in car reflections and neon street scenes. The mod relies on custom neural network execution methods, including AMD-specific HIP kernels, custom memory allocations, and targeted game integration layers.
System Requirements and Software Dependencies
Running the modification requires specific operating system conditions and user-provided binaries. The setup demands Windows 11 and AMD Adrenalin 26.1.1 drivers or a newer release. Users must supply their own proprietary NVIDIA dynamic link library files because the project distribution contains no copyrighted NVIDIA code. Frame generation functions additionally require specific GPU spoofing procedures and game-side support. RDNA 2 architecture graphics cards currently receive only experimental support as development continues. A newly integrated NeuralHub interface centralizes installation routines, automatic game detection, and feature configuration.
| Component | Requirement |
|---|---|
| Operating System | Windows 11 |
| Graphics Driver | AMD Adrenalin 26.1.1 or newer |
| Supported Architectures | RDNA 3, RDNA 4 (RDNA 2 experimental) |
| Required User Files | Legitimate NVIDIA DLSS library binaries |
Broader Technical Implications for GPU Ecosystems
The software development demonstrates that advanced neural rendering limits stem from execution environments rather than rigid hardware locks. The project proves that competitive hardware can process alternative instruction pipelines when developers supply dedicated integration layers. Compatibility, image quality, and framerates continue to vary depending on individual titles and specific hardware configurations.
