An unofficial mod has brought NVIDIA’s DLSS 5 neural rendering to AMD RDNA 4 graphics cards, overcoming hardware barriers and pushing frame rates in Cyberpunk 2077 up to 50 FPS by late September 2026 after rapid performance updates from developer Daniel Blanco.
Modders have once again cracked hardware boundaries, successfully porting NVIDIA’s advanced DLSS 5 neural network rendering onto non-native hardware. The custom utility, known as DLSS-NR-on-AMD, makes the proprietary NVIDIA technology work on AMD Radeon RDNA 4 architecture, expanding compatibility well beyond its intended GeForce ecosystem via a mod published on GitHub.
While NVIDIA’s official implementation is restricted to RTX graphics cards—and notoriously exacts a heavy toll on performance, causing top-tier RTX 50-series hardware to lose roughly half their frames—getting the neural renderer running on an AMD card required breaking entirely new software ground.
From Unplayable to 50 FPS in Cyberpunk 2077 on the Radeon RX 9070 XT
When the mod first surfaced in early September 2026, running the tech on AMD’s hardware was more a technical novelty than a practical solution. Early tests conducted by the YouTube channel Ancient Gameplays showed the RX 9070 XT dropping well below 20 FPS in Cyberpunk 2077, making the experience suitable for little more than capturing static screenshots.
However, developer Daniel Blanco—also identified in repositories as danielblnc—rolled out a rapid sequence of alpha updates throughout September 2026 that dramatically altered performance. Version 0.3.0 introduced a crucial pipeline optimization called the Pre-Pass method, which runs the neural rendering model before applying FSR spatial scaling, rather than forcing the neural network to process the image at full-screen resolution. In the benchmark Cyberpunk 2077 on the video card Radeon RX 9070 XT, this optimization allowed frame rates to increase from the original unplayable 13-14 FPS to a stable 30-35 FPS with minimal compromises in image clarity.
By version 0.4.0, performance gains had compounded significantly. Testing in 3440×1440 resolution on a Radeon RX 9070 XT demonstrated frame rates climbing from roughly 30 FPS up to 50 FPS. Subsequent updates like Alpha 0.4.1 added an 8% performance increase relative to version 0.4.0, alongside a new priority queue system capable of extracting up to 9% more performance during high gaming loads.
Software Architecture and Game Compatibility Across the Radeon Lineup
The utility relies on a custom executable paired with a specific build of NVIDIA’s DLSS DLL (version 310.8.0.0), requiring a Windows 11 environment and a DirectX 12 game that supports FSR, with Vulkan support promised in a future release. Unlike older visual mods that disguised FSR as DLSS, the project is described in its repository as a written-from-scratch full implementation of a neural network runtime environment for AMD architectures.
The compatibility list has expanded beyond flagship adapters like the Radeon RX 7900 XTX and RX 9070 XT to include the Radeon RX 9060 XT, with technical support also reaching certain RX 7000 models according to project trackers. Developers have integrated fixes for RE Engine projects including Resident Evil and Pragmata, resolved conflicts with FSR 3 algorithms, ironed out bugs in Marvel’s Spider-Man Remastered and Grand Theft Auto V Enhanced, and adjusted HDR exposure during frame buffer processing.
Technical Hurdles and Hardware Limitations Ahead
Despite the rapid optimization, running third-party custom libraries on non-native hardware carries operational warnings. Tech specialists point out that the utility should be restricted to single-player experiences because multiplayer anti-cheat software can easily misinterpret the injected files as unauthorized system interference.
Current code relies on a direct brute-force reconstruction of neural network logical connections, but the developer is actively redesigning the runtime architecture to migrate NVIDIA’s Tensor Core algorithms directly into the ray tracing acceleration units of AMD’s RDNA 4 architecture. If those upcoming deep optimizations succeed, performance could potentially double again, bringing AMD hardware closer to GeForce RTX-level reconstruction parity without relying on proprietary hardware monopolies.