A new unofficial application has emerged, enabling users to run Nvidia's DLSS Neural Rendering technology on AMD Radeon GPUs, even in games that do not natively support it. Developed by a user named danielblnc and hosted on GitHub, the tool allows for the injection of DLSS Neural Rendering features by integrating necessary DLSS 5 .dll files, reportedly sourced from early access versions of titles like NBA 2K27, into game directories.
The process involves downloading the DLSS-NR-on-AMD file and placing it alongside the nvngx_dlssnr.dll file within a game's relevant folder, such as the 'bin/x64' directory for Cyberpunk 2077 on Steam. While initial attempts may be flagged by security software as a false positive, and some users have experienced difficulties in implementation, it is possible to get the feature working, albeit with stability issues that may require launching games directly via their executable files instead of through launchers like Steam.
However, the ability to run DLSS Neural Rendering on AMD hardware comes at a steep performance cost. In testing with Cyberpunk 2077 at 4K resolution with RT Ultra settings, an AMD RX 9070 XT graphics card was reportedly achieving approximately 61 frames per second with AMD's own FSR 4 set to Balanced. Activating the DLSS Neural Rendering feature in-game caused this frame rate to plummet to a mere 9 frames per second, representing an 85% performance reduction.
This drastic performance drop is attributed to the lack of dedicated tensor core silicon in AMD's RDNA 4 architecture, which is designed to accelerate the matrix calculations required by DLSS Neural Rendering. While older Nvidia Turing architecture cards with weaker Tensor Cores also struggle compared to newer RTX series, AMD's shared resources within its Compute Units are not as robust for these specific AI tasks. Early reports suggest some performance improvements have been made since the tool's initial release, but a significant hardware limitation appears to prevent DLSS Neural Rendering from being a practical solution for current Radeon GPU owners.