DLSS 4.5 Arrives on Radeon RX 9000 Series
NVIDIA DLSS 4.5 technology is experimentally coming to AMD Radeon RX 9000 series cards with the d4r project. Performance increase is aimed at with FP8 support in the RDNA 4 architecture.
The latest update to the d4r project, published by developer “countervolts,” has enabled the experimental operation of NVIDIA DLSS 4 and DLSS 4.5 technologies on AMD Radeon RX 9000 series graphics cards. This project, which integrates NVIDIA’s official DLSS libraries into AMD hardware via a Linux and Proton-based software stack, specifically aims to take advantage of the native FP8 computing capabilities of the RDNA 4 architecture.
This work, which is not yet ready for daily use and is in the proof-of-concept stage, is valued as a valuable technical step for AMD graphics card owners to be able to use NVIDIA’s proprietary landscape scaling technologies.
- The d4r project has managed to port NVIDIA DLSS 4 and 4.5 to AMD RDNA 4 architecture-based RX 9000 series cards.
- Thanks to the native FP8 computation support of the RDNA 4 architecture, performance loss in DLSS processes is minimized.
- The system runs in Linux and Proton environments by translating CUDA workloads via ZLUDA.
- The current version offers unverified performance and does not have the necessary stability for daily use.
DLSS 4.5 Technology is Compatible with RDNA 4 Architecture.
While NVIDIA’s DLSS technology was traditionally designed to work only on GeForce RTX series cards, the d4r project continues to push that boundary. Newly released updates can create dedicated cores, especially for cards based on the RDNA 4 architecture, such as the RX 9060 and 9070.
The most striking aspect of this integration is that the RDNA 4 architecture can natively process FP8 calculations used in contemporary DLSS models without needing to convert them.
This capability can help reduce the load on the processor, resulting in smoother performance. However, it should be noted that the project is still in its early stages. While the developers list cards like the RX 9070 XT and RX 9070 among the targeted hardware, real-time testing is currently very limited.
Development Process is Progressing in a Controlled Manner
The working principle of the d4r project consists of making NVIDIA’s nvngx_dlss.dll file executable on AMD GPUs with the help of ZLUDA. The project attempts to compensate for the performance loss experienced during translation by adding customized cores for RDNA 3 and RDNA 4 architectures. This method, currently tested on the RX 7700 XT model, has helped to achieve higher frame rates compared to pure translation processes.
However, this work is not a complete alternative to AMD’s own FSR (FidelityFX Super Resolution) technology. The unverified performance data and experimental nature of the software make it a tool that appeals more to technical enthusiasts than end users. Whether this compatibility will become more stable with future updates remains to be seen.
What do you think about this experimental development of NVIDIA’s DLSS technology running natively on AMD cards? Do you think projects like this should put pressure on AMD to develop its own FSR technology, or would the combination of different technologies be more beneficial for the gaming world? Share your opinions with us in the comments.
