NVIDIA and AMD are actively developing their technologies for intelligent image scaling, which can significantly improve performance in games. This is especially true for the latest versions of NVIDIA DLSS Frame Generation and AMD Fluid Motion Frames (FMF), which have a significant impact on the final fps. As it turned out, the work of both technologies can be combined to obtain even more impressive quantitative results.
For example, practical tests of QuasarZone show that the average frames per second in Cyberpunk 2077 can be tripled by using both NVIDIA and AMD graphics cards. The colleagues’ test bench simultaneously included a GeForce RTX 4090 24 GB and a Radeon RX 6600 8 GB with a monitor connected to it. The software settings were made in such a way that the RTX 4090 was initially used for rendering with DLSS Frame Generation, and then AMD Fluid Motion Frames was additionally used on the RX 6600 to generate even more frames.
As a result, the initial performance of Cyberpunk 2077 in 4K mode tripled, with the average number of frames per second increasing from 72 to 209 fps.
In Call of Duty: Modern Warfare III and Ratch & Clank: Rift Apart, performance almost doubled, but for the most part, performance increased when AFMF was activated. The Starfield game also responded actively to the inclusion of DLSS + AFMF, and it is obvious that the authors were experimenting even before the official DLSS 3 support appeared in this project.
The idea of combining the efforts of both technologies looks quite interesting, at least in theory. However, there are points that are not covered in the article. It concerns the actual quality of the picture, which is transformed twice in the process. Also, in certain games, it is even possible to reduce the rate of rare events (1%), which may even worsen the overall perception. There are also questions about the potential increase in input lag, which is relevant for modes even with single frame generation.
Nevertheless, at the level of a concept and a separate study, the DLSS+AFMF topic looks promising. It is unlikely that it will come to mass practical application – there are too many additional conditions.