AMD ready to catch up with NVIDIA: Multi Frame Generation finally arrives on Radeon

A detail spotted in AMD’s software tools ignites the powder, Multi Frame Generation, the technique which allowed NVIDIA to take a head start on fluidity, seems about to arrive on the Radeon side.

The signal is concrete, a new “ratio†option linked to the generation of images appears in the FidelityFX ecosystem, which fits exactly with the idea of multiplying artificial images between two rendered images. If you follow the subject, you know why it is sensitive, today AMD remains confined to a generation of images in 2x when NVIDIA is already pushing much more aggressive modes. The index doesn’t guarantee a date, but it is similar to what you see just before a switch, when a feature moves from the lab to kits for developers. And here the question becomes simple, will AMD’s MFG finally close the gap, or just reduce it.

AMD adds an “Upgrade Ratio†option in FidelityFX SDK

The big clue is a new entry in the SDK, with a function with a long name, but a clear meaning, a “ratio†option for generating images. In the associated descriptions, the stated objective is to let people choose an “optimal” multiplier between performance et visual quality. Translated into player language, you are no longer stuck on a single frame generation mode, you can aim for a higher level depending on the margin of your GPU and the rendering of the game.

This point is important for a very practical reason, this type of adjustment is only of interest if AMD prepares several levels, as the competition already does. Today, AMD mainly offers a limited scheme, a single frame generated between two native frames, which corresponds to FSR 4 in 2x. The addition of a “ratio option†suggests a move to higher modes, typically 3 frames generated or more, although the exact figure is not yet publicly announced.

In the daily life of developers, this kind of brick rarely happens “to look pretty†. When an API exposes a ratio parameter, it means testing behaviors, validating artifacts, measuring latency, and adjusting profiles depending on the scene. An engine engineer, “Marc†, sums it up simply, if we give you a ratio selector, it’s because we expect you to use it, otherwise it wouldn’t have passed the screening. In other words, AMD is preparing the ground for a broader deployment of Multi Frame Generation.

NVIDIA DLSS 4.5 et Intel XeSS 3 ont déjà fixé la barre

AMD’s delay can be seen on the technical sheet, NVIDIA has industrialized image generation since DLSS 3 in 2022, then went up a notch with multi-image generation on the series RTX 50. The principle is that you render a native image, then you insert several synthetic images between two renderings. In recent implementations, NVIDIA talks about modes up to 6xwith dynamic adjustment possible depending on the screen refresh rate.

Intel, for its part, has also drawn up an MFG approach via XeSS 3with modes up to 4x on Arc, including on some integrated chips. This changes the perception of the market, because if two manufacturers out of three already offer multi-frame, the third is mechanically considered “late”, even if pure upscaling has progressed. And in heavy games, especially those that push options like path tracing, the race for fluidity becomes an obvious commercial argument.

We still have to keep one nuance, multi-frame, it’s spectacular on an FPS counter, but not free. The more you multiply the images generated, the more you risk defects on rapid movements, particles, HUDs, or micro-inconsistencies on very contrasting scenes. “Marc†says it bluntly, at 6x, you can impress, but you can also reveal weird things about camera rotations. So AMD will be expected on a specific point, visual stability, not just the figure.

What Multi Frame Generation would change for Radeon cards

If AMD really deploys the MFG in its sequel, the most immediate impact will be seen on the sensation of fluidity, especially for 120 Hz or 144 Hz players looking to “stick to” the refresh. The idea of ​​a selectable ratio opens the door to profiles, for example a more conservative mode to avoid artifacts, and a more aggressive mode for competitive games where above all you want a very regular animation, even if the image is not perfect under the magnifying glass.

Another consequence is the implementation on the developer side. If the ratio parameter is exposed in the tools, studios can offer a finer menu, or at least a simple setting, 2x, 3x, 4x, depending on what the pipeline supports. It can also facilitate “dynamic” approaches, where the game changes ratio depending on the load, for example when you move from a corridor to a large open area. The goal is to avoid the roller coaster of FPS without sacrificing everything on the latency.

But there is one point to watch out for, AMD has not yet given a precise public timetable in these technical elements. The clues speak of an “imminent” deployment in the 2026 window, but that doesn’t tell you if it will be activated first on a few Radeon recent ones, or if it will go down wider. And the real judge of the peace will remain the games, if three or four major titles adopt it quickly, the effect will be immediate, otherwise the function will remain a line in a changelog that few people use.