AMD FSR4 and Multi Frame Generation Explained With New Ratio Based Controls

FSR4 machine learning upgrades appear closer as AMD expands frame generation controls without widespread real world deployment.

Hardware by Shinji Okazaki on Apr 22, 2026

The technologies for frame generation and upscaling in AMD are constantly being developed, and according to the latest updates, new capabilities will be rapidly introduced.

Changes in documentation tend to focus on adjustable frame generation ratios, suggesting it may no longer be limited to fixed implementations. Although there is positive momentum in various aspects, releases are still few and far between, with declines between announcements and actual availability.

AMD FSR4 and Multi Frame Generation, Explained With New Ratio Based Controls, NoobFeed

Preliminary indications of Ratio-Based Frame Generation

We hear that AMD is about to release a multiframe generation. The most recent news emphasizes changes to the GPUOpen documentation, specifically the introduction of adjustable FSR frame generation ratios. This implies that we can possibly regulate frame-generation behavior, or even permit dynamic control, as in existing dynamic multiframe generation systems.

Simultaneously, it seems that AMD is creating a lot under the hood, but not releasing fully-fledged features. With FSR Redstone, few games take full advantage of its capabilities. Features such as ray regeneration and radiance caching are hardly implemented. There is radiance caching and only two ray regeneration. Much of the real work is still upscaling and simple frame production.

Seldom Rollout of Features across Games

We still experience a lapse in announcements and releases. FSR Redstone has been mentioned months before, but when it is featured in the titles of such games as Call of Duty: Black Ops 7, it is not in full use. Radiance caching remains sparse, and significant use in a variety of games has not occurred.

On top of that, we have yet to truly have multiframe generation. AFMF3 has been leaked and changed in interfaces, but it is not officially released. Older GPUs are also not supported by FSR4, except where previously leaked to be compatible. Regeneration of rays is still sparse, and even games that do regenerate them do not exhaustively use path tracing.

The state of AFMF and FSR features at the Moment

We also observe how the AMD Fluid Motion Frames react to the absence of suitable in-game data. It goes to a standard mode and offers options such as motion blur or FSR upscaling. This indicates that the system continues to require extensive integration of games and correct configuration.

At this point, multiframe generation has not been officially offered, FSR4 is not widely supported on older hardware, and AFMF3 is still lacking. Even games that use Vulkan lack proper support for FSR4, which limits their use across a variety of games.

New Interface Controls and Expansion of Features

Examining the new materials for ADLX 1.5, we may notice increased control over frame generation. The interface has been updated with links to check support, turn the feature on or off, view existing ratios, available ratios, and set a ratio.

This is the control that we have been lacking. You can view what features are on and set them on or off. This would bring AMD closer to providing similar visibility and control to other ecosystems, where overlays provide preset usage.

Move Towards Frame Generation based on Machine Learning

We also learn that future releases of Adrenalin can automatically enable games that use FSR3.1.x to be upgraded to machine learning-based FSR4 frame generation in DirectX12 titles. The latter is already provided by a toggle in the software, which enables us to inject FSR4 in supported games.

Nonetheless, FSR4 is not yet fully supported in APIs such as Vulkan. This creates an inconsistency, particularly when playing some titles that do not support the feature.

FSR4 is developed to support RDNO4 hardware, and older GPUs use analytical fallback algorithms that effectively emulate FSR3 frame generation. The question of whether frame generation based on ratios will rely solely on machine learning or involve analytical techniques remains open.

Multi-Frame Generation and its limitations Experimentally

We have already witnessed experimental applications in the form of mods, in which multiframe generation applies to FSR3. It is functional, but it adds artifacts, particularly at high ratios, such as 5x or 6x. Latency and visual quality are also problems.

As a result, FSR4 methods based on machine learning may be required to keep latency low and artifacts to a minimum as frame generation is further scaled.

AMD FSR4 and Multi Frame Generation, Explained With New Ratio Based Controls, NoobFeed

Hardware Support and Future Uncertainty

Still, it remains unclear what GPUs will have ratio-based frame generation. It will probably only apply to RDNA4 GPUs, like the RX 9000 series. It is not clear whether older hardware will be given such controls using analytical methods.

We already have a glimpse that tools such as DLSS enablers and lossless scaling indicate that AMD hardware is capable of implementing tionpower remain the determinantsrate and the GPU's power continuecontinue to be key determinants of performance, latency, and image quality.

Final Thoughts

At this stage, AMD seems to have several features in development, though not offering whole solutions. As other companies continue to update and expand their technologies, AMD's releases are minimal.

We await further guidance. Features such as multiframe generation, AFMF3, and wider FSR4 are likely but have not yet been confirmed, with schedules and availability still unknown.

Also, check our other hardware articles: