Introduction: FSR 4 Comes to RDNA 2 and 3 via Mods
AMD’s FidelityFX Super Resolution 4 (FSR 4) arrived with a mix of fanfare and controversy. After the accidental release of its source code, modders created DLL injections that enable FSR 4 on unsupported architectures—specifically RDNA 2 (Radeon 6000-series) and RDNA 3 (Radeon 7000-series), including certain APUs and handheld variants. This opened a path for enthusiasts to trial an upscaling solution that official channels had not yet sanctioned for these GPUs.
What began as a curiosity experiment quickly became a meaningful test for gamers who still run older hardware and want to squeeze more frames from their systems. A number of outlets, including ComputerBase in Germany, stepped in to quantify what FSR 4 actually delivers on non-native platforms. The questions were straightforward: does the image quality improvement justify the performance cost, and how reliable is the experience when you bypass official support?
What the tests found: sharper image, but at a cost
ComputerBase’s findings show a consistent and notable uplift in image clarity when using FSR 4, even when running on RDNA 2 and RDNA 3 hardware via the unofficial route. In practical terms, this means crisper textures, finer vegetation detail, and generally more defined edges in many scenes. However, the improvement isn’t equivalent to the full, native FP8-based implementation available on newer RDNA 4 hardware. The reviewers observed some instabilities, particularly in fine structures and vegetation, and said artifacts can be more visible in motion. This aligns with the broader reality of upscaling tricks: gains in perceived sharpness can come with trade-offs elsewhere.
Compatibility is another hurdle. While several titles ran fine, others required specific hack methods, and some attempts caused crashes. Of the 18 games tested, 14 could run FSR 4, with the caveat that anti-cheat mechanisms and game-specific protections sometimes complicated the setup. This underlines a crucial point for would-be testers: success isn’t guaranteed, and a portion of your library may not benefit reliably from FSR 4 in its unofficial form.
Performance impact: a meaningful FPS drop, even on strong GPUs
From a performance perspective, the numbers matter as much as the visuals. Compared with FSR 3.1, FSR 4 in Performance mode produced a noticeable hit on frame rates. On an RX 7800 XT, observers logged a drop of roughly 9% to 12%. On the older RX 6800 XT, the decline was closer to 10% to 13%. In contrast, the RX 9060 XT—which officially supports FSR 4 through FP8—saw far smaller losses, in the 3% to 5% range, highlighting how much of a delta informal implementations can introduce.
Interestingly, the tests suggested that RDNA 3 did not always behave as expected with the WMMA Int8 instructions, and RDNA 2 wasn’t categorically worse. This hints at a nuanced reality: the underlying hardware differences are not the sole determinants of performance; software-level optimizations, driver interactions, and the specific game engines all play a role in the final experience.
Is it worth it on older GPUs? Practical guidance
Ultimately, the verdict is nuanced. FSR 4 delivers sharper visuals and a better overall image quality in many scenarios when used on RDNA 2 and RDNA 3 via unofficial means. However, this comes with a non-trivial performance cost and a less stable, less uniform experience across titles. For some gamers, especially those with time to tinker and a primary goal of image fidelity, the trade-off may be acceptable. For others seeking a straightforward plug-and-play boost, the risks of crashes and inconsistent results may outweigh the gains.
What’s clear from these investigations is a growing hunger for officialized support: if AMD intends to keep pushing performance-per-pixel at a distance from traditional feature sets, it will need to address both the installation fragility and the variability in frame-time stability that accompanies unofficial hacks. In the meantime, the community can experiment, but with an informed understanding of the potential downsides and a backup plan for reverting to native rendering should issues arise.
Bottom line: a capable but experimental upgrade
FSR 4 on older RDNA hardware is feasible and often improves image quality, yet it isn’t a guaranteed, drop-in upgrade. The observed 9–13% FPS dips represent a meaningful cost in most titles, and some visual artefacts or instability can appear in motion. If you’re curious and prepared to tinker, you may enjoy the sharper look—just don’t expect a flawless, officially supported experience. The broader takeaway: this hack demonstrates the appetite for advanced upscaling on legacy GPUs, while highlighting the need for a more stable, officially supported implementation from AMD in the near future.
