Overview: FSR 4 reaches older RDNA cards via a mod
AMD’s FidelityFX Super Resolution 4 (FSR 4) has generated curiosity far beyond its officially supported hardware. After AMD accidentally released the source code, hobbyists and modders compiled DLLs that inject FSR 4 into GPUs not originally supported, including RDNA 2 (Radeon 6000) and RDNA 3 (Radeon 7000) systems, as well as various APUs and handheld configurations. A German tech outlet, Computer Base, took this a step further by conducting an in-depth evaluation across more titles to determine whether the mod delivers a meaningful increase in image quality on older architectures and whether any practical trade-offs undermine its use.
What the tests reveal about image quality
The core finding is clear: FSR 4 delivers a perceptible improvement in image clarity over FSR 3.1, even when run on unsupported hardware via the community-built mod. The team notes that the upscaler, while not matching the native RDNA 4 FP8 acceleration, still sharpens textures and fine details in many scenarios. In motion, however, visual artefacts become more noticeable, particularly in vegetation and delicate structural elements. In practice, this means players can experience crisper imagery and better overall sharpness, but not without occasional instabilities that can distract during fast action or lush, leafy scenes.
It’s important to underline that the improvements are not uniform. The mod’s performance and stability vary by title, resolution, and the degree to which a game’s anti-cheat or anti-tamper measures resist injection techniques. Some games run without incident, while others crash or fail to launch if certain hacks are attempted. Still, Computer Base found a significant subset of titles that do render with FSR 4 through the mod, suggesting a feasible path for enthusiasts who want to experiment with the technology on older GPUs.
Performance costs: a meaningful but not crippling hit
Where FSR 4 shines in image quality, it costs in frames per second. In their measurements, Computer Base reports a 9% to 12% FPS drop when using the Performance mode on a RX 7800 XT and a 10% to 13% decline on a RX 6800 XT, compared with using FSR 3.1 or native rendering at higher fidelity. In contrast, a newer, officially supported configuration—such as the RX 9060 XT with FP8 acceleration—experiences a far smaller 3% to 5% FPS hit when enabling FSR 4 in its proper context. The takeaway is that the modded path to FSR 4 carries a noticeable performance tax, which may be acceptable for some players seeking improved visuals but not viable for others sensitive to frame rate drops.
RDNA 2 vs RDNA 3: is there a clear winner?
The report challenges a straightforward assumption that RDNA 3 would always underperform or outperform RDNA 2 in this scenario. In practice, RDNA 2 did not appear categorically worse than RDNA 3 in these tests, and the team speculates that RDNA 3 may not be exploiting the expected WMMA Int8 matrix-multiply instructions as efficiently in the upscaling path. This blurs simple expectations about hardware advantage and underscores how software optimization, even in a hack, can yield nuanced results across generations.
Compatibility and practicality for everyday users
One recurring theme is compatibility. While 14 of the 18 tested games worked with FSR 4 via the mod, some titles demanded alternate hack methods, and other games crashed under certain injection approaches. The practical implication is clear: this is not a one-click solution. It requires technical know-how, careful testing, and a willingness to deal with potential instability and game-specific quirks. If you value a plug-and-play upgrade, this hack probably isn’t for you. If you enjoy tinkering and want to push older hardware beyond its official limits, it becomes a more compelling, albeit imperfect, option.
Bottom line: is FSR 4 worth using on older GPUs?
According to Computer Base’s deep dive, FSR 4—when implemented through an unofficial mod on RDNA 2/3 hardware—indeed improves image quality relative to FSR 3.1 and can be a viable path for users with time to tinker. However, the cost in FPS, occasional artefacts, and inconsistent game compatibility temper the enthusiasm. The verdict: it’s feasible and appealing for enthusiasts and niche scenarios, but not a general recommendation for the average gamer seeking a straightforward performance uplift. In other words, this hack demonstrates that FSR 4 can deliver better visuals on unsupported GPUs, but it also highlights why AMD may want to deliver an official, robust implementation to unlock reliable results for a broader audience.
What comes next
As with most emergent tech, the best path forward is transparency and official support. The mod demonstrates demand for FSR 4’s potential and may influence AMD’s future strategy. Until an official, polished release lands, users should weigh the benefits of sharper imagery against the potential stability costs and the time required to manage the setup.
