AMD's FSR 3.1 upscaling tested: improved over FSR 2 but DLSS and XeSS are still ahead

FSR 3.1 improves but still lags behind DLSS and XeSS in quality and stability, particularly in dynamic scenes and objects without good motion vectors.

: AMD's FSR 3.1 has notable improvements in image reconstruction and stability over its predecessor FSR 2, including reduced flickering and pixelation. However, it still suffers in areas like motion vectors, particle effects, and RT reflections, lagging behind Nvidia's DLSS and Intel's XeSS. Core issues remain unresolved, prompting a need for a new approach to remain competitive.

AMD's FSR 3.1, tested in games like Ratchet and Clank: Rift Apart, shows significant improvements over FSR 2, particularly in terms of image stability and reduction of flicker and pixelation in moving objects. However, issues such as aliasing, flickering in RT reflections, and poor handling of particles still persist, making it less competitive compared to Nvidia's DLSS and Intel's XeSS.

FSR 3.1's frame generation now decouples from upscaling, allowing combination with other image reconstruction techniques like DLSS and XeSS, which is a beneficial move for users with older GPUs. Despite these advancements, the technique still struggles with dynamic scenes and objects lacking motion vectors, where it fails to match the quality and stability of its rivals.

The ongoing issues, particularly with moving objects and particles, suggest that FSR might need a significant change in approach, possibly exploring machine learning methods. Both Nvidia and Intel utilize such techniques effectively, and while AMD has made progress, more substantial changes seem necessary to close the quality gap and remain competitive in the upscaling space.

upscalingFSRImage Reconstruction