ComputerBase blind test shows DLSS 4.5 preferred over FSR and native in all six games

DLSS 4.5 first, native second

ComputerBase has published the results of a community blind test that ran for two weeks. Readers voted on which image looked best in six games at Ultra HD output resolution, comparing NVIDIA DLSS 4.5 Quality, AMD FSR Upscaling AI (FSR 4) Quality, and native rendering with TAA.

Across 6,747 total votes, DLSS 4.5 was selected most often at 48.2%. Native + TAA placed second at 24.0%, FSR Upscaling AI at 15.0%, and 12.8% of votes landed on “no visible difference.”

Source: ComputerBase

By game, DLSS 4.5 led every poll: Anno 117 (50.1% DLSS, 22.8% native, 16.5% FSR, 10.7% no difference), ARC Raiders (47.4%, 27.3%, 13.8%, 11.5%), Cyberpunk 2077 (34.4%, 32.4%, 10.6%, 22.6%), Horizon Forbidden West (56.3%, 19.4%, 11.7%, 12.6%), Satisfactory (60.9%, 15.1%, 12.4%, 11.6%), and The Last of Us Part II (40.9%, 25.9%, 25.3%, 7.8%).

The test used side-by-side comparison videos and asked participants to pick a single “looks best” option per title. ComputerBase also notes it has generally recommended using modern upscaling solutions because it considers image quality comparable or better while also raising performance, but this poll focused only on perceived image quality.

One thing must be clear when looking at the results of the reader blind test. The results only show how frequently each option was rated as having “the best image quality.” What the results do not show, however, is what was perceived as the second and third best. Neither the wording of the question nor the surveys themselves were designed to reveal this. Therefore, the conclusion “FSR Upscaling looks worse than native” is invalid.

— ComputerBase (translation)

ComputerBase also adds a key limitation: these results do not show second- or third-place preferences. It says it is not valid to conclude “FSR looks worse than native” from this data alone, because the survey structure only captured first-choice votes.

Source: ComputerBase