AMD has released the new version of their upscaling solution, FSR 2.0. This time around, it uses temporal data (making it technologically more advanced than version 1.0 and more similar to Nvidia's DLSS), but without using any kind of specialized hardware. The latter enables it to support a much wider range of GPUs, such as AMD RX and Nvidia GTX cards. So far, it is supported in Deathloop, where the following can be observed: It is a massive improvement over FSR 1.0, the image quality is a lot better and it's quite usable even at lower resolutions (1.0 works okay image quality wise at 4k, but things go noticeably downhill with the resolution). It also does not suffer a performance penalty compared to the older solution, in fact, the highest quality preset has been marginally, but consistently faster than the FSR 1.0 counterpart. Compared to DLSS, it's similar, but slightly different. FSR 2.0 produced an image that was consistently sharper and smoother (less aliased) than DLSS and, unlike the Nvidia technology, it also suffered less (or not at all) from ghosting. DLSS on the other hand seemed to preserve slightly more fine texture detail and the motion (especially at lower resolutions) seems a little smoother. The technology seems to work equally well on AMD and Nvidia hardware, providing identical image quality and a similar performance uplift. Compared to DLSS, on cards that support both, it seems provide a similar, but slightly lower performance.
