AMD FSR 2.0 goes open source and hopes for greater diffusion

AMD FidelityFX Super Resolution, also known by the acronym of FSR, is a technology introduced with the Radeon RX 6000 series of graphics cards that is proposed as an "open" alternative to NVIDIA's Deep Learning Super Sampling (DLSS), aiming to offer improvements performance thanks to sophisticated image upscaling algorithms. FSR uses temporal upscaling, which requires additional data from motion vectors, depth buffers, and color buffers to generate quality images.

Recently, the Sunnyvale-based company introduced version 2.0, which provides the possibility to choose between four different modes (Quality, Balanced, Performance and Ultra-Performance) to best suit various needs. Unlike DLSS, FSR can be used on a wide range of products, including cards made by other manufacturers such as rival NVIDIA.

| id: th_hardware_d_mh2 "); }
Unlike NVIDIA DLSS, AMD's FSR 2.0 does not consider machine learning as a pre-requisite to achieve good image quality, so there is no AI algorithm for recognizing shapes or objects within a scene , but it is only used to combine the previous frames to generate the upscaled image.

However, this has the undoubted advantage of not needing specific hardware units (such as the GeForce RTX Tensor cores) and results are quite good (although not at the level of DLSS), as also evidenced in a recent test with God Of Wara>.