FSR was a good first effort but it relied on spatial upscaling using only the data present within a single frame as input. It applied an upscaling and customized Lanczos sharpening filter, a relatively low overhead way of doing upscaling. However, artifacts were visible at higher upscaling factors and it did nothing to combat aliasing. DLSS uses temporal upscaling, with data from multiple frames combined with motion vectors and the depth buffer and feeding that into an AI enhancement algorithm. DLSS 2.x looked better than FSR but was restricted to Nvidia’s RTX graphics cards.