When it comes to upscaling technologies, NVIDIA’s DLSS 2.x is considered the industry gold standard. Combining a time frame accumulator with a neural network produces images that look better, if not on par with native. Intel designed its XeSS scaling technology by roughly copying NVIDIA’s approach. AMD, on the other hand, kept things simple by leveraging a standard temporal upscaler with minor tweaks for FSR 2.0. In this post, we compare DLSS to TSR in Unreal Engine 5 (Going through: Perseverance).
Epic’s Unreal Engine is the most widely used game engine, making it a preview of upcoming gaming technologies. DLSS and TSR produce near perfect images, there are slight differences between the two. These close-ups highlight these intricacies. In the comparison above, DLSS is slightly better than TSR, retaining more detail along the branches and left tower. TSR is softer, which results in better anti-aliasing but loses some detail in the process.
The same result can also be seen in the close-up above. The bench under the tree is sharper with DLSS but TSR produces a smoother image along the edges.
With the performance preset, the differences between the two are even starker. DLSS retains most details even at the performance preset while TSR loses a bit around fine meshes like tree branches, metal grills, and banners.
Overall it looks like NVIDIA’s DLSS 2.x will be the leading upscaler in the gaming industry for the moment. Intel’s XeSS could change that, but the limited market share of Arc GPUs will be a major hurdle for Team Blue.
Going through: Perseverance.