XeSS performs worse than native rendering on AMD's Ryzen 7 5700G

pcgamer.com:

Yesterday I looked at the performance of Intel XeSS upscaling tech running on an Nvidia RTX 2080 Ti(opens in new tab). While not quite up to the performance of DLSS, it wasn't far off, and the image quality was impressive. A promising start for Intel. Especially given that it runs on not just Intel's graphics cards, but AMD's and Nvidia's too.

In order for other cards to support the feature, they need to support DP4A(opens in new tab). The GeForce RTX 2080 Ti does, which is used for the upscaling algorithm and is the fallback for non-Arc GPUs. Arc GPUs, such as the upcoming Arc A770(opens in new tab), bring bespoke XMX acceleration to do this work even faster. 

There is another fallback for cards that don't support DP4A though, and that is INT24. You'll find INT24 support in pretty much every modern GPU, and that includes integrated GPUs, you know, the kind you'll find in AMD's APUs.

However, while this third tier of support ensures that XeSS works on plenty of systems, there are questions about how fast it is. In fact, after testing the various quality settings of a Ryzen 7 5700G system, the results are less than impressive. You're looking at performance that is actually worse than running at native. In other words, you'll actually see better performance turning XeSS off. 

Using Shadow of the Tomb Raider as an example, which is one of two games that current support XeSS, and using the highest graphics settings at 1080p, I witnessed just 20fps with no upscaling. The fastest setting for XeSS is the Performance quality setting, which actually saw the performance drop down to 18fps. The Quality preset dropped this even further, down to 15fps. 

Running Shadow of the Tomb Raider with the Highest settings is good for