4TF of GPU power and because it's only GPU, only graphics are gonna be affected... What's not to be understood? Games complexity will remain the same as if S didn't exist
Are we still stuck at a time when GPU’s only processed pretty pixels only and not essentially being a giant vector unit co-processor for intensive general computation?
XSS’s GPU is not only powered by less CU’s than XSX, but it is clocked lower and apparently also has a single shader engine (though I do not have a hard proof of the latter):
Less shader engines would mean less ROP (RB in AMD wording, affects pixel fillrate/blending), Rasteriser and Primitive processing unit, and possibly less async compute engines to dispatch and prioritise async work (the latter is not a given, it was just common for AMD to match the ACE’s count to the number of Shader Arrays which in this case it would go from 4 to 2).
On top of that you have a lower clockspeed (~1.825 GHz for XSX and ~1.565 GHz for XSS). This would reduce the peak performance of all the shared units like the Geometry Engine as well as the other HW that does not get counted in the 4 TFLOPS number. This affects geometry processing too.
Still, you have a point, limited RAM with about 5.5 GB less available RAM for games compared to XSX (
8 GB on XSS, OS taking up all of the 2 GB slow, 56 GB/s, portion of RAM) is going to be a limiting factor first.
Then you would XSS dropping RT or cutting framerate if RT had to be kept (and we are seeing resolution in DRS dropping to less than 720p in launch window software already). Cutting framerate or dropping RT was not the “same experience as XSX but at 1440 (or even 1080p)” as it was stated btw.
Some more data about the specs here:
https://www.eurogamer.net/articles/xbox-series-s-review-digitalfoundry (of course the article rational part and hope springs eternal emotional sing-off at the end is there, but it is a Leadbetter article
)