Regardless of tooling. Logically this shouldn't be happening (for any comparison).
The PS5 is a 10 TF machine with a 3.5 ghz CPU
The XBSX is a 12 TF machine with a 3.8 ghz CPU
They're directly comparable as they're both RDNA2+Zen2 based. Scratching my head right now. It doesn't matter how minimal the difference is based on specs. The XSX should technically be offering better performance here.
There are loads of possible reasons, the issue is we don't know which is correct.
Firstly the GDK is not as close to the metal as the PS API, the hardware model is abstracted more. Furthermore if there are inefficiencies in the abstracted model?
Perhaps it is just immature GDK, but that doesn't fit so well as the case we are seeing has specific performance characteristics.
Although both consoles use the same CPU, the architecture is different in each case. Now the XSX being faster means that, all things being equal, it should be faster! But all this are not equal. It may be more efficient caching or offloading more tasks make the PS5 less CPU bound, it may only apply depending on SMT mode?
The mistake people have made from the start is confusing paper spec "models" for the actual machines, they are not the same. The test of every model is how it compares to the empirical data of reality.
XBox fans were far too sure of XSX superiority.
And to be clear, I expect XSX to regain the lead in performance in the future, will probably take some months though.