I'm calling bull on this one. The ps5 already has a memory bandwidth advantage because the XSX has to depend on memory interleaving thereby halving the effective available bandwidth. The texture pop-ins in the Halo Infinite demo are testament to the fact that the XSX's slower SSD speeds coupled with interleaved memory again is causing extra problems for dev.s aiming for 4k/60fps on any modern game, things will improve no doubt as development tools mature further. But right out of the gate, MS's focus on doubling the TFLOPs at the expense of bandwidth has surely not done it any favours in terms of actual performance versus the leaner but nimbler ps5 architecture. Also two more CUs aren't something that would close a 33% physical CU deficit as easily, parallelism and higher frequency have very different goals in terms of games performance.
And with the ps5 thermals firmly locked in place, adding any more CUs would inversely reduce the maximum consistent frequency across the entire GPU, no matter what kind of smart shift the ps5 may be able to pull off.
And lastly, MS has nothing to show for its TFLOPs advantage at ALL at this point, with Sony's ps5 exclusives collectively looking wayyyy more next gen than anything MS has shown up to this point.
Upping the performance at this stage would be less financially desirable than a mid-cycle pro refresh as well.