Sony didn't go with this set up just to shave a few seconds off of what a more standard SSD could do to loading times it's going to help stream assets into games faster as well and that's the bigger benefit.
Asset streaming relies on random access of data, though, and load times are a metric of sequential access speeds. One is not actually indicative of the other.
It's like how a SSD with a higher sequential read bandwidth can end up having a much slower random read and write bandwidth than the drive with a slower sequential read bandwidth. For stuff like asset streaming, a lot more is at play.
Also, I think in light of the Until Dawn update (the firmware update, that eliminated load times on the game on PS4), this should also show that you don't necessarily need a hyper-fast SSD considering the PS4 is stuck with SATA II standard:
https://www.thesixthaxis.com/2020/1...h-notes-ps4-loading-time-improved-before-ps5/
So much of the I/O stack in games comes down to how devs organize the code in the game itself, it isn't something that has to predicate itself solely on the surrounding hardware. This is probably a chief reason MS has chosen the pursuit of DirectStorage the way they did. This is my own speculation, but it's very possible that games which choose to optimize their I/O stack in how they handle code and what data to prioritize to launch the game state (even if yet more data is streamed in to memory in the background), can basically work around any hard limitations in the base I/O hardware.
This would apply to both systems but I'm framing it in the context of the Series platforms here because it means, particularly for 1P games, I won't be surprised if later in the generation they have load times within the 2-3 second range on average. The same would apply to asset streaming; Epic did say they had to rewrite their file I/O system on UE5, but that should also be indicative of something else: the software stack regards any I/O system is just as important (and arguably MORE important) than the hardware of the I/O itself. Hardware means nothing if the software can't harness it.
While I think both systems will transcend their paper specs WRT data I/O performance with the storage devices, I think there's more of a chance MS's 1P in particular really "punch above their weight" (I'm really loathing this term now xD) later in the generation in this regard. The question is how specific the I/O stacks in software code would be in a game-to-game basis because that ultimately determines if 3P titles on Series systems will attempt similar things or just settle for working with the peak limits of XvA (which are still very good in themselves, mind). With Sony the floor for storage data I/O performance is already higher so that is going to be universal across 1P and 3P titles, so as things go on even if XvA closes the gap over time by a considerable margin, the advantage will always favor PS5. It's in the 1P titles where it gets very interesting, and where the gap will more or less likely vanish into margins of error.