The coretech video shows how important communication is, and cerny made sure to point out that the ssd wasnt the only thing that was a 100x faster, everything in the i/o is also 100x faster which should help the ps5 get more out of its tflops. maybe not enough to surpass it, but definitely close the gap while offering better loading times, better character models, more detail and draw distance elsewhere. its probably wishful thinking but there is a good chance ps5 games might end up looking better even if they dont have better resolution.
On a
technical level? That's going to be
extremely difficult to pull off. Again, it's "just" an SSD. I don't want to sound like I'm downplaying it, but SSDs built off of NAND flash memory interfacing over PCIe memory interconnects have their own way and peculiarities of handling data, and the differences between them and volatile memories like GDDR6 are so vast that you can't directly compare them, so it's not a case if the SSD can "close the gap" with, say, GDDR6, because it's a fundamentally different type of technology memory with different use-cases.
Same way how you can't really compare the SSDs to the GPU CUs; one's a non-volatile storage memory, the other is fine-tuned silicon for running intensive computational code on in real-time. 100% different technologies with different use-cases, you simply can't directly compare them so there's no way to imply one will 'close the gap" on the other. Truthfully you can only really compare one SSD with the other's SSD, since they share the most similarities. But the limitations of NAND memory are going to dictate its use-cases for specific types of things, such as low-priority texture-streaming of assets that don't need frequent modification of the data, don't need bit and byte-level addressability, etc. Or other things like streaming lower-priority audio and text data, for example.
And, essentially, PS5 and XSX are capable of these same things, the PS5 just does so about 2.25x faster going by the SSD specs we currently know. There's still aspects to the SSDs in how they work, and info that's of particular importance (such as non-first read random access speeds, latency numbers on the NAND chips, etc.) we'll probably be waiting a very long time to get.
Now,
artistically is a whole different argument. You can have games less sophisticated on a technical level but blow the pants off things with art direction choices. The AMD RT demo for instance; very strong technical stuff on display, but the art direction was terrible. Versus, say, Breath of the Wild. Pretty simple on the technical level, but fantastic artistically speaking (could say this about almost all Nintendo games actually xD).
DLSS utilizes GeForce RTX's Tensor cores. Those are cores designed to run AI tasks.
AMD's stuff can do DLSS as well. In fact IIRC MS's already mentioned using DLSS on XSX, since they are building ML and AI stuff into the system too (since it's also being used in their server blades, at least some of them anyway).
Sony will probably be using something similar to it too but, like VRS, it'll be called a different thing. Kind of like how the Geometry Engine is a rebranding of pre-existing RDNA2 hardware feature. Slight changes here and there, different naming convention, but based on the same thing at its heart.