I see this as an almost sacrificial, transitional gen from a technical standpoint; where it's just a whole load of teething pains as we see some major paradigm shifts. Which kinda worked out well given how underwhelming the line-up has been..
Going into PS6 we'll have gone through the worst of those teething pains for RT & ML/AI as well as having mature RT hardware and mature ML/AI functionality. Not to mention by then devs may have gotten more accustom to utilising the SSD+I/O in more novel ways. Then add in [hopefully] a lack of pandemic. The cherry on top would be a 2028 launch allowing for better hardware, qualification for the N2P process node with gaafet transistors & backside power delivery and a shift to chiplets to drastically reduce cost; in turn also allowing for better hardware.
We'll probably be starting off with the ability to do 4x resolution scaling with native-like results, the ability to frame-gen 2-3x without a latency bump, have a far greater quantity of far more efficient RT acceleration hardware and have an architecture far more conducive to ML/AI acceleration. There also won't be a widespread resolution jump going into the next gen as far as I can tell; which frees up a big chunk of resources that in all previous gens was spent on just adding more pixels. Nanite/Microgeometry will hopefully be much more performant too given both optimisation and much more powerful hardware. Mesh Shaders have barely been utilised yet too.
We've seen the introduction/popularisation of multiple major technologies this gen, but they've all been half-arsed. Next-gen they'll all actually be possible, performant, effective and worth the cost in real-world scenarios.
So to summarise, apply another 4yrs of experience developing for these technologies as well as another 4yrs of advancement in the technology, hardware and tools..
- Virtualised Microgeometry (like Nanite)
- Ray Tracing & Path Tracing
- ML/AI
- Mesh Shaders
- Then add in better leverage of the SSDs, I/O & compression/decompression.
- Then factor in no increase in the amount of native pixels needing to be rendered (perhaps -- on average -- even a decrease..).
- Then N2P with GaaFET & Backside Power Delivery + Chiplets with mature interposer tech (Infinity Fanout Links?) to offset the cost + lots of cheap 3D Cache dies bolted on.
- Plus another two major gens of CPU/GPU architectures and roughly 4x the raw compute of the PS5.
- Even if we get another long cross-gen, the PS5 Zen 2 CPU won't hold us back as much; relative to what the PS4 Jaguar CPU did. Plus, the PS5 SSD+I/O won't be an issue vs the PS4 mechanical HDD.
By the end of this gen and going in to the next we'll have engines for games built entirely around RTGI + other RT FX, ML/AI scaling, microgeometry / mesh shaders and fast storage + I/O. Simply scaling up from the meagre RT and ML capabilities we have now to what we get next gen will be liberating and likely show some impressive results before further optimisations are made.
I'm a lot more positive going into the next-gen. Everything is coming together and maturing at the right time.