This has nothing to do with RAM. This can easily be streamed from the HDD.
...
Blah blah blah ...
Bullshit!
It has everything to do with RAM size and load times. You can't stream every house interior and exterior for an area in Skyrim from the HDD to the measly few MBs of cache on the processing cores. You stream directly into the RAM. The more RAM you have the more data and assets you can hold in at once, so it provides a much bigger effective buffer to improve gameplay load times.
The rate at which you processing cores need data is many orders of magnitude higher than your streaming rate from the HDD. If on the XB1 for example the GPU munches data at a bandwidth of around 170GB/s aggregate, how on earth are you gonna feed it data from a HDD with a data rate sub 500MB/s?
Point is you can't. It's why you need RAM in the first place. And more RAM means more data closer to your processing cores, for when they need it. Which is what you want.
Saying any amount of RAM is too much is just imbecillic, outside of maybe having like more RAM than data on your entire game disc. The ideal would be able to store the entire 25-50GB game in RAM local to the cores. Then your load times would be blisteringly fast, and annoying stuff like pop-in and the texture-streaming issues UE3 was notorious for would be a thing of the past.
The point is Sony would not have been able to get much more in the way of GPU/CPU if they discarded the RAM type or amount. The fundamental goal of their entire design was to have an APU design, i.e. CPU/GPU on the same die, that would afford them major savings in wafer/chip cost, cooling, power consumption, motherboard design, R&D, IHV royalties, assembly, packaging and distribution. Sony pretty much pushed theirs to its manufacturable limits in terms of performance/associated costs ratio. More ALUs (to make a meaningful difference to performance) would have meant a massively bigger chip, thus lower yeilds, plus higher power consumption which in turns means a more expensive cooling solution, larger box thus higher packaging and distribution costs.
To go any higher than PS4, Sony would have needed to go with a dual chip solution, ala PS360. The costs associated with that would have skyrocketed relative to an APU design. At the same time sure the associated costs would have been similar to what the xbox 360 cost MS, but there are also long term factors that would have massively impeded both their ability to sell consoles, but also their ability to cost reduce the box in a cost effective way; namely process shrinks now being far more expensive and far less viable than they ever have been. The knock on effect being their inability to cut price would mean sales would slow considerably as they would struggle to gain access to the more price conscious mass market consumer. Neither Sony nor MS could afford to take a big loss on HW sold, as with less viable process shrinks the risks would be too great.
It would be bad for the entire industry if both Sony and MS went with dual chip designs and couldn't cost reduce them effectively, as you'd see eventual HW LTD installed bases at likely a half of what they were this gen in the same timeframe. With higher production costs and half the market to sell to, publishers would dissappear to mobile platforms.
So no, Sony and MS couldn't have gone much higher in terms of performance than what they did. They actually had a responsibility to launch their consoles at a more mass market price this time, as the more consoles they sell the healthier the industry would remain and the better it would be for everyone.
Also the good thing about a simpler APU design is it makes process shrinkage much easier as you only have to think about a single chip. They can easily move from 28nm down to 22nm once the latter bulk process matures. Afterwhich, they could simply ride it out until next node is ready and launch new consoles on the newer process. That would be what i'd do. So we could be looking at a solid 5 year cycle again. Plus by then some real good stuff like DDR4 and stacked chips would be available, so an orde of magnitude jump in performance could be very very real next time (think 32GB of stacked RAM with a TB/s bandwidth).