Like this take. We are definitly in sort of a bind. 

If they make the decision to go to PS5 in 2019, then it will be a strange, and short gen, as the tech to blow it out of the water is right around the corner. A Ryzen 2 based PS5, with a ~12Tflop Navi based GPU, 16 - 24GB of RAM, and an HDD, would basically be another 5 years of this gen at 4K with some extra CPU play room. Going this route, would mean PS6 will be needed by 2023 or 2024, but would give other players a lot of room to jump into the market in 2021 - 2023, and greatly outclass them.

If they build a PS4 Premium in 2019, stick with modified Jaguar at 2.7 - 3.0GHz, ~10Tflop GPU with some Navi upgrades, 16GB of RAM, and an HDD. Then they can bring out a worth while PS5 in 2021 or 2022. Something in the ballpark of Ryzen 5 or 6 with 8 - 12 cores, a 20+ Tlops GPU, 128GB of RAM, and an 8 - 32TB SSD (depending on price reduction and capacity). This route to me looks safer, as it gives less of a window for anyone to slide in with a noticably better device. The only company that could try to slide in with this stratagy would be MS, but I think they would have to take a hit on every unit if they tried to slide in, in 2020.

Depending on when we get to 5 or 3nm may also play into this. 7nm will be here in late 2018 or early 2019. If PS4 Premium is done, then the release of 5 or 3nm fabrication will determine when PS5 comes. It also kind of blocks MS from trying to get the jump start, because, they can't do an XB4 in 2020 on 7nm if 5 or 3nm will be available in 2021 or 2022, becuase it would give Sony a massive advantage at the baseline console next gen.

Then there is one more angle to look at. VR. Can Sony hold a strong position in VR with a PS4 Premium and a ~3GHz Jaguar based CPU, or will they have to do a half step generation and release a Ryzen based PS5 in 2019? 4 or 5 years this early into a new format is an eternity. If Jaguar is going to do massive damage to their VR business, then it could be a decision that looses them billions, and a chance to ride VR to Apple, Google, Microsoft like success. 

I know this is a general next gen storage tech discussion, but other tech plays into this as well. As a Sony consumer and share holder I am looking at it from the perspective of PlayStation. If I had to make the decision, I would do a 7nm PS4 Premium, and build a true next gen PS5 with NVMe/m.2 Storage as soon as 5 or 3nm fabrication is ready. 

Yup... thats true. Here's my take on th whole next gen starting point.

I can bet on it and be really sure.... that next gen will not start before 2020. More specifically, next gen will not start before we see whatever AMD has on a 7nm/10nm process. And I thik all the dots are already there  to connect.

The thing here is the PS4. Sony has got everything going for them right now. They are the current market leader, and likely to remain that way for the at least the next 2-3yrs. They are making more money from the playstation than they have ever made and there biggest rival and person to potentially match their sales is someone that doesn't even compete in the exact same gaming space as them (nintendo with the Switch). They have absolutely no reason right now to want to jump start a new gen early.

Then we have the XB1x. MS releasing this year kinda means they are at the bery least 2-3 years away from any other major hardware release. But more importantly, it means sony has seen their hand. Sony will not put themselves in a position where they can release early with current available tech that has a chance of being completely blown out the water if they had just waited a year more as they now know MS has no qualms doing. So sony will wait.

And the PS5 is going to come in at $500. Well, at the very least it will be a $500 box. If sony releases a year before the next xbox then they would price it at $500 and drop the price aroun the time the next xbox is releasing. If they are coming out around the same time they would just price match the next xbox.

So i feel a good way to gauge what will be in the next consoles is look at whatever will be mainstream in say 2020. Hell, a March 2021 releaseisn't beyond reason at this point.

How far in the future? Because if 20 years probably there will be come and gone some other techs on the subject.

For next gen consoles? SATA because of price.

A few things.

1. this 7nm process we are all talking about isn't even gauranteed to be available by 2018/2019. Thats just powerpoint talk right now and barring any surprises they don't always come out exactly when expected. Theer is always some sort of hiccup or delay. Like take the 14nm process for instance, believe it or not that has been available since around 2013 when the PS4/XB1 launched. Albeit only from intel and in their highest of the high end processors. It took it another 3 years to become mainstream (i.e: AMD support). Right now, no one has even gone out of risk production phase for the 7nm chiips. Thats assuming any of them has even started.
   
   
2. Having said that, it would be alsmot ridiculous to be waiting for 3nm that no one is even talking about now. That thing may very well not show up till 2025 or later, especially with how hard it is to work with silicon now at these super small node sizes. At some point even intel anounced they wouldn't use silicon at the 7nm node and opt for an alternative material like graphene or something. They have gone back on that but it just goes to show how hard its getting.
   
   
3. I think a lot here are underestimating the gains with the 7nm node. Think of it this way, if going from 28nm to 14nm allowed for a 4x bump in processing power in these consoles (they could have achived an at least 2 times bump in the CPU if they wanted but those jaguar cores are hard to improve) then going from 14nm to 7nm would see similar results. Or at least close. So basically, take the XB1x (or even the PS4pro) and multiply that by 3 as a worst case scenario. 
   
   That means we can have anywhere between 12TF and 18TF in the next generation of consoles on the GPU side of things. And that is pretty much a gaurantee if they are built on a 7nm process. At that point its just physics lol. Mind you, this is me lowballing it. Factoring a doubling of cores and a small bump to processor clocks.  But the actual thing could be a lot better than just those things as their would be improvements in other areas of the chips that would result in all round better performance.

Also, if we have ceteris paribus for all else on the architeture, looking at how much more flops would be necessary on a "scalar approximation" isn't really that off.

Keep in mind that Intels 14nm is NOT the same as TSMC's/Global Foundries/Samsung/IBM and so on's 14nm.
They are using the "nm" as a marketing gimmick and isn't really representative of geometry sizes.

I actually had to google "ceteris paribus" as it's a term I wasn't familiar with. In English it's a statement often associated as "everything else being equal".

However even then flops isn't really an accurate representation of anything. - You could have double the flops but only see 20% of the gain in gaming performance as your are bottlenecked elsewhere in the GPU pipeline.

An 8TB M.2 Chip would even in 2023 cost way too much for a console, though I could see an M.2 or possibly even U.2 slot (basically the sucessor of SATA (Express)) for storage upgrades by then. For high storage spaces HDDs will stay on top for a while as Flash will stay expensive. For comparision's sake a 4TB HDD (I had a hard time finding an 8TB SSD) costs little over 100$, a 4TB SSD? over 1500$! The difference is just too big for a console, which are on a tight budget

Also, to cram as much Flash memory storage on a M.2 is potentially risky: The smaller the production of Flash memory in nm, the shorter the life expectancy of it's cells. But to get that much memory into the small arry of chips on an M.2 means you'll need to produce them in 7 or even 5nm, at which point the write cycles are in the hundreds if no major change in the hardware is done by then. The big size will mitigate it, but once the M.2 is close to full every save can become deadly to it's cells. And due to electron migration (which gets common at such small distances), it may not even need those saves for the cells to die. While there are successors to Flash in the works or even released (3D X-Point) who don't have that problem or to a much smaller degree, these will be even more expensive than Flash (3D X-point for instance costs about 5 times as much as Flash memory on their cheapest!)

Apple TV and Amazon Fire TV use Flash (=/= SSD!) because they use much less storage space, more in the line of the Switch (which also uses Flash Memory), and because a HDD would be too bulky for them. But that doesn't make them any faster (well, they can load a bit faster, but certainly won't run faster, and certainly nowhere near a console

For real???? And they let intel get away with that? I mean I have always kinda been suspicious of intel; like how can they just be so far ahead of everyone else, but now I am livid.

Sigh, they do have some really interesting tech in the pipeline though....

You both make solid points overall. It's not like guessing the next gen storage solution is impossible, but every year this gen continues, the storage solution becomes more likely to be future based tech and not an HDD. PS4 could very well last until 2021, but I have a hard time believing that PS will be ok with Pro as the flagship, if that is the plan. If there was any truth to Pro launching to compete with PC and push VR, then PS will have no choice but to continue to push out frequent hardware to make sure they do not fall behind in that sector. A 4 or 5 year gap between Pro and PS5 leaves the door wide open for competition in more ways than just VR. I think one way or another, PS is going to want to launch another box 3 to 4 years max after Pro.

With the rumor of a next gen console using a separate CPU and GPU, and seeing the announcement of that today into the market by Intel, a separate CPU, AMD GPU, and HBM2, all on Intels version of Infinity Fabric basically, I could see PS5 using that type of layout with a 12nm or 7nm Ryzen and Vega + HBM2. That could launch mid to late 2019 in a PS5, which would make using an HDD or SSHD quite likely. The system could definitely house some on board flash memory as well to help, but I would think the main mass storage would be either an HDD or SSHD in that case.

I understand the reasons why extending PS4 makes sense in terms of market share and mind share, and it does make sense, but with Ryzen launching early 2017, and PS not using that tech at all until 2021 or later, seems like lost potential. Considering the first APU's launched in 2011, PS and XB must have been working on them with AMD by then for their 2013 launches. I also can't believe that AMD isn't doing everything they possibly can to get Ryzen into the next gen consoles asap. It would help AMD in many ways as well.

You also have to wonder about PS and their praise from devs on how easy the PS4 was to work with. PS made it quite clear that they wanted to make a console that the devs would be happy to work with. Seeing how easy XB1X is to work with now, minus the upgraded, yet weak Jag CPU, you have to wonder how long PS can drag out the PS4 life span before devs demand better. Sure the pubs are happy with a huge audience, but where is the line between them?

PS4 has a few more years in it regardless, but given the chance to extend that or jump into some newer tech, based on the PS2 to PS3 transition, I think that PS is not going to want to hang on to PS4 for dear life like they did with PS2, holding PS3 and potentially PS5 back. PS3 had a world of other problems on top of that, but PS2 having the backing it did, for as long as it did, hurt PS3 no doubt. PS3 was also sold at a major loss and needed to be paid off, while PS4 is sold at cost if not for a small profit. Keeping PS4 alive as long as possible, would be quite greedy. PS learned many PS3 lessons, but did they learn these as well?

The concept of a system with a dedicated CPU and GPU sounds great, albeit expensive. I am also not sure they can get that kinda setup to work with HBM. An APU yes, but not discrete CPU and GPU. I could be mistaken though.

And I wouldn't say that the more this gen drags on the more likely it is that we would use "future" tech on the storage front. Technically we are only going to use stuff thats on the market right now. The question is how much said stuff is gonna cost.

Even if sony doesn't use Ryzen, doesn't mean there wouldn't be something available for them to use in 2019/2020 that is better than ryzen in every conceiveable way. And I think AMD is really happy with how things are right now. Their hardware is powering 4 very successful consoles right now (PS4/PS4pro/XB1/XB1X) thyey are probably taking orders of over 30M chips each year right now so things are probably really good for them on that front for the time being.

And this isn't about being greedy. Don't forget that its a business. And the real money comes from selling software. Having a platform that is doing so well commercially just means they are under no real pressure to push out a replacement. Now more than ever they are in the best position possible to wait it out until the right combnation of tech not only becomes available but gets to the right price.