GDDR6 is good enough, but it does not fit the console design nicely. HBM is like a perfect marriage for an APU and a console. It is the ultimate in effeciency in current tech. 

I think even if they launched with GDDR6, they would eventually switch to HBM in a revision. The way I work, I wouldn't even waste the time with GDDR6, and just do things right the first time with HBM. There is no rush, so no reason to do a short term work around.

As far as Storage goes, I think that has to be the absolute minimum. If NVMe is not at least supported, I am going to be very, very disappointed. I look at it the same way as the RAM though. There is no hurry for a next gen device, so don't force something if you don't have to. A new platform splits the userbase, and even with full B/C, you are starting a zero again. Best not make that commiment until you have something well worth the it.

That’s what VR is for.

You couldn't be more wrong on that. GDDR6 fits perfectly with how consoles are made. HBM is always going to be more expensive and complicated than just soldering chips onto a MB. And with HBM you introduce the issue of lower yields per package. More things can go wrong. And that doesn't sit well with how consoles are to be designed. Further more, the benefits doesn't justify it adoption for consoles. 

As i adhered to previous and this time i would be more detailed. 16GB GDDR6 with 256-bit bus will give you 576GB/s. 24GB of GDDR6 with a 384-bit bus will allow for around 860GB/s of bandwidth. All options would cost less than going with HBM. They can one day transition to HBM, but only when the tech and its implementation is cheaper than GDDR6.

You also seem to be of the impression that by 2020 there wouldn't be enough in the tech space that can be implemented in a $400/$500 price point that would justify the jump to next gen. I think you are wrong and would be pleasantly surprised.

My theory, is that the PS5 will come in at $499. I believe that GDDR6 would be well developed tech by then and as such cheaper than it obviously would be today. I also believe that we will see a return to a split CPU and GPU in the consoles even tho they are both from AMD. This would drive the cost of the CPU+GPU up by around 50% but would result in a significantly more powerful console at launch. So rather than seeing a 12TF GPU we could be looking at an 20 to 24TF one instead. If and only if they go with a split architecture. 

If they do go for a 2019 - 2020 release then I hope you are right about splitting up the CPU and GPU. They could definitely hit next gen levels if they go that direction. It would be quite the departure from the direction they went with PS4, but perhaps it makes since this time around.

I honestly didn't even consider splitting things up. If they split things up and give us 32GB of GDDR6 and a NVMe slot, then I would be very happy with that device. Even if they ship with a SATA M.2 chip, it would still be about 5 times faster than the current HDD. With a smaller than normal increase in RAM, that would make load times even shorter than this gen, and vastly improve streaming.

Looking at things this way, I can get excited for a 2019 - 2020 PS5! 

What if the mid-range hardware of the Xbox One X and Playstation 4 Pro isn't enough to satiate ones desires for better hardware though? :P

What did I just read?

HBM is also expensive. The opposite that a console needs.

If you spend $50 extra on HBM DRAM... That is $50 worth of graphics hardware not going into your consoles box, you know that chip that draws all the pretty pictures? Yeah that thing.

You need to balance the price/performance of a console, you CANNOT have expensive high-end hardware.

GDDR6 will be cheaper than HBM. Sony and Microsoft will opt with whatever is cheapest.

GDDR6 will also be mass produced to a higher degree than HBM, so it will have the advantage of scales of economy.
GDDR6 will also see improvements over it's lifetime, not just in terms of cost... But power consumption, densities, bandwidth... And that should transition nicely to a Slim revision.

That could actually be a very real possibility.
But it actually might reduce the costs.

Building one, giant monolithic chip does have it's advantages... But they also tend to have poorer yields than smaller chips... Which is why AMD took the approach they did with Threadripper by using it's "fabric" and up-to 4 potential smaller CPU dies working together.
More workable chips you can get out of a wafer the better... And that means having the smallest possible chips the better.

The Xbox One and Playstation 4 had relatively large chips (for the time and process) on release.

Interested to see what approach they take on this front though to be honest.

SATA 3 TLC SSD would still allow for a 5-7.5x performance increase over the current hard drives being used in the current consoles.
I don't think people fully understand how shit the hard drives Microsoft and Sony chose really are.

The base Xbox One's Hard Drive is only pulling 80MB/s in sustained.

I think an SSHD is probably the best we can expect... And in some instances that can provide improvements of 3x.
Even the Xbox One X's 5400rpm hard drive is nothing to write home about. (I can't believe 5400rpm Hard drives even exist in 2018.)

I also can't wait for 2020, technology seems to be lining up for that year, so hopefully Sony and Microsoft bring the goods to wow us.

RAM: 12-16 GB HBM2, CPU: Custom 8 core Ryzen, GPU: AMD Vega based, Estimated single-precision performance: around 13 TFLOPS (7 times faster than PS4)... if we see PS5 by 2020. So again mid-high end PC of 2018, if they want to keep less than $400 price tag. That should be enough to comfortably run 2020+ games in 4k 30fps.

Honestly if the memory chip module manufacturers can't provide HBM at reasonable prices for console manufacturers then instead of using a short term workaround like GDDR6, I would wait and hope for the JEDEC board to standardize GDDR7 by 2020 and wait for console manufacturers to release their next gen systems by 2021 to coincide with the successor to the Navi microachitecture ... (next generation maybe our last generation since we could run out of maximizing transistor technology so we need a reasonable baseline for next generation in terms of both new hardware features and raw power to provide a high incentive for customers to purchase new systems but hopefully Samsung will ramp up EUV for this year so that we can use the technology to enable higher performance and density for DRAMs) 

why Vega based? maybe Navi is ready for oct-nov 2020. 

One one side 400 could be a posibility. But 2020 is 7 years after 2013, and If we think about inflation, those 2013 400 USS are close to 2020 500 USS. Usa inflation usually is near 3% every year if im not wrong. 

IF PS5 is fast enough, they could price it at 500 USS after all XBox one X was priced at 500 in 2017. 

32GB Ram still cost at least 250€, that's way too much for a console. If the prices don't start dropping soon, next gen might as well just have 16GB

16gb of VRAM is more than enough to display 4k and have high quality textures.

You only need to look at Xbox One X to have an idea of what the next generation will be capable.

The extra juice will come with better CPU per core performance. 10 cores could be good enough to reach 60 GPS.