SONY can learn from Xbox One X technique on TDP optimization by soldering the APU close to the PSU.

You might be right, but you also need to remember we are talking about AMD GPU here, not Nvidia , 15 teraflop Nvidia will be equal to 20 to 22 teraflop on AMD gpu , and also if PS5 releasing in 2020 or even 2021 , 20 teraflop + will be a mainstream GPU or standard even on PC. And high end will be like 24 to 30 teraflop on AMD PC GPU.

if Xbox One X can have 12 GB GDDR5  on this gen why not go nuts on next gen. Also future PC GPU will also go nuts. Even AMD has GPU that support 64 GB HBM2. And also even if it's imposible "a man can dream " . XD 

According to the article, Navi might be a revision on the RX500 series and still a GCN.

Yeah , i understand cost is the problem here, but PS4 also first came with 8 GB GDDR5 when most  of High end PC GPU are 6 to 4 GB GDDR5 in 2013. And still they are managed to make a a high margin profit each of consoles sales. 

I predicted between 12-24GB of RAM, because of prices as you mention we might actually just get 12GB. But don't we need 24GB ram as minimun for next-gen graphics? I read that from other people in resetera forum that it's required for better textures, more polygons and lightning effects.

A. If such a thing were to happen, MS would most likely go with an AMD CPU and an Nvidia GPU rather than give Sony a 2 year sales headstart on next gen or use an outdated Vega chip

B. Such a scenario would never happen anyway, MS has alot more money than Sony, Sony can't afford to pay AMD to make the entire Navi line PS exclusive, MS could easily outbid them. MS needs both an AMD CPU and an AMD GPU in order to maintain backwards compatibility, they won't let Sony lock down the entire Navi family. 

Polaris 10 is a 5 tflop chipset, I think the most we can realistically expect for Navi 10 is 10-12 tflop, which would be a 200-240% increase over the 3 year old (in 2019) Polaris 10.

Recent reports put Navi 10 performance level somewhere in the ballpark of Geforce 1080 and Geforce 1080ti, which are 8 and 10 tflop respectively, and historically speaking it takes more tflops for AMD to achieve a similar result to Nvidia in games, so that sounds just about right for a 10-12 tflop AMD GPU performance wise.

MS would be able to brute force their way to a similar 10-12 tflop level in 2019 by cutting down and underclocking Vega 20

and they'd have to sell at a small loss in order to match PS5's price, and the TDP would be higher so they'd have to use a bigger PSU to power it and vapor chamber cooling to keep it cool, but the end result would be 2 consoles that are roughly equivalent in terms of graphical power.

The most realistic specs, yet. I'd only add 4 cores to that CPU, but otherwise, excellent guess!

I'm slightly afraid that if the next CCX on 7nm Zen is 6 cores, then 6 cores it is going to be on consoles, instead of 8 or 12. It's not like all the three manufactures have shown much predisposition for increasing CPU power dramatically, even though there is sort of a media and fan consensus now they should.

well its complicated, it seems like because of sony's release road map for PS5, Navi had to be pushed back, So amd has 3 navi tech, Navi 10 which will be ready by the end of 2019, Navi 14 maybe same year and Navi 20 which will be the high end gpu and will arrive after navi 14.

Talking about the power. Navi 10 will be equivalent to vega at 7nm. So considering the navi will use next gen memory, even if sony used laptop version ryzen CPU, 4k 60fps could be achieved. But i think Sony's focus will be to push for higher frame rates for VR.

It seems that Navi 10 which will be cheaper, will be exclusive to PS5 with a custom design from sony coz the info that came out of computex seems like Navi 10 will be for professional use, Navi 14 which will be releasing soon after navi 10 will be available for gamers. (i wont be suprise if MS goes for navi 14.)

Navi 20 will be ready by 2021

The Playstation 4 took advantage of higher density memory modules for a similar cost.
Otherwise they would have had to double the amount of memory chips and costs would have blown out.

12GB or 24GB... I just don't see happening from a cost point of view, you would need a 384-bit memory bus to power that.

I'm slightly afraid that if the next CCX on 7nm Zen is 6 cores, then 6 cores it is going to be on consoles, instead of 8 or 12. It's not like all the three manufactures have shown much predisposition for increasing CPU power dramatically, even though there is sort of a media and fan consensus now they should.

Duuude you can do 12GB or 24GB of ram on 256-bit memory bus, this is possible on GDDR6 but not GDDR5.

Since you usually ask for source for some random reason (you should find a source not me), this what I found

"GDDR6 breaks with that tradition and offers in-between options. The standard allows a capacity of 8 to 32 Gbit, but 12 Gb and 24 Gb are possible as well. This will probably make GPU makers happy since it will increase the ability to segment the market based on the amount of memory."

"Today, a GPU with a 256-bit bus can only cleanly support 4GB, 8GB or 16GB. With GDDR6, they will also be able to support 12GB, while still maintaining a full balanced load with identical sized memories connected to each controller."

2400MHz is measly. 3.0 or, as Pema would say, GTFO. These things don't need to be gimped anymore, and I promise you, even 2x the power of current console GPUs (Pro/X) will be an issue for an under-clocked/specked CPU. Forget 4K/60 if that's all they can muster. 

...Umm. What.

Watch digital foundry on how Xbox solution by optimizing their TDP by soldering their APU close to the power source 

Flops are flops. They are absolutely identical regardless if it is nVidia or AMD.
The issue is... The flops that gets thrown around casually like you have done so is merely a theoretical, unachievable number in the real world.

Agree flop are flop but Nvidia and AMD are different , most AMD GPU tend to have bigger TF then Nvidia but Nvidia is more efficient, so if we want to compare the TF between Nvidia and AMD GPU then we use that rule of thiumb, for example GTX 1070 6,5 TF are comparable to RX Vega 56 9 TF on Vega 56. 

The Xbox One X is a "Premium" high-costing device.

The wider memory bus? Costs more.

It's cost more but it will be tackled by next gen memory that will have wider memory bus (GDDR6)

 
The extra memory chips? Costs more.

Yes , every chip cost more, but then is a nutural for every generation they added more RAM

The extra layers on the PCB? Costs more.

Again agree, but it's a natural law, more chip more cost but it will benefit more

The more complex crossbar memory controller? Costs more.

Again agree
The superior power delivery? Costs more.

Agree, but then again by time it will be cheaper 

The superior cooling mechanism? Costs more.

Agree, but better performance, higher clock  and life expectation 

And it has the price to match... Which is why it will never hit the low price points of the Xbox One S.

That's why i am expecting price around 500 USD to 550 USD  which is fine, (Xbox One X sold alot whit this price)

For a next-gen console that isn't targeting a lower price point, then you bet'cha an odd 12GB/24GB set-up will simply be unfeasible from a cost perspective. (Plus ignoring the insane pricing for DRAM right now.)

why would SONY will be using the same price method if Xbox One X can be popular in 500 USD price range, is and also SONY can just close their marginal profit for one year , until each of component becoming cheaper.

And guess what Navi is? Graphics Core Next 6.0.

Most article point out that Navi will start a transition to a new architecture, but i am sure they will have some similarities from old architecture 

They are all part of the same overarching architecture, with small enhancements at every evolutionary iteration.

Yes, but  it will be slightly different or most article point out that it will start to differ from previous gen architecture 

So yes, in short, Navi is a "revision" of the RX 500 series and by extension a revision of the same GPU that is found in the Playstation 4.

Probably, but with newer architecture , again read the interview on the article it will be different albeit will have some similarities. 

The Playstation 4 took advantage of higher density memory modules for a similar cost.
Otherwise they would have had to double the amount of memory chips and costs would have blown out

Agree, that's why PS5 will follow the same route

12GB or 24GB... I just don't see happening from a cost point of view, you would need a 384-bit memory bus to power that.

then they increase it to 384 if it need to