Speaking of which ...

Looks like someone on NeoGaf and Beyond3D spotted the "SCORPIO" here

https://forum.beyond3d.com/posts/1916838

SCORPIO: A 36 core "research chip". 

A 36 core Polaris would give you (whataya know) 6.187 TFLOPS. 

It's all lining up now.

It doesnt work that way tho. Shikamarus assessment is closer to what it will be. But even at that it will cost more. 

But even if we use his and your assessment, there is one thing you guys are not considering.

The first AMD 28nm chips were available in 2011. Two years before they were seen in consoles. And over 4 years after the tech was developed. It's not like Sony or MS can just go pick these chips off a shelf you know? It takes a while to improve yields of any kinda chip especially if you intend on ordering 10s of millions of them to go into a console. 

People that buy GPUs basically pay a premium due to how poor yields are while the GPU manufacturers try and get the latest and greatest tech to market. 

This can be alleviated a little tho, since MS and Sony will probably still make the base PS4/XB1 along side the revisions. That allows them not have to make too many of the new consoles while waiting out how long it may take for yields to improve. But even that has its own drawbacks. 

I expect Microsoft and Sony to continue to use the same arch.  Isn't it weird that the PS4 Neo is also rumored to have 36CUs.  

I think that the Neo and the new Xbox One are actually much closer than we think in power.

PS4 Neo is also projected to be Polaris.  Read this from DF:

GPU: This is the most exciting aspect of the spec. Compute unit count doubles from 18 to 36, and clock-speed increases from 800MHz to 911MHz - a 14 per cent increase. That's an overall increase of 2.3x in FLOPs. The question is, what technology is being used here? AMD has created both of its current-gen console processors so far by taking older, off-the-shelf components and disabling a couple of compute units. In effect, Xbox One got the Radeon HD 7790, while PlayStation 4 got a more capable, semi-custom Radeon HD 7870. Here's where things get interesting - the 36 compute unit count cannot comfortably fit any of AMD's existing GPUs. It suggests that Sony and AMD have pushed the boat out, that they are usingthe upcoming Polaris technology.

So basically they are both the same thing.  the question will be which system is clocked faster and what type of memory both are using.  

That really doesnt mean anything. We also know based on leaks that the PS4neo has a 36core GPU. And if the rumors are true too then its also using polaris. 

I am just finding all this a little too hard to believe..... its actually more likely for somy to go from 18 to 36 than it is for MS to go from 12 to 36. And if they are both using 36 then the onoy other possoblilty is that MS is running theirs at a higher clock that what sony is goimg for.

And let's not forget, the reason the PS4 APU could hold 6 more CUs than XB1s was because of the memory system employed. To comoensate for the slow DDR3 ram ms has to put SDram on the die of the APU which took up space. The only way they can have an APU that is idemtical to what the Neo is rumoured to have is of they are doing away with the sdram.... that however stands to break compatibility for every already existing XB1 game, or they get all of them patched to support the new hardware, or they use a memory system so powerful that the new hardware can emulate the sdram...... complicated. 

lol some people are really nervous by a more powerful XBox.

From what we know Neo is not a full blown Polaris, it might be based on certain things but it's basically just a PS4 GPU x2 overclocked a bit.

A Polaris 10 at 32 CUs is 5.5 TFLOP, so that's obviously not the part Neo is using. 

XB2 from the sounds of it will be more like a brand new Polaris 10 (custom), which would enable it to hit 6 TFLOPS without much problem at a reasonable price/power consumption.

Sony is likely hamstrung here by having to stick with something close to the older PS4 architecture, otherwise at 36 CUs they should be getting better performance than just 4 TFLOPS-ish. 

This is the thing. A 6TFLOP GPU seems like complete overkill for a "mid-gen" update. Either:
1) MS are really banking on 4k native gaming, even if it comes with no real graphical upgrades from what we see on PS4. 
2) there is no X3 and developers will start building games from the ground up for the the X2 starting 2018/9 onwards and porting to X1 if they can be bothered
3) All this power and cost is going to waste and just for boasting. 

The most logical thing is No.2, that there is no real next gen (for MS anyway), in which case "next-gen" is going to may be horribly crippled by half baked system, assuming they can't easily quadruble the CPU like how they intend to do GPU.

I'm all for the interative approach but I like to start with well balanced systems designed with future scalability in mind. 

Whichever one utilizes THE CLOUD better!

I believe Spencer when he said he he wanted to make a more powerful Xbox they would just make a new one. I think this is just an Xbox One slim model.

It's ESRAM. Actually it would be at least possible for Scorpio to maintain ESRAM or eDRAM as cache. Question is, is it neccessary? If the system memory is fast enough there should be other solutions like mapping a little bit of system memory for caching.

Full blown Polaris is 40 CU and 2560 shader units. AMD has already deactivated some compute units for the first GPU's. Looks like hey're going for highest possible yields first. Leading to the 36 CU and 32 CU versions.

FLOP/s (single precisision) number of shaders*2*clockspeed in GHz=TFLOP/s

So there's no X compute units is Y TFLOP/s. One of the last benchmarks showed a 36 CU Polaris 10 at 1.266 GHz.

first of all both have to be real. regardless, the new console will be more powerfull than just a new model of an existing console.