For Ram configuration I think PS5 have 16GB Vram 256-bit bus (572GB/s memory speed) + 4-6GB ddr4 and Anaconda either 14GB Vram 320-bit bus (560GB/s), 10GB Vram available for games or 16GB Vram 384-bit bus (672GB/s Memory bandwidth), 12GB Vram for games.

These numbers make no sense at all, no sense at all.

On a 384-bit bus it would be 18gb Vram not 16gb.

Likewise you wouldn't get 14gb on a 320-bit bus either, 15gb.

You're the hardware expert, figure it out...

My numbers are accurate and I'm sure you can't do 384-bit bus on 18GB Vram as 1.5GB gddr6 Vram sticks are not available.

These numbers make no sense at all, no sense at all.

Digital foundry - "However, more pertinently, memory prices are increasing dramatically and a move to 12GB, 18GB or 24GB (all are possible on a GDDR6 384-bit memory interface)" - link

Your numbers aren't accurate.

Yep I know Eurogamer said that, but 1.5GB Vram sticks are not available, 12x1.5 = 18GB Vram. But there is no 1.5GB Vram available.

No.

It's called turning those effects off to increase CPU headroom or shifting those effects over to another processor. - The game fundamentally operates the same.

Take for example GPU accelerated particle effects, typically these were done on the CPU (Actually, most games still put it on the CPU), that removed the burden form the CPU in idTech powered games so that the CPU can be tasked with other jobs or makes it easier to hit 60fps if you are CPU and not GPU bound.

Conversely, many games use the CPU for post-process effects, like Anti-Aliasing, Blur and more, you can change the type of AA your game uses in order to reduce the CPU burden.

Console games are all about a game of "Compromise". - There is always something you can reduce/remove/move to another chip in order to reduce your hardware demands, which is why developers are able to take a game that runs on a 2.3Ghz 8-core Jaguar CPU in the Xbox One X and scale it downwards to the 1Ghz 4-Core Switch CPU.
It's fundamentally the exact same game.

Good. Because if you were asserting that Lockhart will be equivalent to the Playstation 4 in overall capability based on flops alone... Then you would certainly be highly misinformed.

Scaling up happens all the time. PC thrives on it.

Anaconda is using the exact same hardware base as the Playstation 5. Thus if Anaconda is a glorified PC, then so is the Playstation 5.

I would argue that the consoles are just semi-custom PC's at this point anyway, when looking at it from a hardware perspective.

I would love for the umbilical cord to the Xbox One to be cut off. - But that likely has a customer base that could be 50~ million by the time Anaconda is fully entrenched in the market place. (I.E. A few years from now.)
You want those customers (and more) to transition over, Microsoft is good at making money so I am sure they will make the best decision for themselves either way.
Personally I couldn't care. PC games aren't any "lesser" for supporting old hardware, the PC still has the best looking games on the market if you have the hardware to push it.

Flops are a useless metric, more so with next-gen consoles. You need to stop hanging off it.

For example, Flops doesn't account for the Ray Tracing performance of chips, for all we know Anaconda may have twice the Ray Tracing performance as the Playstation 5... And Ray Tracing is going to be the "big thing" next console generation.

It's because you are basing all your assumptions on flops.

We get new GPU's every year.
Architectures however last allot longer than that.

VLIW/Terascale lasted from 2007 with the Radeon 2900 series right up to the Radeon 6970 in 2011, 4 years isn't bad.
GCN went from a Radeon 7970 in 2012 right up to Radeon RX 580 in 2018, 6 years isn't bad.
Heck, it could be argued that NAVI is still GCN based as it still retains GCN's instructions, it's a hybrid GPU between GCN and RDNA 2 essentially.

As GPU's get larger, architectures get more complicated, the longer an architecture is likely to stick around as AMD and nVidia need to get a return on their R&D investment.

In general console manufacturers pay a lump sum to AMD for their console GPU design, then they pay for the manufacturing, AMD will assist (In return for some money) to shrink their designs down to a smaller manufacturing node.

The profit margins though on AMD's console chips is actually really small compared to a high-end GPU, but boy did they help keep AMD afloat during it's bulldozer days.

In saying that... There are roughly about 75-100~ million discreet PC GPU's sold per year split between nVidia and AMD that accounts from the lowest-end to the highest-end, the bulk of which are low-end and mid-range parts.

12nm is just a refined 14nm process. It's just advertising at it's best.
The Radeon RX 590 is AMD's only consumer 12nm GPU, which is basically an overclocked Radeon RX 580 which in turn is an overclocked Radeon RX 480... Normally a new node should bring with it better power characteristics, but the RX 590 still consumes more power anyway.

Look, there's an easy way to see that your numbers aren't a valid configuration.

Valid configuration: XBO X, 12gb 384-bit bus

12,288 divided by 384 = 32 (CLEAN DIVISION)

Your Configuration: 16gb 384-bit bus

16,384 divided by 384 = 42.666 recurring. Not a clean division, not valid.

Your numbers aren't right, they don't make sense.

You must have missed my previous comment to Drohler, you have 12 gddr6 sticks. 8 of them are 1GB Vram and 4 of them are 2GB Vram, this gives 16GB of vram. 12 sticks x 32 = 384-bit bus, it's CLEAN. You then run them in FLEX MODE, 12GB Vram full speed for games, 4GB slow speed for OS.