There's quite a bit there, but what I'm interested in is the 12 TFLOPS for a GPU. There have only been three consumer GPUs ever released rated at 12 TFLOPS or higher. Two of them are GCN derivatives, so much less efficient on a per-flop basis than RDNA (or Turing). If the next Xbox were released today, it would likely have the second fastest consumer GPU ever made, trailing only the GeForce RTX 2080 Ti. Faster than a Radeon RX 5700 XT, faster than a Radeon VII, and likely faster than a GeForce RTX 2080 Super. That's not bad.
The other notable thing about those high TFLOPS GPUs is that they use a lot of power. And I mean really a lot of power. There has never been a GPU rated at 12 TFLOPS or higher single-precision with a TDP lower than 250 W. That's not just for consumer GPUs, but also professional graphics, server, or laptop cards.
And that raises the question of just how much power the next Xbox is going to burn. (Yes, I know what the console is called, but its name is stupid enough that I refuse to type it.) Game consoles generally want to keep their power output low. That helps to keep the size, noise, and cost down. If the next Xbox has a 250 W GPU, then once other things are added, it will be by far the most power-hungry game console ever made.
Which is why I don't think it's going to have a 250 W GPU. The big question is how much more efficient AMD can make their RDNA architecture. Right now, Navi is about even with Turing on energy efficiency, but that's in spite of being a process node ahead. Normally, if you have to be a process node ahead in order to be even on efficiency, that means that you're losing badly with the efficiency of your architecture. AMD is way ahead on die size efficiency, so they certainly got something out of the process node.
But there are two reasons to believe that AMD could have a lot of energy efficiency to gain with RDNA2. One is the process node. It's not clear whether it's going to be built on the same TSMC 7 nm node as Navi and Vega 20, or on the new 7+ nm EUV node that AMD is moving to for their upcoming Zen 3 CPUs. Microsoft said that the next Xbox will have Zen 2 cores, which is already on TSMC 7 nm, but AMD had no problem making Jaguar cores first on Global Foundries 28 nm, then later porting them to TSMC for Microsoft and Sony.
But the other reason is AMD's history. AMD has historically had no problem with launching new GPU architectures that do a lot wrong on energy efficiency, and then going back in subsequent generations and cleaning it up. Their first VLIW card, the Radeon HD 2900 XT, was famously hot, late, and slow. A little over a year later, the Radeon HD 4870 that was heavily derivative of it had caught Nvidia in energy efficiency and about doubled (!) what Nvidia had in die size efficiency. Yes, the die shrink helped, but Nvidia did as much die shrinking in the intervening year as AMD did.
Or look at their next major architecture, GCN. Their lead chip, Tahiti (Radeon HD 7970) was something of a power hog. Eventually, Fiji (Radeon R9 Fury X) would about double its performance on the same process node and in the same power envelope as before. That AMD could do that on the same process node means that there was a lot to fix in the initial chip. HBM helped, but that's a small fraction of the energy efficiency gains that AMD was able to get.
If AMD can deliver that sort of efficiency gains for RDNA, then they might not be at an efficiency disadvantage to Nvidia, even after Nvidia catches up on process node. Microsoft didn't announce a TDP, but they certainly already have guidance from AMD. If that 12 TFLOPS GPU is rated at 120 W or 150 W, that would be a lot more in line with previous consoles. To scale Navi up to 12 TFLOPS, you'd be looking at something not far shy of 250 W. That's quite a difference.