They do compare the Anaconda to X1X 6TF, so it's 1:1 in both examples. Which would give about 2x more GPU power difference for both X1S vs Lockhart and X1X vs Anaconda. And the difference between Lockhart and Anaconda would be 3x, compared to 4x between X1S an X1X. And it all sounds just too perfect to support the idea of two next-gen models, a FHD one (instead of 720-900p, hence more TF difference in next-gen "base" model) and a 4K one.
It would be weird marketing [....]
Nobody is marketing anything, that's the point. The only thing Sony and MS
THEMSELVES claimed so far is the Zen and Navi architectures, SSD storage, and RT support, but no numbers were even given by those companies.
I bet none of the two will ever compare their next-gen consoles to the competitor - as usual, they will compare their new console to the predecessor, and I'm not even talking about the mid-gen refresh models, because the difference would be so narrow. I mean - 10x more GPU power (vs XB1) sounds waaay better than just 2x more GPU power (vs X1X)..
And I don't expect any specific numbers to be thrown when revealing PS5 and XB2 to be hones, rather than the "X times more" etc., or like when the PS4 was revealed (and bare in mind they didn't compared it to anything at all back then, not even the PS3) - 8-core X86 CPU, a PC-like GPU, 8GB GDDR5, just some ordinary stuff that gives people general idea of what's under the hood.
Which even if you look at it even today looks like quite a powerful machine, better than most ordinary Joes have, but it was exactly the details that indicated that those consoles aren't really that powerful - those 8 cores were actually as "powerful" as a 2C/2T Celeron, out of those 8GB only 5 were actually useful, and the GPU's were below PC's mid-end spectrum, to say the least.
The TFlops were somewhat good from marketing standpoint because that was the only difference between the consoles - they both had the same 8C jaguar, same 8GB unified memory pool, same 1TB laptop HDDs, same Blu-Ray drives, and even the GPUs were exactly the same, except one had more CU than the other, hence the TF difference., because MS wasted a lot of die space for ESRAM.
But now, when Sony learned from both PS3 mistakes and PS4 success, and MS learned from both XB1 mistakes and X1X success, I have absolutely no doubts neither of the two screw up this time around, and won't forcingly try to reinvent the wheel, the consoles will be as simple and dev-friendly as possible. Which also means the consoles will be even more similar than PS4 and XB1 already were, simply because they cannot ask AMD for something better than AMD themselves don't have yet.
Totally doable in the power budget.
So you're basically saying that AMD purposely let's NV be the undisputed king on the GPU market? Because by your logic/math, all they should do is to just downclock the 5700 GPUs (that already run at moderate 1,7GHz) and add CUs (well it's AMD after all, so the "add more cores!" strategy would fit perfectly here ;D), but yet somehow... they don't! And we are talking about PCs here right now, where they are not so limited by space, power usage, thermals etc.
The problem is that AMD's architectures don't scale well, if any, after a certain point - we have a 3rd iteration of Zen architecture, 2nd die shrink (14>12>7nm), and still ~4GHz is the wall. And the frequency is what drives IPC, you know, the Instructions Per CLOCK - it's like torque and HP in cars, you can talk about either of the two, but it's the combination of both that makes all the final performance. And same thing applies for their GPUs - there seems to be a wall between the frequency and CU count, which AMD's GPUs seem to be unable tu surpass, they can wither put a lot of CUs but at low clock, or low CUs at high clock, or some sort of a middle ground, but the wall always appear at the same spot, which Fury X already hit in mid 2015.
BUT - all that being said, and what we all seem to forget, is that Navi, a.k.a. RDNA1 is STILL a GCN-based architecture - a greatly improved and optimized, sure, but still a GCN at its core. So maybe that's it - the next-gen consoles will be based on RDNA2, which is suppose to be AMD's truly new GPU architecture, a Ryzen equivalent for the GPU market, which IMO is the only logical explanation of such a high rumored TF numbers? Yeah, I think that's what I'm going to bet on from now on. Especially that RDNA2 is also suppose to have RT support, it all just fits too good.