Just going to have to take it for what it is. An Arm processor tied to a mobile SOC isn't going to offer sustained performance of a modern X86 CPU. Especially not Zen 2 with eight cores with MT.

There's also the issue of storage space, speed and I/O which is going to another factor in the equation. I can't imagine Nintendo investing in even more expensive NAND cartridges for games. That's gonna get expensive, fast. Nintendo likes making money off hardware.

Then there's publishers hurdle. I can't see a modern GTA released on a Nintendo console anytime soon. I mean how is it that GTA V is on every console including a toaster but not on Switch,

Aside from the issue that the Switch can't really run GTA Online comfortably, which is where Rockstar makes the bulk of their money, GTA may be a bit of a unique case in a different way.

DMA Design which is basically the developer that morphed into the GTA team was once very close to Nintendo in the mid-1990s, being one of Nintendo's so-called "Dream Team" for the N64.

They had a falling out with Mr. Miyamoto apparently because development of the title Body Harvest (which was to be published by Nintendo) ran into many conflicts. Body Harvest is in a lot of ways the predecessor to GTA. Nintendo dumped the title, which probably upset DMA, but it was published on the N64 by a 3rd party. 

The story goes DMA also pitched Nintendo the original Grand Theft Auto itself (then a 2D top down view game) and Miyamoto did not approve of the game's hyper violence, so Nintendo would not publish the game (imagine for a moment how different game history might be had they said yes).  

Many speculate that the GTA team to this day still harbors a grudge and resentment towards Nintendo for that past history, whether that's true or not, I don't know and I doubt you will ever get any kind of confirmation from any side. They were willing to make a GTA game for the DS but that may be because the system was so wildly successful even they couldn't try to take a swing at that gravy train (the game didn't sell as they would have liked apparently). 

Interesting, didn't know that. That might be partially the reason we never got GTA's on Nintendo consoles from GC onwards.

Kinda reminds me of the reason why Playstation was created :P

This needs a reply, because its silly (imo).

1) Xbox Series S (lockheart) has the same cpu, as the series X. A 8c16t amd zen 2, which is like ~10 times more powerfull than the cpu inside the PS4/Xb1 and likely also the Switch.

2) DLSS isnt something magically that no one else has.  Playstation 5 AND Xbox Series X+S, both have machine learning, to do their versions of DLSS.

"Series S XBox model, which is supposedly about 1/3-1/4 the power of a Series X. Suddenly the power gap really isn't large at all, with DLSS in fact the Switch 2 may actually be more powerful than the Series S."

The achitecture of the GPU inside the PS5/XSX+S offering around ~50% more bang-4-flop than the older architectures in the PS4/XB1.
That 4 Tflop rumor Xbox Series S, is like todays 6 Tflops GPUs in consoles.

What your basically saying is that a Switch 2, will be more powerfull than a Xbox One X, in terms of graphics performance.
Thats not gonna happend, there is no mobile chip that powerfull in existence (or development, that I know of).

Switch 2, would need to be like 15 times more powerfull than the current Switch, for that to happend.

I'm not sure technology has moved that far along, in the meantime.
Even if nintendo used the very newest of technologies, which nintendo usually doesn't.

*edit:
on the other hand, stuff like ray traceing is pretty demanding.
If nintendo Switch 2, runs these games without that(while the others do), it might make up some of the differnce, along with lower resolutions and such.

It might be possible for a Switch 2, to be able to run some 3rd party games, much like the current switch runs some too.

*edit2:
"If MS had a Series S equivalent for the current XBox One, it would've been basically 400 gigaflops ... which is the same as the Switch docked."

Thats probably true... 1/3th the power, of a Xbox One, is roughly a Switch.
1300/3 = 433 Gflops.

AMD has not shown a direct comparable to DLSS 2.0 as of now. Microsoft has a machine learning solution but they have also not been forthcoming in showing it working really. They had a panel in 2019 GDC but have not mentioned it again at all and not in any XBox Series X discussion whatsoever which I find very odd if this is supposedly a feature its supposed to have. 

I wonder if the performance hit the machine learning takes is greater than what they have been letting on, Nvidia's solution requires dedicated Tensor cores (actual hardware) to achieve DLSS 2.0 results, AMD does not have Nvidia's investment in A.I. hardware.

Switch 2 will also benefit from new architecture, quite frankly Nvidia has better architecture than AMD does, the RDNA 1.5 architecture you're talking about as being so much better than GCN ... guess what? Nvidia's existing Turing based architecture from 2 years ago is what RDNA 2 is basically catching up to. 

Switch 2 will likely be Ampere or Hopper based which is ahead of Turing architecture wise.

The bottom line though is this gap is a lot smaller than the last generation. 

The Switch was 394 GFLOPS docked last time around. The XBox One, the lowest speced of the two systems was 1.2 TFLOPS.

This time around you're probably looking at a 2-3 TFLOP docked Switch 2 with DLSS 2.0 versus a XBox Series S which is 4 TFLOP.

This is a much smaller gap any way you try to spin that. 

How would you do that with a 20-30watt device though?

Mobile is more power effecient at lower performance ranges, but they cant scale up, while keeping the same effeciency.
They are designed for low power, and work best at that range.

Once you start scaleing up some of these Arm CPUs and mobile GPU they turn no better than current x86 CPUs.

Could Nintendo magically make a Switch 2 thats 15 times more powerfull than the current Switch? while still being a switch?
I dont think so... they at the very least would need to make it bigger though (more cooling, ect), add bigger batteries (more weight) ect.
The problem is you then run into the issue of size, handhelds are expected to be a certain size to make the portable aspect real world usable.

You could have a laptop sized Switch 2, that would be as powerfull as a Xbox Series S without much issue.
Is that still a switch though?

Supposedly Microsoft has yet to show ANY game running on a Xbox series X.
All the footage shown at their event was run on PCs.

Yes, its questionable, and yes neither Sony or MS, have shown their machinelearning techniques off, working in a game.
I think back to the PS4pro with checkerboardering.... did we know well in advance how that worked? were we shown it working in a game well before the console launch?

15 times? No. 7-8 times though? Yes, quite easily. Tegra X1 will be 8 years old in 2023, that's even older than the PS4 is to the PS5. 

7-8 time performance leap gets you into 2.7-3 TFLOP range, when you factor in DLSS 2.0 (or 3.0?) though that means that 2.7-3 TFLOPS performs more like double+ that performance.

You're asking the processor to only run at maybe even 576p instead of 1080p, that's a massive difference. 

This gap is far, far smaller than the current Switch-XBox One. Way smaller. 

DLSS 2.0 can resolve even a very, very nice looking 1440p image from as low as 576p. 

They don't even have to show it off, why haven't they mentioned it even once? I mean if the XBox Series X/S supported that, I would think they would be yelling it from every roof top.

My guess is the dirty little secret that MS didn't tell in their old GDC demo was that the machine learning they're talking about takes a bigger processing hit than they stated. 

Checkerboard rendering is a thing ... and DLSS 2.0 kicks the shit out of checkerboard rendering. Digital Foundry did a test on Death Stranding on PC using DLSS 2.0 versus the PS4 Pro version which has checkerboard rendering, the DLSS 2.0 version trounced it. Nvidia is way ahead of other companies because they have actual hardware based solution (Tensor cores) and have invested a ton of R&D into that whereas AMD being small potatoes compared to Nvidia can't compete with that. 

Soundwave, I tried looking all I found was a referance to something on neogaf, with a guy saying he found some patent.

https://patent.nweon.com/8108

https://patents.google.com/patent/US10685425B2/en?inventor=Mark+Evan+Cerny&sort=new

"In graphics processing data is received representing one or more vertices for a scene in a virtual space. Primitive assembly is performed on the vertices to compute projections of the vertices from virtual space onto a viewport of the scene in a screen space of a display device containing a plurality of pixels, the plurality of pixels being subdivided into a plurality of subsections. Scan conversion determines which pixels of the plurality of pixels are part of each primitive that has been converted to screen space coordinates. Coarse rasterization for each primitive determines which subsection or subsections the primitive overlaps. Metadata associated with the subsection a primitive overlaps determines a pixel resolution for the subsection. The metadata is used in processing pixels for the subsection to generate final pixel values for the viewport of the scene that is displayed on the display device in such a way that parts of the scene in two different subsections have different pixel resolution."

Varying effective resolution by altering rasterization parameters.

This sounds like a cost saveings technique.
Ei. instead of checkerboarding.

*edit:
Apparently this cerny patent, links back to another they have on VRR (variable rasterization rate).