StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (New Staff Post, Please read)

[kirbyfan407]

Interesting news today. The BOTW demo allows Nintendo to show the power of the system without revealing any new software using a game many people are familiar with, and the Matrix demo allows Nintendo to say "Yeah, you can get your PS5/XSX games running on our hardware." The only disappointment for me is that it doesn't seem like the demos were running on actual hardware, which is preventing me from having any firm expectations for the hardware performance at this point.

It'd be nice if the BOTW demo became an actual release via a patch. When the stories broke today, I began to wonder about what a Tony Hawk's Pro Skater 1+2 combo remake of BOTW and TOTK for Switch 2 would look like and I dreaded the thought of a continuous 150-hour main storyline Zelda game.

Wouldn't it be better for them to release this info during the Direct hype so that it could possibly be hidden from more people and maybe not attracting Nintendo's attention or something for the source (And maybe the sources would then be less pissed about the info coming out so early?)

From a media perspective, I imagine a primary goal would be driving traffic to the site. While a journalist may have those concerns about sources, I'm guessing others at the outlet (editor? I'm unfamiliar with how media orgs are structured) would want to push a story like this at a time when they can get fresh attention they wouldn't already be getting from Direct news. There may also simply be the desire to be the one to break the news, so it's in your interest to be the first to push publish and not sit on a story if it's ready.

In any case, necrolipe's most recent post about it being a joke maybe makes the reasoning discussion a conversation without a serious foundation. But, hypothetically, I think it makes sense for outlets to push the story when they did.

 

[ILikeFeet]

Wouldn't it be better for them to release this info during the Direct hype so that it could possibly be hidden from more people and maybe not attracting Nintendo's attention or something for the source (And maybe the sources would then be less pissed about the info coming out so early?)

if they don't want to attract Nintendo's attention, they wouldn't say anything in the first place

 

[Pokemaniac]

Making use of more resolution by using the same design but "more" sounds ideal. I wouldn't be surprised if game icons stayed the same size, so to remain sharp their presentation in the UI would have to shrink. Definitely what I'd be leaning towards- smaller icons, maybe a grid like that, and themes lathered on top (only on NG Switch, only for NSO members via the NSO app.)

I'm fairly certain the icons are higher resolution than the Home menu is currently able to take advantage of. Don't recall the exact numbers off the top of my head, but I think they'd be fine if they bumped the menu resolution up to 1080p (which I think is the most like scenario, should stay whatever the native resolution of the screen is) without changing anything else.

 

[justanotherperson]

Could Nintendo, with the power of DLSS, push last gen games to around 4K120 via DLSS Ultra-Performance?

 

[Kise Ryota]

for Splatoon 4, people just want better netcode and tick rate

Dedicated servers, please.

 

[kirbyfan407]

Could Nintendo, with the power of DLSS, push last gen games to around 4K120 via DLSS Ultra-Performance?

Even if the GPU hurdle was overcome using DLSS (assuming it's just the AI upscaling solution), I think you'd start to run into major CPU limitations with a 120fps target. But I'm sure, in theory, some game could run at 4k120 on Switch 2 if the hardware supports that output. Someone more informed about Switch 2's rumored CPU capabilities and the CPU demands of older games could perhaps point to a specific generation of games that would fit this criteria.

 

[BreakAtmo]

Could Nintendo, with the power of DLSS, push last gen games to around 4K120 via DLSS Ultra-Performance?

Switch games maybe. Not PS4 stuff, except maybe the sort of thing that the PS4 could run at a locked native 1080p60.

 

[fwd-bwd]

Updated compilation of Switch NG release window leaks/predictions

[Publications]
MoneyDJ: “Q1 next year” (Foxconn’s FY starts in January)
TechInsights: “March or April next year”
Nikkei Asia: “next spring at the earliest”
Eurogamer: “latter part of next year” but “sooner if possible”
Nikkei Shimbun column (not report): “the industry” expecting “second half of 2024”
VGC: “second half of 2024”

[Individuals]
Pokemon leaker: “early 2024”
necrolipe: “H1”
NWeedle: “before summer”
Nate: “informed speculation” being “late 2024” (this is quite old; not sure if he’d want to revise)

 

[Zedlars]

400: pretty likely
450: possible but less likely
350 or less: very unlikely
500 or more: absolutely no chance

Yeah... 400 seems like the sweet spot.

 

[Serif]

2025 is dead.

 

[BreakAtmo]

Even if the GPU hurdle was overcome using DLSS (assuming it's just the AI upscaling solution), I think you'd start to run into major CPU limitations with a 120fps target. But I'm sure, in theory, some game could run at 4k120 on Switch 2 if the hardware supports that output. Someone more informed about Switch 2's rumored CPU capabilities and the CPU demands of older games could perhaps point to a specific generation of games that would fit this criteria.

It's likely that the A78C in the Switch 2 is going to be tremendously better than the PS4 Jaguar CPU. Maybe not 4x better, but we might be surprised.

 

[Skittzo]

2025 is dead.

Long live 2026

 

[TheSentry]

Nintendo should rebrand DLSS in the Switch 2 as… “Blast Processing”

 

[epicmartin7]

Nintendo should rebrand DLSS in the Switch 2 as… “Blast Processing”

Nintendo does what Nintendon't

 

[LiC]

Nintendo does what Nintendon't

*what Nintendidn't

 

[Serif]

Nintendo should rebrand DLSS in the Switch 2 as… “Blast Processing”

 

[Dmax3901]

Nintendo should rebrand DLSS in the Switch 2 as… “Blast Processing”

 

[SiG]

Needs to read "Super DLSS"

 

[Atlus]

One thing I am curious about is if the dock act as a supplemental computing device this time around to give visual parity to PS5/ Xbox Series when docked. Because it would be interesting to see if all the hardware is in the console itself. Shouldn’t be too long before we know. However the announcement trailer may not exactly reveal that.

 

[D36]

Super FX was actually a thing, though.

 

[Serif]

Super FX was actually a thing, though.

Exactly. Hardware acceleration to improve the graphical capabilities of a console. We're dealing with Super FX 2: Tensor Trouble.

 

[justanotherperson]

One thing I am curious about is if the dock act as a supplemental computing device this time around to give visual parity to PS5/ Xbox Series when docked. Because it would be interesting to see if all the hardware is in the console itself. Shouldn’t be too long before we know. However the announcement trailer may not exactly reveal that.

That would be expensive, difficult to implement and useless. It seems like a better CPU would be a better investment for the console, as it seems the console has RAM parity with Series X and PS5 (even though we might still be in basic-ass OS territory), so no need for 16GB of RAM.

 

[Serif]

One thing I am curious about is if the dock act as a supplemental computing device this time around to give visual parity to PS5/ Xbox Series when docked.

Unnecessary with DLSS.

 

[Deleted member 887]

Hmm I feel like they will come up with at least some marketing explanation why certain games have higher performance or better resolution than others. Especially if BC is a thing and not every game (likely most BC games, maybe even some new releases) would be using DLSS or other possible hardware features.

It could be a label with something as simple as "optimised / improved for Switch NG". Power will likely be not the focus overall but I can see them using comparison videos and such if they made games like for example BOTW look and run way better because it is a strong selling point.

Optimised versions of BC Games maybe even could be tied to a NSO subscription.

I'm not sure such a label qualifies as marketing their upscaling solution

 

[Individualised]

Exactly. Hardware acceleration to improve the graphical capabilities of a console. We're dealing with Super FX 2: Tensor Trouble.

Super FX 2 was also a thing. Yoshi's Island and Doom use it.

 

[Vanillalite]

That would be expensive, difficult to implement and useless. It seems like a better CPU would be a better investment for the console, as it seems the console has RAM parity with Series X and PS5 (even though we might still be in basic-ass OS territory), so no need for 16GB of RAM.

*It might be in terms of the amount of ram, but we won’t be there in terms of bandwidth. Not that this is Nintendo skimping or anything. Every device on the sun that’s mobile all uses LPDDR. Is what it is.

 

[Giancarlo]

So who has started to save up for this thing, got a few months but yeah hope its around 300 dollars that will be the sweet spot price

i dont expect $300 for Nintendo Switch sucessor, given all the tech is gonna gonna have, $400/450, is what we should expect

 

[Deleted member 887]

Could Nintendo, with the power of DLSS, push last gen games to around 4K120 via DLSS Ultra-Performance?

If you're asking if a developer could, with a custom port designed to use these technologies, take a PS4 era game and make it 4k120fps on something roughly Switch 2 performance, with DLSS?

Theoretically yes. Would it happen? I would bet money against it, for myriad technical reasons, but also because I think it would look like vaselined donkey ass.

Even if the GPU hurdle was overcome using DLSS (assuming it's just the AI upscaling solution), I think you'd start to run into major CPU limitations with a 120fps target. But I'm sure, in theory, some game could run at 4k120 on Switch 2 if the hardware supports that output. Someone more informed about Switch 2's rumored CPU capabilities and the CPU demands of older games could perhaps point to a specific generation of games that would fit this criteria.

It's likely that the A78C in the Switch 2 is going to be tremendously better than the PS4 Jaguar CPU. Maybe not 4x better, but we might be surprised.

I think 4-5x is actually about what you'd expect.

This benchmark is a little hard to read. You've got the 8 core Orin NX variant clocked at 1.98 Ghz. Technically, Switch 2 is a slightly better performing CPU, and we don't know final clock speeds, but I think this is a really good estimate for what Switch 2's CPU perf would look like.

The other half of the benchmark is the A9-9820, which was a super-rare, Chinese only desktop kit from AMD. It turns out, it was literally AMD trying to fob off old Xbox One APUs they had lying around. So exactly the same CPU as the PS4/Xbox One... except the default clock here is much higher than either machine by about 30%.

So while this benchmark shows a 3.5x leap, in practice it's probably closer to 4.5x.

 

[OctoSplattack]

Hmm I feel like they will come up with at least some marketing explanation why certain games have higher performance or better resolution than others. Especially if BC is a thing and not every game (likely most BC games, maybe even some new releases) would be using DLSS or other possible hardware features.

It could be a label with something as simple as "optimised / improved for Switch NG". Power will likely be not the focus overall but I can see them using comparison videos and such if they made games like for example BOTW look and run way better because it is a strong selling point.

Optimised versions of BC Games maybe even could be tied to a NSO subscription.

I can see them advertising upgraded versions of older Switch games, but I don't see why they have to have a marketing explanation for why some games look better than others. Isn't that just a situation that literally every piece of hardware that plays games has dealt with?

 

[karmitt]

Already happening, lol…

From “elsewhere”:

“I hate this future where we’ve just decided upscaling everything is fine instead of asking for more native resolution performance (within reason)”

I saw this post and damn it’s the dumbest thing I’ve seen in a bit lol. Yes please let’s arbitrarily put value in this older version of a machine producing output over this newer one. Upscaling efforts are brilliant and should be championed.

 

[Estew]

I was moving and getting internet set up and saw... 20 pages have passed since I was last able to check in. Guess this is gonna be a fun read.

 

[kirbyfan407]

It's likely that the A78C in the Switch 2 is going to be tremendously better than the PS4 Jaguar CPU. Maybe not 4x better, but we might be surprised.

If you're asking if a developer could, with a custom port designed to use these technologies, take a PS4 era game and make it 4k120fps on something roughly Switch 2 performance, with DLSS?

Theoretically yes. Would it happen? I would bet money against it, for myriad technical reasons, but also because I think it would look like vaselined donkey ass.



I think 4-5x is actually about what you'd expect.

This benchmark is a little hard to read. You've got the 8 core Orin NX variant clocked at 1.98 Ghz. Technically, Switch 2 is a slightly better performing CPU, and we don't know final clock speeds, but I think this is a really good estimate for what Switch 2's CPU perf would look like.

The other half of the benchmark is the A9-9820, which was a super-rare, Chinese only desktop kit from AMD. It turns out, it was literally AMD trying to fob off old Xbox One APUs they had lying around. So exactly the same CPU as the PS4/Xbox One... except the default clock here is much higher than either machine by about 30%.

So while this benchmark shows a 3.5x leap, in practice it's probably closer to 4.5x.

I hadn't realized the CPU leap was expected to be that significant and that far ahead of the PS4/XB1 generation. If all these rumors and leaks pan out for the final design, I guess the next Nintendo hardware is going to be a major leap forward for them. I'm curious how their internal studios will handle it.

 

[LiC]

I saw this post and damn it’s the dumbest thing I’ve seen in a bit lol. Yes please let’s arbitrarily put value in this older version of a machine producing output over this newer one. Upscaling efforts are brilliant and should be championed.

There's a lot of bad upscaling out there, and depending on your exposure to it I can see why people would be wary. My go-to example is the spatial-only DLSS 1 and how awful, hideous, and hallucinatory it was. It was so bad it made me doubt the potential of DLSS 2 for a long time.

But people are being unreasonable when they don't care to look at what the practical reality of good upscaling looks like, and just believe it inherently leads to bad results because it's faking something as opposed to native rendering. And then a lot of people are just console warriors or PC master race morons who think of upscaling as cheating because it lets weaker hardware look better than they think it's allowed to.

The second group of people can't be reasoned with. For the first, I would say they need to realize that -- besides the fact that temporal upscaling done right can 100% pass the eye test and technical arguments are unnecessary -- all 3D rendering is "faking it," because there's a finite number of pixels on a screen, and the process of choosing what to put in them is all about fakery and compromises to make it look good. That's what "sampling" is. The idea that upscaling by fiddling with the output after it's rasterized is somehow more fake and bad than, like, anti-aliasing is silly.

 

[Lancelot]

If you're asking if a developer could, with a custom port designed to use these technologies, take a PS4 era game and make it 4k120fps on something roughly Switch 2 performance, with DLSS?

Theoretically yes. Would it happen? I would bet money against it, for myriad technical reasons, but also because I think it would look like vaselined donkey ass.



I think 4-5x is actually about what you'd expect.

This benchmark is a little hard to read. You've got the 8 core Orin NX variant clocked at 1.98 Ghz. Technically, Switch 2 is a slightly better performing CPU, and we don't know final clock speeds, but I think this is a really good estimate for what Switch 2's CPU perf would look like.

The other half of the benchmark is the A9-9820, which was a super-rare, Chinese only desktop kit from AMD. It turns out, it was literally AMD trying to fob off old Xbox One APUs they had lying around. So exactly the same CPU as the PS4/Xbox One... except the default clock here is much higher than either machine by about 30%.

So while this benchmark shows a 3.5x leap, in practice it's probably closer to 4.5x.

Is that actually single core? If it is then... It's huge to say the least.

 

[Multiverse]

Translated via Twitter's built-in translation:

yay Joy-Con drift gone

Wasn’t the new patent something else entirely? Someone here already clarified:

Cool find. Thanks for sharing. Hall effect analog stick isn’t new, and there are already prior patents. This new patent is about an analog stick that contains “a resistance section using a magnetorheological fluid whose viscosity changes in accordance with an intensity of a magnetic field”. So after you move the stick and it bounces back, the feel and velocity of the feedback force is changeable dynamically via an electromagnetic field. The patent is not about eliminating drifts, which has been solved by other patents.

Edit: One example I can think of is when Mario is swimming in a watery stage, the stick can be made feeling sluggish.

 

[Serif]

I saw this post and damn it’s the dumbest thing I’ve seen in a bit lol. Yes please let’s arbitrarily put value in this older version of a machine producing output over this newer one. Upscaling efforts are brilliant and should be championed.

They would have some kind of point if upscaling was always worse than native (can at least chalk it up to preference). Indeed I'm not fond of a lot of the oversharpening that some solutions perform or the lack of simple integer scaling options. But DLSS has demonstrated to be equal or better. At this point they're literally arguing against free performance. That's like not wanting to download more RAM. Who wouldn't want to download more RAM?

 

[TLZ]

Ooh some of my friend on my discord server spotted a new patent published by Nintendo for a analog stick that mentions "magnetic field generation" ie hall effect

 

[ReyVGM]

So who has started to save up for this thing, got a few months but yeah hope its around 300 dollars that will be the sweet spot price

I'm buying $10 worth of Amazon gift cards each month. I already have $100 saved up.

 

[Estew]

I was moving and getting internet set up and saw... 20 pages have passed since I was last able to check in. Guess this is gonna be a fun read.

Caught up now I think and holy crap this system sounds incredible.

 

[OctoSplattack]

Actually, I think those naysayers have a point. The Switch NG ain't gonna be an honest console for honest folk. Using DLSS all willy-nilly like that to manipulate the system just ain't right. It'll be a console built for gangsters, thieves, and hooligans. Not like the PS5 and Xbox Series X|S, which are made for upstanding citizens who know the noble value of native resolutions and raw power. Nintendo and Nvidia, I dare say, are cheater-cheater pumpkin eaters, and we shouldn't stand their tomfoolery.

 

[SoldierDelta]

Actually, I think those naysayers have a point. The Switch NG ain't gonna be an honest console for honest folk. Using DLSS all willy-nilly like that to manipulate the system just ain't right. It'll be a console built for gangsters, thieves, and hooligans. Not like the PS5 and Xbox Series X|S, which are made for upstanding citizens who know the noble value of native resolutions and raw power. Nintendo and Nvidia, I dare say, are cheater-cheater pumpkin eaters, and we shouldn't stand their tomfoolery.

REBEL AGAINST THE SYSTEM!! BUY THE NEXT-GEN NINTENDO SWITCH ENTERTAINMENT SYSTEM TO OWN THE MAN!! GRAAHHHH!!!

 

[eye0h]

Actually, I think those naysayers have a point. The Switch NG ain't gonna be an honest console for honest folk. Using DLSS all willy-nilly like that to manipulate the system just ain't right. It'll be a console built for gangsters, thieves, and hooligans. Not like the PS5 and Xbox Series X|S, which are made for upstanding citizens who know the noble value of native resolutions and raw power. Nintendo and Nvidia, I dare say, are cheater-cheater pumpkin eaters, and we shouldn't stand their tomfoolery.

Found the copypasta I’m using for the next 6 years

 

[Lugia667]

Actually, I think those naysayers have a point. The Switch NG ain't gonna be an honest console for honest folk. Using DLSS all willy-nilly like that to manipulate the system just ain't right. It'll be a console built for gangsters, thieves, and hooligans. Not like the PS5 and Xbox Series X|S, which are made for upstanding citizens who know the noble value of native resolutions and raw power. Nintendo and Nvidia, I dare say, are cheater-cheater pumpkin eaters, and we shouldn't stand their tomfoolery.

so that means it can get "underground" games like Yakuza right?

 

[Deleted member 887]

I hadn't realized the CPU leap was expected to be that significant and that far ahead of the PS4/XB1 generation.

AMD did the same thing with binned PS5s that they did with the binned Xbox One, which means we can head to head that comparison as well. You can see the leap for Sony and Microsoft was even larger.

AMD basically made every bet wrong on the Jaguar CPU design. So part of that huge leap is because the last gen consoles had very bad CPUs. The current Switch CPUs, despite being little mobile cores from 11 years ago, are fairly competitive against them

Is that actually single core? If it is then... It's huge to say the least.

Single core and multi-core both have similar leaps against last gen, in this benchmark. Which matches with roughly what the theoretical leap is between those two, so I large expect it to play out in games.

If all these rumors and leaks pan out for the final design, I guess the next Nintendo hardware is going to be a major leap forward for them. I'm curious how their internal studios will handle it.

Locking in their hardware partner has been a huge help here I imagine. We know from the gigaleak that Nintendo was given demos of DLSS 2 well before it released, and started exploring building the console around it then, and there have been indications that Nintendo has been consolidating around one big internal engine. And not for nothing has this generation been a little longer than expected, giving them a little more time to adapt.

 

[karmitt]

There's a lot of bad upscaling out there, and depending on your exposure to it I can see why people would be wary.

This shouldn’t weigh into the conversation though, as we have real examples of how it can produce a better image with lower system requirements. It’s indisputable that it’s a great direction for these engineers to pursue.

I said the efforts should be championed, not that every effort was universally superior. DLSS being a highlight for the next Switch, which this post was targeting specifically, is not remotely eligible for the poster’s disdain (or wariness as you put it)

Edit: and I know the rest of your post largely supports what I’m saying. I just had to call out that the original Era post is, in fact, very dumb.

 

[Dekuman]

 

[kirbyfan407]

AMD did the same thing with binned PS5s that they did with the binned Xbox One, which means we can head to head that comparison as well. You can see the leap for Sony and Microsoft was even larger.

AMD basically made every bet wrong on the Jaguar CPU design. So part of that huge leap is because the last gen consoles had very bad CPUs. The current Switch CPUs, despite being little mobile cores from 11 years ago, are fairly competitive against them

Single core and multi-core both have similar leaps against last gen, in this benchmark. Which matches with roughly what the theoretical leap is between those two, so I large expect it to play out in games.


Locking in their hardware partner has been a huge help here I imagine. We know from the gigaleak that Nintendo was given demos of DLSS 2 well before it released, and started exploring building the console around it then, and there have been indications that Nintendo has been consolidating around one big internal engine. And not for nothing has this generation been a little longer than expected, giving them a little more time to adapt.

I've repeatedly heard that the Jaguar CPUs were really bad, but your numbers are putting into perspective how bad they truly were.

I didn't realize that Nintendo was given an early preview of DLSS 2. I'm sure the Nvidia partnership is helping a lot, and it's clear both companies are heavily invested in it (so far). I suppose my curiosity does come from a place of concern due to my understanding that higher fidelity has led to a lot of slowed production and increased cost for developers over the past 15ish years. It'll be interesting to see how much longer the game credits for Nintendo games grow and how their release cadence changes in the next generation. They know their software plays a major role in their success, so I'm sure they've spent years thinking through these challenges and preparing for them.

 

[Serif]

Using DLSS all willy-nilly like that to manipulate the system just ain't right.

Admittedly hardware discussion has kind of felt this way with the 'real' power of the system stated as its capabilities 'before DLSS'. (Not here, necessarily. Specifying before/after DLSS is helpful to clarify comparative expectations and is not a dig at the system). But it is framed that way, with DLSS treated as this kind of afterthought or bolt-on as opposed to a part of the console's features, in fact it's not surprising to hear suggestions of a dock upscaler as DLSS is associated with generating very high-res visuals after the fact when it could still be put to good use in handheld mode. I'm not shocked that people think this way, since it does reflect how it is on PC, where it's not guaranteed every player has an RTX card so a DLSS implementation would only benefit a subset of the audience, and the focus is usually developing a game where the minimum requirements don't require that. Not really something a game is built around.

This is the first time we have tensor cores on a console, where every dev developing a 3D game has access to DLSS in some way, even if they have to take steps to inject it. And they know every player will benefit from it no matter what mode they use. This is exciting to think about it. I don't expect every dev to use it since there is work involved (or it's impossible, as in 2D games) but - if a 3D game ships on Switch 2 without DLSS, no matter how well it performs, I will not consider it a 'true' reflection of the system's power, because it's not even leveraging the system's full capabilities and the hardware acceleration*. As well said in this thread before, all graphics are 'fake', ultimately mosaics of teeny tiles meant to fool you. That's why there's variable rate shading, foveated rendering, and AI upscaling, to further take advantage of how easily fooled we are. So as far as I'm concerned the next 3D Mario will be in real 4K and it's time to dream big.

*I am speaking somewhat passionately here, I know this is engine dependent, though I have yet to see an implementation of DLSS where it looks bad or worse than just bilinear upscaling the lower res, even if there are artifacts

 

[My Tulpa]

Well not really, DLSS need tensor cores. The better those are, the better DLSS works. And the DLSS 3 feature, frame generation, needs specific hardware (only available on their 40serie cards).

So even if Switch2 is based on Ampere, I really hope the rumors are true and that Nvidia and Nintendo made it as future proof as possible with some mix-up Lovelace tech, like 4th gen tensor cores.

That will makes the difference it the long run, much more than just raw power.

T239 (switch Drake) would be gen 3 tensor cores.

The rtx 2000 gpus are gen 2 tensor cores and have benefited from every advancement in DLSS. EXCEPT only for frame generation.

I wouldn’t worry about this new Switch losing most support and DLSS advancement upgrades for another 10 years or so.

 

[Look over there]

Alright, we've slowed down enough that I can drop off one of the things I wanted to remark on, since it'll be fairly quick:

Baldur's Gate 3 on PS5 is effectively the PC version at ultra settings

Baldur's Gate 3 is one of the most highly acclaimed games of the year, marrying the best of video games with the comple…

www.eurogamer.net

So not surprisingly, the toughest parts of Act 3 bring the PS5 version's FPS down quite a bit. Interestingly, it can be dragged down to the 20's, while apparently their testing for the Ryzen 3600x only got down to ~low 30's?
That emboldens my suspicion that it's the increased NPC density/AI scripts running. AI scripts give me the impression that they're the sort of thing that would be constantly accessing recent data.
Remember that the PS5 uses monolithic based Zen 2, so a given core has access to only the 4 MB of L3 cache within the CCX it's in. 3600x, being chiplet based Zen 2, has 16 MB of L3 cache for each CCX. Then there's also the latency difference between using GDDR (like the PS5) and regular DDR (like usual desktops).
So, I think that what we're seeing is a massive spike in spilling out of cache and into RAM. The 3600x would seem to suffer somewhat less badly due to having more L3 cache to pull together a higher hit rate plus being less penalized by going out to RAM in the first place.
(of course, what tipped me off to think in this direction in the first place is seeing how much better the X3D chips were doing)

To bring it back to Nintendo and hypothetical ports; I think that the expected 8 MB of L3 cache and LPDDR ram would mitigate things a bit as far as how bad the worst situations can drag things down from the baseline. But I don't want to oversell it; additional clever work should still be needed.