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

[darthdiablo]

About the scalping issue, we have to think outside the box. Scalpers exist because people buy from them and they make profit. These scalpers are only there for the money. Now, let's think "who" buys from them. Typically the diehard fans and tech enthusiasts that couldn't buy the product, basically the hardcore side of the consumer. What do these guys most likely have? A Nintendo account. What if Nintendo makes the Switch 2 available for preorders online (from their store, making Nintendo accounts able to link with Amazon, etc) and prioritize orders from accounts that are ~6 months old, while limiting 1 console per account? This way, they can decrease the amount of people that would potentially buy from scalpers.

There's an easier way to address scalping: Making sure supply is plentiful - in such a way that scalpers cannot reasonably snap up the entire supply.

This has happened before - I cannot recall which console it was (it might have been Switch actually? - edit: nah, it was PS5's) but the scalpers got burnt HARD because they bought tons of consoles and had trouble selling them, because people were able to find console easily at stores.

Supply is key. For this reason I think Nintendo might start mass production and make reveal sooner than October 2024.

 

[lattjeful]

It’s only positive news for me.

I’ve been waiting for that threshold where my major issues with game visuals are resolved - lower resolutions being near the top, followed by poor framerates. For Nintendo’s purposes, I really think this is it. This is where I’m expecting real diminishing returns moving forward, and that’s a great thing. I don’t want the next Wind Temple to have me reeling as the image turns to garbage during the crescendo of a boss fight. I’m still a little bitter about Tears of the Kingdom having to be enjoyed in that state.

Maybe I’m being overly optimistic, but I expect to care far less about technical advancements on the next-next hardware from Nintendo, at least with respect to Nintendo’s own games and AA game quality. The AAA third party scene is reaching an impasse of sorts, and for all we know the majority are forced to stop raising the bar because it’s just costing too much. That could see Switch 2/3 being right where it needs to be, even if it’s not at the bleeding edge.

Agreed. Low resolution, poor draw distance/texture filtering (honestly the bad texture filtering in a lot of Switch-era Nintendo games is more distracting than the lower resolution overall LOL. For me I think it's the lack of consistency that bad texture filtering has VS a game generally being low res), and soupy texture work are my big complaints when it comes to bad graphics. As far as Nintendo games go, I expect all of those to be resolved with the Switch 2 and that's all I need.

And even then, very few games got distractingly bad. The only games that come to mind are Tears of the Kingdom's Water Temple with its dodgy framerate (V1 Switch here, so TotK generally isn't as stable across the board) and distracting aliasing, and Xenoblade Chronicles 3 Erythia Sea with its general blurriness and poor texture filtering got me. I don't even think they look bad though tbf, I think it's just that the other 99% of each game looks so good that during the 1% where the Switch's aging hardware rears its ugly head, it ends up being distracting.

 

[Serif]

maybe debunked is the wrong word but there does seem to be a trend to knee-jerk dismiss things or deem them unlikely if a source it deemed not trustworthy enough. or maybe they got something wrong in the past so it's easier to ignore it if the information is 'out there' or doesn't align with the most positive outcome ie the node. that's fine and all but sometimes it's ok to believe and be wrong, likewise the skeptics can (and have) been wrong too. sometimes it's easier to be a skeptic when we're dealing with things that can't be immediately proven. i guess i see things more as ok that's possible and work backward from there ie the magnet thing that sounded very Nintendo and believable from the get.

The reaction to 8nm rumors is knee-jerk at this point because we've had similar discussions for years now, and there hasn't been a credible leak one way or the other, just page of analyses. I guess I don't see what the problem is with dismissing untrustworthy sources. Sometimes the accuracy of information is not related to how trustworthy they are, nor how desirable the info is. I think we've scrutinized every source that claimed this matched or exceeded the Series S in raw performance, even though we would all love that. And it is possible for a 'leaker' to make a very good guess or be accidentally right.

It is indeed ok to believe and be wrong - the truth is the system is not out so nothing is final yet no matter how good the info is. All I personally wish is that people would weigh info when participating in the discussion. I think it's fair to evaluate any new claims based on what we've gathered so far - like if a person says that Nintendo switched to AMD, or is making a home console. Not technically impossible, but we can reasonably dismiss these. We cannot completely rule out 8nm, but we can rigorously question it.

I don't have a good memory of the magnets conversation but I don't think it was because magnets sounded implausible, but that the source of information seemed dubious. Magnets are indeed a Nintendo thing, theyve Incorporated it into all their clamshell designs. I know some people doubted how safe they were with the pinching fingers, but that is something I think Nintendo can figure out.

 

[HankyPanky]

There's an easier way to address scalping: Making sure supply is plentiful - in such a way that scalpers cannot reasonably snap up the entire supply.

This has happened before - I cannot recall which console it was (it might have been Switch actually?) but the scalpers got burnt HARD because they bought tons of consoles and had trouble selling them, because people were able to find console easily at stores.

Supply is key. For this reason I think Nintendo might start mass production and make reveal sooner than October 2024.

much sooner, i'd imagine. in my mind the only question is whether it's July, August or September with more emphasis on the earlier months. as soon as this June Direct is out of the way it'll be all systems go, methinks. it's been implied this is the last time they'll focus on the OG Switch so it's not a stretch to believe the reveal could be coming shortly after.

 

[ILikeFeet]

as I said before, if Digital Foundry had insider information, it would be its own article/video and not them speculating in their video case. so at this point I don't think DF knows any more than we do.

and that wouldn't be unusual as, if I remember correctly, their details on Switch didn't come until around the proper reveal of the system in October 2016

---

I'm seeing a lot of people believe in 4TFLOPS in docked mode, and I don't think I can bring myself to believe that. now with the Series S hitting 4TFLOPS at a much higher power setting. that would imply dynamic clocks and Drake would downclock often to fit in the power budget

 

[Concernt]

PlayStation's Focus Has Shifted From Game Sales to Actual Play Time, Sony Says - PlayStation LifeStyle

Sony has said that due to changes in consumer behavior in PS4 era, Sony's more focused on increasing PS5 play time rather than selling games.

www.playstationlifestyle.net

This makes sense but is another notable change in the AAA industry. Nintendo has apparently been tracking engagement for a long time, but without significant monetised live service elements, they probably don't profit in the same way. Unlike Sony, they also don't provide notable current generation games through their subscription service.

 

[Dekuman]

as I said before, if Digital Foundry had insider information, it would be its own article/video and not them speculating in their video case. so at this point I don't think DF knows any more than we do.

and that wouldn't be unusual as, if I remember correctly, their details on Switch didn't come until around the proper reveal of the system in October 2016

---

I'm seeing a lot of people believe in 4TFLOPS in docked mode, and I don't think I can bring myself to believe that. now with the Series S hitting 4TFLOPS at a much higher power setting. that would imply dynamic clocks and Drake would downclock often to fit in the power budget

Didn't rich work with Fami members to create his T239 target specs/performance video?
It pretty much confirms they don't have insider info or aren't hearing anything substantially different.

Rich's reference to node process seems to be just riffing on Kopite or MLID's own guesses which is the worst case of laundering speculation through a trusted source to turn it into a fact.

 

[Koti]

There's an easier way to address scalping: Making sure supply is plentiful - in such a way that scalpers cannot reasonably snap up the entire supply.

This has happened before - I cannot recall which console it was (it might have been Switch actually? - edit: nah, it was PS5's) but the scalpers got burnt HARD because they bought tons of consoles and had trouble selling them, because people were able to find console easily at stores.

Supply is key. For this reason I think Nintendo might start mass production and make reveal sooner than October 2024.

That is also important, but a good marketing method is utilizing scarcity. Scarcity increases demand for the product, it creates hype and discussion. This is especially important to appeal to the casual consumer. I call it the Tamagotchi scheme, but there's probably a proper name. This is the reason why the Switch kept momentum at launch and during the covid times. Same with the current-gen consoles, but those suffered from actual stock issue like chip shortage. Balance is important. Nintendo should produce lots, but they shouldn't sell all of them immediately to retailers and market.

 

[RedSpring]

maybe debunked is the wrong word but there does seem to be a trend to knee-jerk dismiss things or deem them unlikely if a source is deemed not trustworthy enough. or maybe they got something wrong in the past so it's easier to ignore it if the information is 'out there' or doesn't align with the most positive outcome ie the node. that's fine and all but sometimes it's ok to believe and be wrong, likewise the skeptics can (and have) been wrong too. it's much easier to be a skeptic when we're dealing with things that can't be immediately proven. i guess i see things more as ok that's possible and work backward from there ie the magnet thing which sounded very Nintendo and believable from the get.

Ninspider 's one word posts got pages of discussion so calling it a knee jerk dismissing is a bit unfair. Nintendo stated that they tried magnet and didn't like it with the Switch. Maybe they discovered a solution but it's easy to be skeptical of the post given the context. It could be that magnet refer to some other usage but it's hard to tell, it's a one word post. For other speculation, I personally didn't dismissed them for being negative. I dismissed them for being contradictory (e.g. 4TF but on 8nm) or completely goes against much more reliable source (e.g. 10SM T239 compared to 12SM straight from NVIDIA's data)

re the Moore is Dead guy, he did make a couple of videos that addressed the 8nm thing. there were a ton of people in the comments fighting the good fight presenting some of the arguments here about 8nm not being viable due the power usage. in the second video he came back to reiterate from one of his sources that it was indeed 8nm and presented this as fact from someone who knew and wasn't guessing.

whether any of that is true or trustworthy or simply furthering a narrative is another thing but he was claiming that at least one his sources (maybe there were three in the end i forget) knew this information.

He also come back later with the 4TF docked claim. Either one of his source is lying and he doesn't have the technical knowhow to see the contradiction or he just repeat anything he's told or Nintendo did magic with 8nm which is fine with me.

 

[ILikeFeet]

Didn't rich work with Fami members to create his T239 target specs/performance video?
It pretty much confirms they don't have insider info or aren't hearing anything substantially different.

Rich's reference to node process seems to be just riffing on Kopite or MLID's own guesses which is the worst case of laundering speculation through a trusted source to turn it into a fact.

yea. until they put out their own article/video, their current discussions is going to go around the same circles as we go through. they've kinda exhausted talking points already, if you ask me.

if they wanted a better speculative test, I think they could skip the 2050 and go straight to mobile testing now lumen is available for android and there are a couple of games with RT out there

 

[HankyPanky]

re the Moore is Dead guy, he did make a couple of videos that addressed the 8nm thing. there were a ton of people in the comments fighting the good fight presenting some of the arguments here about 8nm not being viable due the power usage. in the second video he came back to reiterate from one of his sources that it was indeed 8nm and presented this as fact from someone who knew and wasn't guessing.

whether any of that is true or trustworthy or simply furthering a narrative is another thing but he was claiming that at least one his sources (maybe there were three in the end i forget) knew this information.

 

[HyruleanKnight]

as I said before, if Digital Foundry had insider information, it would be its own article/video and not them speculating in their video case. so at this point I don't think DF knows any more than we do.

and that wouldn't be unusual as, if I remember correctly, their details on Switch didn't come until around the proper reveal of the system in October 2016

---

I'm seeing a lot of people believe in 4TFLOPS in docked mode, and I don't think I can bring myself to believe that. now with the Series S hitting 4TFLOPS at a much higher power setting. that would imply dynamic clocks and Drake would downclock often to fit in the power budget

4TF may not be real, but I don't think it's fair to compare with the series S either. That one has an x86 CPU in there, along with high-powered GDDR6, and it's on 7nm. Power efficiency was never a consideration in it's design, but then again according to MS's own spec sheet a typical gaming session should pull no more than 75W in total for the whole system.

If I wasn't thinking about mobile hardware, I would've been very impressed.

 

[Serif]

If Nintendo created an 8nm device with the performance characteristics of 5nm, then cool. I wouldn't care that it's 8nm then. If someone like MLID is claiming 4TF and 8nm and that ends up being true, and the device was somehow still the thickness of the current Switch, cool. But I'm not convinced that this scenario will happen. If it's 8nm, something's gotta give for a GPU of this size.

 

[oldpuck]

Maybe but I have troubles seeing Nintendo going with A78AE instead of A78C for Switch Pro.

Yeah, they're never going to use AE cores unless they wind up coincidentally in an off-the-shelf SOC they want to use. AE has complex features they could never need, but is also missing some highly useful features of C that I wouldn't want to give up

 

[darthdiablo]

That is also important, but a good marketing method is utilizing scarcity. Scarcity increases demand for the product

This is how scalping takes hold in the first place - low supply or artificial scarcity.

No, the issues that comes with artificial scarcity are not worth it. Why take consoles away from buyers, the same buyers who would have bought games if they had a console?

The "create demand via artificial scarcity" scenario never made sense to me, not when consoles are involved, because the fewer consoles that are out there, the fewer # of games they can sell. Games have significantly bigger profit margins compared to consoles.

 

[ExoticSpice]

Guy I replied to was talking about Macbook Pro, and those use standard registered DDR4 2933 chips

Don’t base M chip MacBooks Pros use LPDRR5 6400? Soon be LPDRR5X whenever they release the M4 Pro/M4 Max. Also the higher end pros have insane bandwidth for laptops, probably adds to the BOM as well.

I think the Intel’s MacBooks did use standard DDR4 RAM for the Pros.

Either way, you are correct about the upsell. Apple is very good at it.

 

[Kise Ryota]

8nm is simply impossible, 4N is most likely.

I agree 4N seems to be the most reasonable node for that SoC, but SEC8NM can work.

Using Thraktor's numbers from the Jetson Tool, we could have:

GPU - 420.75MHz - 5.70W
CPU - 1113.6MHz - 2.2W

These are the lowest clocks the tool would show.
That's 7.9W, and I think they could shave it a bit for having the clock gate backported from Ada, and also for CPU having 8 cores in one single cluster (differently than the Orin chip). Maybe even other improvements because it's a custom chip, not an Orin that wasn't exactly made with extreme power consumption in mind.

But let's say they manage to only shave 0.9W. Then we have 7W for the SoC, and let's add 4W for the rest of the system. In order to get the same battery life of the 2017 model, we need a 6757 (3.7v) mAh battery. Not absolutely impossible. Also, the cooling on the Switch v1 already showed that it can handle a much higher power consumption than what Nintendo went with. With a small upgrade, it could handle docked mode just fine.

So, it's not really impossible. And if you believe they possibly could double the battery capacity (from the V1 model), then you could even have a battery life closer to the V2 model with the clocks above (or have the V1 battery life with better clocks).

Now, do I like these clocks? No.
Do I think it makes more sense to go with 4N, so they can squeeze all what the silicon they are paying for can delivery? Yes, I do (even though they could also not squeeze it 100%, giving up a bit of performance for more battery life)

But SEC 8nm was only impossible when this thread was looking at a Switch 2 being exactly the Switch OLED's case but with the new hardware inside of it, and having the same power consumption/battery life of the V1 model.

 

[heyonme]

But I'm talking about last year, long before the delay to 2025 news.

I get being dismissive of certain forms of pessimism, but the confidence people have where they think they know better than what seems like a fairly large number of Tech Insiders who seem to be pointing towards NIntendo going for the cheap short term route when it comes to node size just seems odd.

The tone and general direction of the thread has been pretty shocking in recent days. A group of power users who get high off their own supply have been so effective in cementing a certain set of beliefs that everyone seems so angry when a dissenting thought or opinion is presented. A lot of the commonly accepted 'facts' are rooted in leaks and shaky information that are years old at this point. This community has long had a problem of excluding and chasing away people who don't fall in line with the power users and it seems worse than ever now...

 

[Jersh]

V1 Switch here, so TotK generally isn't as stable across the board

Can someone remind me why this is again? People are quick to point out the RAM modules in newer Switches are higher bandwidth, but Nintendo runs them at the same clocks on all Switches so that's not actually true in practice. Is it just lower latency/faster timings?

 

[entut1]

Congratulations, @LiC , you've made it to Arlo's latest video XD.

 

[Dimensio]

Looks like my guess was right it's another Mario Party, I assume for the Switch 2

 

[Concernt]

Congratulations, @LiC , you've made it to Arlo's latest video XD.

OH I HATE THAT THUMBNAIL.

REALLY, ARLO?! Come ON. If you're reading this stuff you'd know Series S is a more salient comparison because the features are more comparable!

Bwuh, bwuh I say!

Edit: and the mistakes come rolling in...

 

[Steve]

Looks like my guess was right it's another Mario Party, I assume for the Switch 2

Whelp, we pretty much know what the big holiday release will be.

Maybe there’s some hope for Metroid prime 4

 

[ComradeKoopa]

Looks like my guess was right it's another Mario Party, I assume for the Switch 2

Man I'd rather they just add more boards and minigames to Superstars, but I'll take what I can get.

 

[Lugia667]

will Arlo crash the site again? lol

 

[redmutineer75]

Whelp, we pretty much know what the big holiday release will be.

Maybe there’s some hope for Metroid prime 4

Metroid is way too niche to ever be the big holiday title. I expect an August or September release for it, then Mario Party in November or December.

 

[Steve]

Man I'd rather they just add more boards and minigames to Superstars, but I'll take what I can get.

Maybe it’s a club house 51 type of releas?

But Mario party is more likely

 

[imfake2]

Looks like my guess was right it's another Mario Party, I assume for the Switch 2

tbh it’s probably this year they could do a really quick rollout for a mario party like they did for wonder

 

[fsshariq]

Really looking forward to seeing Gamers Nexus tear this thing apart upon release. Hopefully they can go even further in depth vs their OG Switch teardown given the expansion of their testing equipment.

 

[carbvan]

OH I HATE THAT THUMBNAIL.

REALLY, ARLO?! Come ON. If you're reading this stuff you'd know Series S is a more salient comparison because the features are more comparable!

Bwuh, bwuh I say!

Edit: and the mistakes come rolling in... * Hidden text: cannot be quoted. *

This video sucks so bad. He really says he never expected the Switch 2 to be more powerful than an Xbox One because Nintendo "is always 1 generation behind". Ffs, how the hell would The Witcher 3 or Doom Eternal run on a 360? And then he just doesn't bother to do the research to understand what any of the actual specifications means. Arlo is pretty entertaining when it comes to his videos on actual games, but his hardware videos make me wanna bash my head into a wall.

 

[Cheetah6442]

Whelp, we pretty much know what the big holiday release will be.

Maybe there’s some hope for Metroid prime 4

If it's a spinoff, which it most likely will be, then I doubt it's the big holiday title.

 

[Cheetah6442]

tbh it’s probably this year they could do a really quick rollout for a mario party like they did for wonder

What do you mean by a quick rollout?

 

[Concernt]

This video sucks so bad. He really says he never expected the Switch 2 to be more powerful than an Xbox One because Nintendo "is always 1 generation behind". Ffs, how the hell would The Witcher 3 or Doom Eternal run on a 360? And then he just doesn't bother to do the research to understand what any of the actual specifications means. Arlo is pretty entertaining when it comes to his videos on actual games, but his hardware videos make me wanna bash my head into a wall.

Less powerful than an Xbox One would be a less than twice the performance of Nintendo Switch.

On the one hand I think the general public, Arlo included, will be in shock and awe when this thing comes out and is standing nearly toe to toe in output resolution with other consoles.

On the other hand WOW Nintendo will have a HELL of a time hammering in the marketing of "yes this is a current generation home console you can take on the go, yes it's 4K, no it's not a generation behind". As much as they let the games do the talking I think they need to flat out make a statement about the performance

 

[gonanzir]

To be honest, I no longer care about the opinions of those who haven't paid attention to the Switch 2 leak regarding its performance; most of them just say 'Nintendo doing Nintendo'.

 

[redmutineer75]

If it's a spinoff, which it most likely will be, then I doubt it's the big holiday title.

Why would it be a spinoff? It has the main Mario Party developers on it, and the last one was Superstars in 2021. A new mainline game in time for Holiday 2024 makes complete sense.

 

[Serif]

Switch 1 has been memed into sub PS3 performance and there is shock and awe when 15 year old games from that generation perform well, I am not surprised that people thought sub XB1 was in the cards for Switch 2.

I mean some people still think that even with these specs, but careful. If you merely question this expectation, you will be called a power user agonizing against a dissenting opinion. God forbid we have any kind of conversation or debate.

 

[OctoSplattack]

Congratulations, @LiC , you've made it to Arlo's latest video XD.

Among everything else, he also seems confused about what the new information even is, noting both the current shipping information that confirms "12 GB RAM/256 GB storage" and the old RedGamingTech specs that we were chatting about a few pages back, and being unsure which one is accurate. I know the videos on his TopicArlo channel are meant to be more casual, but I wish he took the time to double-check.

 

[darthdiablo]

haven't watched video yet, but Arlo isn't sure which is correct, RGT specs or shipment data? Yikes

 

[Cheetah6442]

Why would it be a spinoff? It has the main Mario Party developers on it, and the last one was Superstars in 2021. A new mainline game in time for Holiday 2024 makes complete sense.

I could be wrong, but I don't think Nintendo has never made more than two mainline Mario Party titles for a single system(at least not since maybe the GameCube days). They have been making spinoff Mario Party titles since the start of the 3DS/Wii U era. It would be surprising that the Switch don't even have 1.

 

[Dekuman]

Arlo isn't an insider, he just likes to make anticipation videos. His NX videos is what got my to watch him, not his whining about Paper Mario.
His videos shouldn't be relevant in this thread.

 

[oldpuck]

More than just clock speed.

8nm means shorter battery life. More bulky. More cooling required, gets hotter.

I wanna push back on this slightly, not because you're wrong, but because I think some folk - not you! - are seeing it in a really binary way.

8nm means all these thing, but not all at once - that's the "downward pressure" that darthdiablo is talking about. It's kinda like a waterbed (are waterbeds still a cultural reference that make sense anymore?) - you push down in one place, it comes up in another.

8nm can deliver the same performance as 4nm... just in a bigger, costlier device. 4nm can cost the same as 8nm - in a more expensive package. Some of these things Nintendo is flexible on, some of these less flexible on, and some have hard lines. Price absolutely has a hard line - look at Sony making more revenue than Nintendo but less profit - that's because Sony doesn't have a hard line and loses money on their hardware and screws themselves in the proces.

The only advantage that 8nm provides is cost. If we wind up with 8nm it's actually a sign that Nintendo prioritized performance, because that let them get the most chip per dollar. If they go with 4nm, then performance might come along for the ride, purely as a side effect of clock speeds, but that would make me think they prioritized battery life, since they went with the most power efficient option per dollar.

 

[OctoSplattack]

What do you mean by a quick rollout?

Probably meant quick announcement-to-release. Which I feel is most Nintendo games these days, but to look at NDcube specifically, Super Mario Party and Superstars were "June announcement, October release", while Clubhouse Games: 51 Worldwide Classics and Everybody 1-2 Switch (which they co-developed with EPD4) both had one month windows.

 

[tetsugo]

On the other hand WOW Nintendo will have a HELL of a time hammering in the marketing of "yes this is a current generation home console you can take on the go, yes it's 4K, no it's not a generation behind". As much as they let the games do the talking I think they need to flat out make a statement about the performance

Yeah, this will be their big challenge to promote it as having something to appeal all the fathers, kids and not just the Nintendo hardcore fans out there. The switch communicated it pretty well on the marketing the new stuff that made Switch really appealing.

I know that Nintendo gave up on doing techdemos, since those set pretty bad expectations for the Wii U, but maybe they will have to do one of those Zelda techdemos that showcase what the hardware can do, and how it will be significantly better than previous generation. Honestly, I don't how Nintendo will handle the marketing on this thing to avoid people talking things like "I already have a Switch, why I'd want this thing that does exactly what I've playing for the last 7 years, but with just better graphics?". I know that for us, is easy to see this as a dream coming true, but I don't know if this is exactly what the big mass really wants.

 

[BreakAtmo]

I agree 4N seems to be the most reasonable node for that SoC, but SEC8NM can work.

Using Thraktor's numbers from the Jetson Tool, we could have:

GPU - 420.75MHz - 5.70W
CPU - 1113.6MHz - 2.2W

These are the lowest clocks the tool would show.
That's 7.9W, and I think they could shave it a bit for having the clock gate backported from Ada, and also for CPU having 8 cores in one single cluster (differently than the Orin chip). Maybe even other improvements because it's a custom chip, not an Orin that wasn't exactly made with extreme power consumption in mind.

But let's say they manage to only shave 0.9W. Then we have 7W for the SoC, and let's add 4W for the rest of the system. In order to get the same battery life of the 2017 model, we need a 6757 (3.7v) mAh battery. Not absolutely impossible. Also, the cooling on the Switch v1 already showed that it can handle a much higher power consumption than what Nintendo went with. With a small upgrade, it could handle docked mode just fine.

So, it's not really impossible. And if you believe they possibly could double the battery capacity (from the V1 model), then you could even have a battery life closer to the V2 model with the clocks above (or have the V1 battery life with better clocks).

Now, do I like these clocks? No.
Do I think it makes more sense to go with 4N, so they can squeeze all what the silicon they are paying for can delivery? Yes, I do (even though they could also not squeeze it 100%, giving up a bit of performance for more battery life)

But SEC 8nm was only impossible when this thread was looking at a Switch 2 being exactly the Switch OLED's case but with the new hardware inside of it, and having the same power consumption/battery life of the V1 model.

The problem with this is that you're trying to work out how to make the T239 design work on 8nm after the fact, when what Nintendo and Nvidia have actually done is decided the node first and then designed T239 around that. So we have to look at the chip design for answers, and they all point to 4N, because on 8nm you'd be better off with 8SMs at higher, sweet-spot clocks, especially since that chip would actually be cheaper.

Plus there's the fact that the system uses LPDDR5X and there's apparently no 8nm LPDDR5X controller, and it has 120GB of bandwidth which is way more than the weaker system you're talking about needs. Sure you could say "well, it's not impossible that they designed an 8nm LPDDR5X just for Switch2 so that they could use pointlessly high-bandwidth RAM when LPDDR-6400 would've been just fine", but at a certain point there's just too many not-impossibles and not enough actual evidence.

 

[lattjeful]

Can someone remind me why this is again? People are quick to point out the RAM modules in newer Switches are higher bandwidth, but Nintendo runs them at the same clocks on all Switches so that's not actually true in practice. Is it just lower latency/faster timings?

Yep, that's exactly it. Lower latency from the RAM modules.

 

[Dekuman]

The only advantage that 8nm provides is cost. If we wind up with 8nm it's actually a sign that Nintendo prioritized performance, because that let them get the most chip per dollar. If they go with 4nm, then performance might come along for the ride, purely as a side effect of clock speeds, but that would make me think they prioritized battery life, since they went with the most power efficient option per dollar.

wouldn't 4N also run cooler and perhaps forgo the need to have the rumoured fans on the dock?

 

[HyruleanKnight]

Don’t base M chip MacBooks Pros use LPDRR5 6400? Soon be LPDRR5X whenever they release the M4 Pro/M4 Max. Also the higher end pros have insane bandwidth for laptops, probably adds to the BOM as well.

I think the Intel’s MacBooks did use standard DDR4 RAM for the Pros.

Either way, you are correct about the upsell. Apple is very good at it.

You're right, I don't know why I thought the M series macbooks were still using DDR4 2933, was probably a brainfart moment

Even so, a single LPDDR5 6400 8GB module is somewhere between $30-35 if you were to buy a single unit. Apple definitely gets large OEM discounts though.

 

[HyruleanKnight]

I agree 4N seems to be the most reasonable node for that SoC, but SEC8NM can work.

Using Thraktor's numbers from the Jetson Tool, we could have:

GPU - 420.75MHz - 5.70W
CPU - 1113.6MHz - 2.2W

These are the lowest clocks the tool would show.
That's 7.9W, and I think they could shave it a bit for having the clock gate backported from Ada, and also for CPU having 8 cores in one single cluster (differently than the Orin chip). Maybe even other improvements because it's a custom chip, not an Orin that wasn't exactly made with extreme power consumption in mind.

But let's say they manage to only shave 0.9W. Then we have 7W for the SoC, and let's add 4W for the rest of the system. In order to get the same battery life of the 2017 model, we need a 6757 (3.7v) mAh battery. Not absolutely impossible. Also, the cooling on the Switch v1 already showed that it can handle a much higher power consumption than what Nintendo went with. With a small upgrade, it could handle docked mode just fine.

So, it's not really impossible. And if you believe they possibly could double the battery capacity (from the V1 model), then you could even have a battery life closer to the V2 model with the clocks above (or have the V1 battery life with better clocks).

Now, do I like these clocks? No.
Do I think it makes more sense to go with 4N, so they can squeeze all what the silicon they are paying for can delivery? Yes, I do (even though they could also not squeeze it 100%, giving up a bit of performance for more battery life)

But SEC 8nm was only impossible when this thread was looking at a Switch 2 being exactly the Switch OLED's case but with the new hardware inside of it, and having the same power consumption/battery life of the V1 model.

All this would've made sense, given we know next to nothing about clockspeeds and can only base our assumptions on current Switch's power envelope. Heck Nintendo could even clock the CPU below 1GHz and it'd still make sense.

The part that doesn't make sense is the 1536 shader cores on 8nm. You can check this out to know why: https://famiboards.com/threads/futu...staff-posts-before-commenting.55/post-1096639

 

[LinkURL]

wouldn't 4N also run cooler and perhaps forgo the need to have the rumoured fans on the dock?

No, if they really want to push clocks to the extreme, NVidia's new N4P GPUs are rumored to reach up to 2.9GHz.

 

[NineTailSage]

Oh, ok thanks for the correction.

Was I the only one who saw the actual video of MLID? He was guessing that it would have fan in the dock.

Yeah this is pretty much what I got from watching that video.
The problem I have with MLID is that he had some speculation on what he thought could be used in the dock to achieve close to 4TFLOPS and depending on what information comes out about final hardware. He will try to claim it as a win of him being right about what Switch 2 is capable of even if it ends up on a completely different node than he claimed...