• Hey everyone, staff have documented a list of banned content and subject matter that we feel are not consistent with site values, and don't make sense to host discussion of on Famiboards. This list (and the relevant reasoning per item) is viewable here.
  • Furukawa Speaks! We discuss the announcement of the Nintendo Switch Successor and our June Direct Predictions on the new episode of the Famiboards Discussion Club! Check it out here!

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

I think a simple "groups" button on the home screen would be nice, along with moving "All Software" to the leftmost slot (rather than the rightmost). Perhaps these could be two rectangles. A large "All Software" button with a smaller one on top of it for "Groups", together they form one home screen icon square. Juuust to the left of the first icon.

That, themes, and moving Bluetooth settings to Quick Home are my only real "asks" for the UI.

Speaking of moving "All Software" to the leftmost slot, I recently discovered I can hit d-pad left on the first icon and it wraps back around to "All Software". This console has been out since 2017 and I just learned that.
 
Imo the third party headliner should not be a last-gen / cross-gen game. Would love to see Elden Ring in a sizzle reel but give me a significant current-gen only game to shout it loud and clear 'this is a current gen portable console'.
 
My projected launch window for Switch 2
-Metroid Prime 4 (along with Switch version)
-Visions of Mana
- FC 2025
-Persona 3: Reloaded (with DLC)
-New Sonic game
-Mario Kart 10
March Launch

3D Mario
Metroid Prime 4
Ridge Racer 8 Exclusive
Various Third-Party Ports (RE9)

Fall-Holidays 25

Mario Kart 10
BOTW Deluxe
 
Imo the third party headliner should not be a last-gen / cross-gen game. Would love to see Elden Ring in a sizzle reel but give me a significant current-gen only game to shout it loud and clear 'this is a current gen portable console'.
Yeah, the innards of this system are pretty nice so there's no reason we couldn't have something new for 2025 like Monster Hunter Wilds as the technical third party showcase.
 
Imo the third party headliner should not be a last-gen / cross-gen game. Would love to see Elden Ring in a sizzle reel but give me a significant current-gen only game to shout it loud and clear 'this is a current gen portable console'.
Alan Wake 2
 
Node shrinks are getting harder, more expensive and producing diminishing returns for that huge investment.

It's not so much that it's not possible but there could be cheaper alternatives to get similar results.

If the chip is already pretty efficient and solid state batteries end up being cheaper to manufacture, it could be a far more cost effective way for Nintendo to produce a model with enhanced battery life.

Redesigning a chip on a new node, going through tape out and testing etc doesn't come cheap, why do that when you can just slap a battery in with triple the density that costs less than your standard lithium ion battery.
A valid argument. But I do stand by my point because:

When we say a node is "expensive," what we mean is the cost of a single wafer is high. Now, if I were to be able to put. let's say, 50% more transistors within the same area, that means I'm theoretically saving 33% die space. Now, actual number might be a little below that but just bear with me for a bit. More transistors within the same area means smaller chip, right? So while the cost of a single, 4 year old wafer may still be "high" by today's standards, the cost per chip isn't going up proportionally because we're making more chips per wafer. Makes sense?

You also need to consider than by 2030 4N may not even be commercially viable, as it will be 8 year old technology and basically antiquated. The cost of keeping up 4N production will eventually exceed that of a shrunk down 3E chip. I understand where your concerns regarding ever increasing chip costs is coming from, but you're forgetting that we're only doing a node shrink, not a node shrink + more powerful chip. This is my biggest reason why they can't keep producing 4N forever, so they will have to do a die shrink to keep up.
 
I don't understand this. Why is a node shrink not possible? TSMC 20nm was introduced in Nov 2014, Switch released with TX1 in March 2017 - that's a 2.5 year gap. TSMC 16nm was introduced in Q3 2015, Switch Lite released with TX1 "Mariko" in Q3 2019, 4 year gap.

TSMC 4N began production sometime in 2022, and Switch 2 is supposedly going to use it in 2025 (was supposed to come out late this year) so a 2.5-3 year gap - close enough. TSMC 3nm, specifically 3E, began production last year. If a Switch 2 Lite/Mini/electric boogaloo comes out 2 years after the Switch 2 with a TSMC 3E chip, that's.....you guessed it, 4 years. And it's a genuine node shrink too, considering the 50% greater transistor density over 4N. I expect 3E production to be hella cheap in 2027 (remember, it's 4 year old tech by this point); we'll have left the nanometer scale and entered the age of the Angstrom.

Chip design/manufacturing/tape out all in all takes time. These decisions are made years in advance. I wouldn't be surprised if work on the node shrink of the T239 has already begun by now.

Unless my math isn't math-ing, care to explain what I'm missing here?
There's no SRAM scaling with TSMC's N3E process node, with TSMC's N3E process node and TSMC's N5 process node family (including TSMC's 4N process node) sharing the same SRAM area of 0.021 µm².

And considering that most SoCs nowadays are designed with 70% SRAM and 30% logic in mind, how much the hardware components can shrink, and effectively the amount of cost savings from transitioning from TSMC's 4N process node to TSMC's N3E process node, is rather limited.
 
Speaking of moving "All Software" to the leftmost slot, I recently discovered I can hit d-pad left on the first icon and it wraps back around to "All Software". This console has been out since 2017 and I just learned that.
Yeah, it's a great shortcut, but I'd prefer it were more visually clear - and worked with touch.
 
0
Does limiting memory frequency make sense for power consumption?The experience of overclocking a PC makes me doubtful about this.
If the frequency difference is not significant, it's possible that the overall power consumption falls within the margin of error.
How much power saving can be achieved by reducing the frequency from 7500 to 5500?
Probably a shitload. Power curves are, well, curves, and 7500 is the absolute top of that curve - by definition, Samsung is the primary driver of the standard, the maximum clock speed that Samsung thinks they can safely manufacture gets set as the max in the standard.

Nintendo almost definitely will lower the memory frequency for the purposes of battery life. They ought to, 120 GB/s is absurd for a handheld. It's just wasteful, lowering it is basically free battery life.

That doesn't mean they'll knock it down in docked, I wanna be clear.
 
My point was, if 7nm is so prohibitively expensive after all this time, how on earth is 4N, a far more expensive node, a viable option?

If 4N can be viable today despite 7nm being "expensive," likewise, 3E will eventually become viable. All it'll take is passage of time. We're probably not going to see any nanometer-scale chips from any of the major companies by 2027/28; 4N will be ancient by those standards.
 

* Hidden text: cannot be quoted. *
EDIT :
I probably shouldn't quote people on hidden posts.
Sorry.
 
Last edited:
There's no SRAM scaling with TSMC's N3E process node, with TSMC's N3E process node and TSMC's N5 process node family (including TSMC's 4N process node) sharing the same SRAM area of 0.021 µm².

And considering that most SoCs nowadays are designed with 70% SRAM and 30% logic in mind, how much the hardware components can shrink, and effectively the amount of cost savings from transitioning from TSMC's 4N process node to TSMC's N3E process node, is rather limited.
I see. That makes sense. In that case I'm out of options; chiplet is probably not happening but I also don't think Nintendo will be able to keep costs down once 4N reaches antiquity status. This one's beyond me.
 
So the magnet subject could be more about scroll wheels, and not necessarily for attaching the Joycons on the tablet?
Nah I didn't mean that, I just wanted to point out that something eben smaller than the Switch exists that boasts an electro-permanent magnet. Some folks said that these magnets would be too big or something along those lines and I just don't think that that's the truth. We can easily expect the to be present in the console body as well as the joycons
 
There's no SRAM scaling with TSMC's N3E process node, with TSMC's N3E process node and TSMC's N5 process node family (including TSMC's 4N process node) sharing the same SRAM area of 0.021 µm².

And considering that most SoCs nowadays are designed with 70% SRAM and 30% logic in mind, how much the hardware components can shrink, and effectively the amount of cost savings from transitioning from TSMC's 4N process node to TSMC's N3E process node, is rather limited.
Wait, 30% logic and 70% SRAM? Are you sure you don't have those numbers flipped? I think I saw the opposite here: https://fuse.wikichip.org/news/7343/iedm-2022-did-we-just-witness-the-death-of-sram/

Edit: I read the Anandtech article. That's news to me.
 
Last edited:
0
Imo the third party headliner should not be a last-gen / cross-gen game. Would love to see Elden Ring in a sizzle reel but give me a significant current-gen only game to shout it loud and clear 'this is a current gen portable console'.
This was my thought as well. The biggest untapped part of the market for Nintendo right now is probably 3rd party AAA sector. They'll want the general public to think Switch 2 is a 'portable PS5' and not 'last gen port machine'.
Monster Hunter Wilds could be one of those games.
 
0
My projected launch window for Switch 2
-Metroid Prime 4 (along with Switch version)
-Visions of Mana
- FC 2025
-Persona 3: Reloaded (with DLC)
-New Sonic game
-Mario Kart 10
Mine:
  • 3D Mario
  • New EPD4 game involving the microphone
  • Breath of the Wild Remastered
  • Final Fantasy 16
  • Assassins Creed Shadows
  • Fortnite
  • Indie games
 
Even if it's on 4nm, it isn't going to be close to 3GHz. Hell, even on 4nm, it will probably be 2GHz max and even then I doubt it.

They aren't going to make a system that uses significantly more power hungry than the current Switch (I do expect a bigger battery this time around to offset increased power needs) and the current Switch is what, 8W or so in portable mode?

Some of your projections are way out to lunch.
Original Switch was like 10W, with 7W for SoC
 
How long do you guys think it will take for someone to hack/hard mod the Switch 2 this time around? I'm always surprised when I see how much potential the homebrew scene is able to get out of the Switch and I assume Switch 2 would be an incredible device to mod/hack.
We're still waiting for Mariko to be hacked without a hardware mod

A valid argument. But I do stand by my point because:

When we say a node is "expensive," what we mean is the cost of a single wafer is high. Now, if I were to be able to put. let's say, 50% more transistors within the same area, that means I'm theoretically saving 33% die space. Now, actual number might be a little below that but just bear with me for a bit. More transistors within the same area means smaller chip, right? So while the cost of a single, 4 year old wafer may still be "high" by today's standards, the cost per chip isn't going up proportionally because we're making more chips per wafer. Makes sense?

You also need to consider than by 2030 4N may not even be commercially viable, as it will be 8 year old technology and basically antiquated. The cost of keeping up 4N production will eventually exceed that of a shrunk down 3E chip. I understand where your concerns regarding ever increasing chip costs is coming from, but you're forgetting that we're only doing a node shrink, not a node shrink + more powerful chip. This is my biggest reason why they can't keep producing 4N forever, so they will have to do a die shrink to keep up.
When we say a node is expensive, we mean R&D costs. You don't just take a design and put it in a new machine. You have to redesign the chip. You're effectively making a while new chip, which is increasingly expensive
 
For those interested in CPU frequency/power draw, I found these numbers from a Geekerwan review a little while back

The SoC is a Dimensity 8100 with a cluster of 4x A78 cores and 4x A55 cores. It's fabbed on TSMC N5.

A78 Power Draw (per A78 core)
@ 2.1GHz = 0.49W
@ 1.9GHz = 0.4W
@ 1.3GHz = 0.19W

I multiplied the per core power draw by 8 for T239's CPU config.
@ 2.1 GHz = 3.92W
@ 1.9 GHz = 3.20W
@ 1.3 GHz = 1.52W
Then I made a graph to extrapolate power draw of the 8x A78C cluster at other frequencies
v8IM4QH.png


This is a rough estimate of course. Power draw shown here is a bit higher than what T239's CPU cluster would actually draw on 4N, since 4N would achieve slightly higher clocks at iso-power or conversely, better efficiency at iso-frequency. It was also brought to my attention that the single cluster of 8x A78C will not incur as significant of an efficiency penalty as the dual CPU cluster Dimensity 8100 due to the interconnect between the clusters consuming additional power.

Regardless, I would say that based on the numbers we have, T239's A78C cores are likely to be clocked around 2.0 GHz max, if not slightly below that
 
Last edited:
We're still waiting for Mariko to be hacked without a hardware mod


When we say a node is expensive, we mean R&D costs. You don't just take a design and put it in a new machine. You have to redesign the chip. You're effectively making a while new chip, which is increasingly expensive
I had taken that point for granted. TX1 was based on the 22nm family, on TSMC's 20nm. That whole design was moved forward to the 14nm family to TSMC 16nm.

I now understand that was a different time. Modern nodes have trouble shrinking the SRAM, so the viability of a node shrink for the T239 is questionable. I honestly don't know where to go from here.
 
For those interested in CPU frequency/power draw, I found these numbers from a Geekerwan review a little while back

A78 Power Draw:
@ 2.1GHz = 0.49W/c; 8x = 3.92 W
@ 1.9GHz = 0.4W/c; 8x = 3.20 W
@ 1.3GHz = 0.19W/c; 8x = 1.52 W

I don't remember which SoC it was specifically but I do remember it was manufactured on a TSMC 5nm-family node.

Anyways I made a graph to extrapolate power draw at other frequencies
v8IM4QH.png
Did the SoC have an 8-core A78 design? And if so, was it in 2 clusters of 4 or a single cluster of 8? I'm curious if 1x8 has any power savings over 2x4
 
Did the SoC have an 8-core A78 design? And if so, was it in 2 clusters of 4 or a single cluster of 8? I'm curious if 1x8 has any power savings over 2x4
None of them had 8 A78s. They were 1 cluster of 4Core A78 and 4Core A55. A single cluster would have power savings over two clusters of 4Core A78s though as the interconnects have costs and latency penalties
 
Why do you think 4N will become an antique, at least in the Switch 2's lifespan? TSMC still market larger nodes https://www.tsmc.com/english/dedicatedFoundry/technology/logic

TSMC 20nm became antiquated within the Switch's lifespan.

Though you do have a point, they still have some extremely old nodes still in active production because some critical and/or legacy applications still need them, but otherwise mass-market nodes like 20nm got axed. And after what I learned about the difficulty of shrinking SRAM, maybe 4N will stick around a little longer after all.
 
One thing about power consumption is that not every game hammers the system the same way even if the clocks are the same.

From my experience monitoring games on my V1, although MK8DX uses the same 460 MHz profile than Mortal Kombat 11, the former draws an avg of 5.8W (fan at 25%) while the latter stays at 6.8W (fan at 30%) [max brightness, max volume, wi-fi & BT off, joy-cons fully charged]

So, from my experience, the Switch isn't a 9W machine. I never saw this power consumption actually. MK11 is one of the most taxing games on Switch and I still get 2h21min.

So, with a 6.8W of power consumption, I imagine that the SoC is using ~4W. Now, if Drake has 4.2W (660 MHz) for the GPU and 2W for the CPU (~1.7GHz going by the Jetson Tool and cutting it in half for 4N), and 4W for the rest of the system (specially with the ~1.8" bigger screen compared with the V1), then we would have 10.2W. With have an increase of 50% in power consumption. We would need at least a 6465 mAh battery to have the same battery life (and this is all supposing that the 4.2W figure for the GPU is the max power draw at full usage)

I can't see Nintendo offering a worse battery life than the V1, and I also don't think they were happy with the more demanding games having ~2h25min of battery life. IF they could fit a 7500 mAh (3.7v) in there, that would help a bit.

Anyway, just napkin math to kill the time lol (not much happening today at the office)


-----------------------------------------------------------------------------------------------------------------------------------------------------


Talking about killing time (and napkin math), I just got my hands on a notebook from a coworker. When he told me it had a 4050 I had to ask to test it. And I did, with two games that shall not be named.

I set the TGP to 45W. Both games were using 98%~100% of the GPU. The average clock was at 2GHz. This GPU is 3 times bigger than Drake. I don't know if we can just divide by 3 to have an approximate idea of power consumption in a 12 SM GPU. I imagine the GDDR6 was drawing ~5W, so if we divide 40 by... 2.5 maybe? That's 16W for a constant 2GHz clock. I just thought it was interesting to share...


Also, that frame generation is another black magic from Nvidia. I couldn't tell the difference.
The current batteries found in Steam Deck and Rog Ally are not 3.7v, but 7.7v. So, while switch has a 16w-h battery, the current portables use a 40w-h battery.

If OZ at full power consume only 10w, this means a 4 hours in a battery like theses. But the problem is the weight. While Switch battery has only 75g, these from deck has a weight of 160g.

Maybe Nintendo can use a 4000 mAh and 7.7v for a 3 hours on 10w if weight became a problem.
 
0
Why do you think 4N will become an antique, at least in the Switch 2's lifespan? TSMC still market larger nodes https://www.tsmc.com/english/dedicatedFoundry/technology/logic

I just realized the 10nm technology (which SEC8N is optimized version of.. it's not actually 8nm) started in 2016/2017, the same year orig Switch came out 😯

It cannot be SEC8N, please.

5nm technology (which TSMC 4N is optimized version of) started in 2020.

T239 completed validation testing at nvidia India in 2022.
 
Yo … what happened to racoon T_T

(Tinfoil hat narcissism) - Maybe it’s because he is directly responsible for my username.
Nothing crazy. Just dipped outta the thread once it became abundantly clear that the successor will be noticeably larger than the current Switch. Not what he's looking for, which is more than fair. I'm not gonna lie and say that the size isn't a concern for me, but it's still also gonna be smaller than every other home console and I play docked 99% of the time.
 
I'm excited to see what the next Mario Kart will be like. They've had ages to brainstorm about it. I would think it must be at the least an early launch title.
 

TSMC 20nm became antiquated within the Switch's lifespan.

Though you do have a point, they still have some extremely old nodes still in active production because some critical and/or legacy applications still need them, but otherwise mass-market nodes like 20nm got axed. And after what I learned about the difficulty of shrinking SRAM, maybe 4N will stick around a little longer after all.
Interesting article, if technically "old news", particularly regarding the longevity of some 28nm and larger nodes. 4N is an Nvidia specific process, so how long it lasts will be up to big green. I'm hoping with Blackwell ramping up on 3nm there'll be space for T239 on 4N from Q3 this year. But as noted above, these production runs are planned years in advance
 
For those interested in CPU frequency/power draw, I found these numbers from a Geekerwan review a little while back
Did the SoC have an 8-core A78 design? And if so, was it in 2 clusters of 4 or a single cluster of 8? I'm curious if 1x8 has any power savings over 2x4
Only the C variant of A78 supports an 8 core cluster and the only product I know of that uses it in that config is Nvidia Bluefield which is not a device anyone is doing benchmarking on.

Nvidia publishes Orin power curves, if you want to see what 8nm would look like. You can use their power tool to look at various configurations.
 
I can see something like:

  • MP4 Fall 2024
  • MK10 and BOTW Remaster+Zelda's Fury Spring 2025
  • New Mario 3D Holiday 2025
  • New DK Game, Zelda OOT Remake (40th Anniversary) Spring 2026
  • Luigi Mansion 4 and New Smash Bros Fall 2026
  • New AC 2027
  • TOTK Remaster or Zelda 2D 2027
  • New Zelda 3D 2028
I thought about Monster Hunter Wilds, however, I feel a late port will come in early 2026 or later
 
Interesting article, if technically "old news", particularly regarding the longevity of some 28nm and larger nodes. 4N is an Nvidia specific process, so how long it lasts will be up to big green. I'm hoping with Blackwell ramping up on 3nm there'll be space for T239 on 4N from Q3 this year. But as noted above, these production runs planned years in advance
We won't see any 3 nm Nvidia products come to market until 2026.
 
EDIT :
I probably shouldn't quote people on hidden posts.
Sorry.
I don't think there is anything wrong with that, as long as the hidden content stays hidden.

I can't add anything to the shipping manifests, I think @darthdiablo and @LiC have taken all the meat off of that bone, but I've seen the LinkedIn resume before, it's been discussed here. I think I found it? Maybe I didn't, I don't want to take credit away from whomever did, it's been a while.
 
5nm technology (which TSMC 4N is optimized version of) started in 2020.

T239 completed validation testing at nvidia India in 2022.
Just for comparison the last generation(s) of the iPad and Google Pixel lines used 5nm process and LPDDR5 RAM, even the Pixel 6a which launched in 2022 for $450 used 5nm. The current ones use 3nm / 4nm and just switched to LPDDR5X memory. The Pixel 8 still uses UFS 3.1 storage.

The Switch 2 will most likely have 5X memory according to shipping data, just timeline wise in comparison with mobile products, TSMC 4N (5nm) makes sense. Modern RAM with higher bandwidth, reasonably modern node.
 
Potential timeline when referring to the current latest facts aligning from suppliers and manufacturing and not "plans changed" or rumors from "insiders"

  • Q2 (Mar-June 2024): Pre-Prod QC + First mention of the console (WE ARE HERE)
  • Q3 (July-Sep 2024): Mass Production + Reveal/Presentation
  • Q4 (Oct-Dec 2024): Launch
I'm putting money on a November launch
Don't you need more time between mass production and launch? I forgot how long it takes. Maybe 6-8 months?
 
I thought about Monster Hunter Wilds, however, I feel a late port will come in early 2026 or later
How come? MHWorld has sold 25 million copies total, while the Switch version of Rise has sold at least 1/3rd of that just on Switch. Switch sold 7.5 million on Switch in 2 months, and then I can't find anymore information on just the Switch release. Personally I'd guess it's 10+ million on Switch after all this time if it sold 7.5 million before Sunbreak came out, as it's topped 15 million recently. Anyways, my point is that Nintendo consoles, especially because they're the only portable of the big 3 left, is where Monster Hunter sells best. They want those sales day 1, and Capcom is super close partner with Nintendo, there's no doubt in my mind they got a dev kit day 1 to start porting Wilds over.
 
I think the most telling thing re: launch price will be if there is ever any permanent discount offered on V2/OLED Switch models.
I'm guessing
V2 250 USD
Oled 300 USD
Switch Lite 150 USD

Include a game bundle and you'll be easily get the 13M units.
 
0
good lord, seeing this headline gave me a jumpscare lmfao
eeee.png

here's the link for anyone interested
what kinda event is this?? There's only 2 Nintendo games coming out this year (announced thus far at least, and there's not Direct happening before the 18th), so is it just gonna be demos for Paper Mario and Luigi's Mansion? I don't see how this warrants any special event, 1 is a remake and the other is a simple remaster. Maybe I'm alone here, but that seems like too little to host an event over.

editing to add: I'm not speculating that something BIG is gonna happen in the next 4 days, I'm just confused at why Gamestop would choose now of all times to host such an event lol. I expect nothing to happen between now and this event.
 
Last edited:
Potential timeline when referring to the current latest facts aligning from suppliers and manufacturing and not "plans changed" or rumors from "insiders"

  • Q2 (Mar-June 2024): Pre-Prod QC + First mention of the console (WE ARE HERE)
  • Q3 (July-Sep 2024): Mass Production + Reveal/Presentation
  • Q4 (Oct-Dec 2024): Launch
I'm putting money on a November launch

Am I the only one still hoping for a Spring launch 🥲
 
what kinda event is this?? There's only 2 Nintendo games coming out this year (announced thus far at least, and there's not Direct happening before the 18th), so is it just gonna be demos for Paper Mario and Luigi's Mansion? I don't see how this warrants any special event, 1 is a remake and the other is a simple remaster. Maybe I'm alone here, but that seems like too little to host an event over.

Yeah this is weird, does GameStop do those kind of events on a regular basis for regularly scheduled Switch games before?

Cuz Paper Mario and Luigi Mansion I don’t think warrants a special event either, I’m with you
 
Please read this new, consolidated staff post before posting.

Furthermore, according to this follow-up post, all off-topic chat will be moderated.
Last edited by a moderator:


Back
Top Bottom