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

[Alovon11]

"It's old"

Yeah, it's very weird to me, like.... We already are going to see a 3x+ boost over og Switch most likely and that's on 8nm.

Why is that bad all of a sudden?

 

[ILikeFeet]

Yeah, it's very weird to me, like.... We already are going to see a 3x+ boost over og Switch most likely and that's on 8nm.

Why is that bad all of a sudden?

because everyone else are on better nodes.

you know, the same old talking points

 

[Tentacle-tropes]

Think of some good console warring posts and that is probably your answer

 

[Kevin]

Anything new in here?

More or less no reliable information since before the Switch OLED release. Just "insider" claims, nvidia patents and chip details, shady reddit posts from someone who has an uncle at Nintendo, some very broad Nintendo job postings.

Having followed all of this for years my assessment is that there will likely be no new Nintendo hardware for a very long time. Probably several years. The evidence is sketchy at best.

Switch continues to dominate sales even while commanding the same asking price over five years. Nintendo is in no hurry to move on from this gravy train. No new hardware until at least 2024 is my prediction.

 

[Skittzo]

More or less no reliable information since before the Switch OLED release. Just "insider" claims, nvidia patents and chip details, shady reddit posts from someone who has an uncle at Nintendo, some very broad Nintendo job postings.

Having followed all of this for years my assessment is that there will likely be no new Nintendo hardware for a very long time. Probably several years. The evidence is sketchy at best.

Switch continues to dominate sales even while commanding the same asking price over five years. Nintendo is in no hurry to move on from this gravy train. No new hardware until at least 2024 is my prediction.

If you've followed this for years your conclusion should be the opposite. They put out new hardware every 2 years (or less) almost like clockwork.

 

[Dekuman]

I’m out of step here… why are we considering that the Samsung 8nm process node is out?

i think it lines up, but nvidia themselves are still on consumer ampere cards which is using 8nm. it makes sense for Dane to pick up the capacity once they move off of it. That or they go onto an intermediate process, if one exists. Maxwell was on 28nm, Pascal was on 16 and X1 was on 20.

The age doesn't matter right now i think, more about how nvidia positions their fab transitions since they will be supplying Nintendo the chips. They are probably more in the driver seat of this than Nintendo is

 

[Terrell]

OK, so we’re thinking new node for no raisin.

8nm is fine, it gives room for a new node later down the line, hopefully by the time the chip shortage gets a little less intense. It’s a substantial enough jump from where we are now, and with Samsung’s new node technologies, I’d like for Nvidia to stick with them and get capacity up, since I think they’re going to end up with a winning technology for new process nodes over TSMC, at least in the short term.

 

[ShadowFox08]

It was very close to finalized, and maybe it is now. There’s no way to know.

If they’re changing process nodes this late in the process, they have to scrap/ redo a lot of work.

Isn't the 8nm Samsung Nvidia just as performant efficient as 7nm AMD GPUs (citation needed?)

As long as they go balls to the walls on power to match AMD Steam Deck GPU and get minimum 2.5-3 hour battery life for switch, then It will be alright by me i guess. The revision 2 years later could have the battery life increase similar to switch on a 4-5nm mode I guess.

 

[ReddDreadtheLead]

Isn't the 8nm Samsung Nvidia just as performant efficient as 7nm AMD GPUs (citation needed?)

As long as they go balls to the walls on power to match AMD Steam Deck GPU and get minimum 2.5-3 hour battery life for switch, then It will be alright by me i guess. The revision 2 years later could have the battery life increase similar to switch on a 4-5nm mode I guess.

you aren't getting a revision with a better battery life 2 years later if it is on Samsung 8nm... (with certain caveats)

The only way you will get that is if they expend a massive R&D for a chip to be redesigned for a process that uses a different lithography than the 8nm. Do you really expect them to do such a massive R&D for simply a revision especially in these times? No company would waste that much for that and expect it to make good business sense. They'd get questioned by investors on such an exuberant spending for that. Even if they split the bill with Nvidia, it won't really be cheap.

And that would be on top of already spending a lot on R&D for this "Dane" chip.

But to the earlier subject, the only possibility I could see them go with this newer process is if the special node that is exclusive only to Nvidia and Nvidia clients was made with this "dead-end" in mind and they already have it prepared for a denser process. Now, Samsung did have a 10nm move to a 7nm (Exynos9820->9825), but it was the only chip to ever use it and Samsung seemingly brushed the 7LPE under the rug and wants no one to think about it. They only talk about 7LPP or later.

However, there is a chance that Nvidia already has an ORIN variant on a different node, but then, for who?


Mind you, it isn't like they don't really have engineers on both Samsung and TSMC, where they work with the GPUs and the CPUs. They do have them, and they are already working on future products with both SEC and TSMC.

 

[Dakhil]

Isn't the 8nm Samsung Nvidia just as performant efficient as 7nm AMD GPUs (citation needed?)

Outside of idle power consumption, which is ~28.57% higher (9 W vs 7 W), the GeForce RTX 3070 is actually quite a bit more power efficient than the RX 6700 XT according to TechPowerUp's power consumption benchmarks. The reason why I mentioned the GeForce RTX 3070 is because GDDR6X can significantly increase power consumption, as shown with TechPowerUp's power consumption benchmarks for the GeForce RTX 3070 Ti.

 

[#Salt]

What about 6 LPP? We haven't heard of other products using that. What I mean to say is that it might not be as constrained as 5nm, or even 8nm.

 

[ShadowFox08]

you aren't getting a revision with a better battery life 2 years later if it is on Samsung 8nm... (with certain caveats)

The only way you will get that is if they expend a massive R&D for a chip to be redesigned for a process that uses a different lithography than the 8nm. Do you really expect them to do such a massive R&D for simply a revision especially in these times? No company would waste that much for that and expect it to make good business sense. They'd get questioned by investors on such an exuberant spending for that. Even if they split the bill with Nvidia, it won't really be cheap.

And that would be on top of already spending a lot on R&D for this "Dane" chip.

But to the earlier subject, the only possibility I could see them go with this newer process is if the special node that is exclusive only to Nvidia and Nvidia clients was made with this "dead-end" in mind and they already have it prepared for a denser process. Now, Samsung did have a 10nm move to a 7nm (Exynos9820->9825), but it was the only chip to ever use it and Samsung seemingly brushed the 7LPE under the rug and wants no one to think about it. They only talk about 7LPP or later.

However, there is a chance that Nvidia already has an ORIN variant on a different node, but then, for who?


Mind you, it isn't like they don't really have engineers on both Samsung and TSMC, where they work with the GPUs and the CPUs. They do have them, and they are already working on future products with both SEC and TSMC.

Ah, I remember this dead end conversation now. So that makes it less likely to be an 8nm then. Because switch 2 will get a revision.. Only one day to find out then

 

[Terrell]

you aren't getting a revision with a better battery life 2 years later if it is on Samsung 8nm... (with certain caveats)

The only way you will get that is if they expend a massive R&D for a chip to be redesigned for a process that uses a different lithography than the 8nm. Do you really expect them to do such a massive R&D for simply a revision especially in these times? No company would waste that much for that and expect it to make good business sense. They'd get questioned by investors on such an exuberant spending for that. Even if they split the bill with Nvidia, it won't really be cheap.

And that would be on top of already spending a lot on R&D for this "Dane" chip.

But to the earlier subject, the only possibility I could see them go with this newer process is if the special node that is exclusive only to Nvidia and Nvidia clients was made with this "dead-end" in mind and they already have it prepared for a denser process. Now, Samsung did have a 10nm move to a 7nm (Exynos9820->9825), but it was the only chip to ever use it and Samsung seemingly brushed the 7LPE under the rug and wants no one to think about it. They only talk about 7LPP or later.

However, there is a chance that Nvidia already has an ORIN variant on a different node, but then, for who?


Mind you, it isn't like they don't really have engineers on both Samsung and TSMC, where they work with the GPUs and the CPUs. They do have them, and they are already working on future products with both SEC and TSMC.

They took the Erista version of the Tegra X1 (TSMC 20nm) and made Mariko (TSMC 16nm, first process node from TSMC to use FinFET). Would updating a non-FinFET chip to a FinFET chip on a newer process node not have required a similar level of re-designing to achieve?

 

[Bosintang]

More or less no reliable information since before the Switch OLED release. Just "insider" claims, nvidia patents and chip details, shady reddit posts from someone who has an uncle at Nintendo, some very broad Nintendo job postings.

Having followed all of this for years my assessment is that there will likely be no new Nintendo hardware for a very long time. Probably several years. The evidence is sketchy at best.

Switch continues to dominate sales even while commanding the same asking price over five years. Nintendo is in no hurry to move on from this gravy train. No new hardware until at least 2024 is my prediction.

I mostly agree. While a serious, incremental revision may have been something seriously considered by Nintendo at some point, that ship sailed the moment the OLED model was announced. There's too many roadblocks with the chip shortages and too much success from the existing, simple to produce models to release a super Switch this year or early next year. End of 2023 or March 2024 for a full successor appears to be the safest bet. And I would not be surprised if it was end of 2024.
I'm quite bummed because I was really looking forward to a more powerful system around 2020-21. I'll probably cave in and buy a Series S if there are good offers eventually.

 

[Hermii]

They took the Erista version of the Tegra X1 (TSMC 20nm) and made Mariko (TSMC 16nm, first process node from TSMC to use FinFET). Would updating a non-FinFET chip to a FinFET chip on a newer process node not have required a similar level of re-designing to achieve?

Yea, I think this "you cant shrink 8nm" argument is wildly exhaggarated.

 

[NineTailSage]

Yea, I think this "you cant shrink 8nm" argument is wildly exhaggarated.

Yeah I think most people are currently just looking at how expensive it is today to fabricate on these much newer manufacturing nodes.
The question is how long could they push an 8nm design and still stay relevant, but we see with the Switch we have now have far Nintendo will push a product.

 

[Simba1]

Talking about chip shortages, Sony just cut PS5 forecast for current FY from 16m to 11.5m PS5 units.

 

[Dakhil]

What about 6 LPP? We haven't heard of other products using that. What I mean to say is that it might not be as constrained as 5nm, or even 8nm.

Samsung's 6LPP process node's practically only available to Samsung, especially with Samsung's 6LPP process node no longer being in Samsung's technology roadmap. Although Samsung's 6LPP process node could also be available to some of Samsung's largest consumers, theoretically speaking, as mentioned previously, there's no chip from Samsung or any other company confirmed to be fabricated using Samsung's 6LPP process node, which makes me believe Samsung's 6LPP process node practically doesn't really exist.

As for Samsung's 7LPP process node, I went into detail why I think Samsung's 7LPP process node isn't a serious consideration here.

 

[Intoxicate]

Talking about chip shortages, Sony just cut PS5 forecast for current FY from 16m to 11.5m PS5 units.

Sony is doomed

 

[bodokoh]

So, is Dane the next Nintendo Switch revision or successor? I don't believe Nintendo would waste a lot of efforts in creating 4K Switch revision when they can use that for the Switch successor.

 

[Skittzo]

So, is Dane the next Nintendo Switch revision or successor? I don't believe Nintendo would waste a lot of efforts in creating 4K Switch revision when they can use that for the Switch successor.

We don't know and in practical terms there's no real difference between the two. It's all marketing.

 

[ILikeFeet]

So, is Dane the next Nintendo Switch revision or successor? I don't believe Nintendo would waste a lot of efforts in creating 4K Switch revision when they can use that for the Switch successor.

"revision" and "successor" are pretty irrelevant now. the Xbox Series S is about as powerful as the Xbox One X and it will, eventually, get games that the One X can't run

 

[Dakhil]

Hopefully MVG hears more details that he can share during GDC 2022 at late March 2022.


Outside of whether the launch window has changed or remained the same, and maybe estimated performance metrics, I don't expect much information.

 

[Dekuman]

Hopefully MVG hears more details that he can share during GDC 2022 at late March 2022.


Outside of whether the launch window has changed or remained the same, and maybe estimated performance metrics, I don't expect much information.

Doesn't he work at Nightdive? i think he's said before he could be fired if he revealed anything he know about a potential Switch Pro, assuming he knew anything.

If nightdive were to also be developing on the new hardware the appearances of an employee hinting at stuff or leaking could be problematic even if the information was obtained from a different source. I assume any leaks from GDC would be laundered through journalists.

 

[Dakhil]

They took the Erista version of the Tegra X1 (TSMC 20nm) and made Mariko (TSMC 16nm, first process node from TSMC to use FinFET). Would updating a non-FinFET chip to a FinFET chip on a newer process node not have required a similar level of re-designing to achieve?

One difference is that Samsung's 7LPP process node and onwards and TSMC's N6 process node and onwards use EUV lithography instead of DUV lithography. (TSMC's N7+ process node also uses EUV lithography. However, only Huawei's Kirin 990 was publicly announced to be fabricated using TSMC's N7+ process node, which makes sense, considering TSMC's N7+ process node is not IP compatible with TSMC's N7 process node. And TSMC mentioned expecting customers to migrate from TSMC's N7 process node to TSMC's N6 process node.) On the other hand, both TSMC's 20 nm** process node and TSMC's 16 nm** process node use DUV lithography (DUV lithography uses argon fluoride (ArF) lasers.)

And the differences between DUV lithography and EUV lithography aren't trivial, considering that EUV lithography does require at least considerably more logic process steps in comparison to DUV lithography, especially when EUV lithography was starting to become adopted. However, as usage of EUV lithography increases, EUV lithography can also reduce the amount of photomasks, which can allow more 2D designs to be printed, which is definitely important, since DUV lithography does require significantly more logic process steps due to DUV lithography requiring more photomasks than EUV lithography at 7 nm** and more advanced.

Doesn't he work at Nightdive? i think he's said before he could be fired if he revealed anything he know about a potential Switch Pro, assuming he knew anything.

If nightdive were to also be developing on the new hardware the appearances of an employee hinting at stuff or leaking could be problematic even if the information was obtained from a different source. I assume any leaks from GDC would be laundered through journalists.

That's why I mentioned "...that he can share..." But yeah, realistically speaking, there's probably not much MVG can share. And I personally expect developers to divulge any relevant information from GDC 2022 indirectly via news outlets (e.g. Eurogamer, etc.).

 

[Look over there]

They took the Erista version of the Tegra X1 (TSMC 20nm) and made Mariko (TSMC 16nm, first process node from TSMC to use FinFET). Would updating a non-FinFET chip to a FinFET chip on a newer process node not have required a similar level of re-designing to achieve?

Good question, and googling a bit led to me learning something new today:
First version of TSMC 16nm was basically 16 nm FEOL + 20 nm BEOL. So there wasn't that full node's worth of changes to design rules.

It also seems like the rate of change for design rules has been rising, which presumably in turn at least partially explains the accelerating design costs.

 

[Thraktor]

I kind of doubt a company as particular and controlling as Nintendo would do this.

Also bringing up Game Boy Color brings with it a particular set of circumstances (shelving of the 32-bit Atlantis included) that I don't think quite fits to Switch. For one Game Boy tech was fairly old by the time Color came out and GBC was effectively the same Game Boy tech with more VRAM, doubled clocks and doubled WRAM and even then Nintendo didn't miss the fact that it was color in the branding. While they counted its sales towards GB's total, i would argue it was positioned an upgrade even if successor may not fit.

A similar positioning for the next Switch would have made more sense if it was based on the X1 design from 2015 with everything more or less doubled with some kind of 4K upscaler for TVs. Given what we know of the Dane SoC it seems like a true successor, unless that tech gets shelved like the Atlantis and we go back to the worst timeline of a die shrunk X1 (lol).

I'm not saying they will explicitly engage in generaitonal talk like sony has, but if they do a low key upgrade , it will be more like Apple's strategy of clear cut generations within its product lines but one that is largely detatched from what others are doing. So a Gen 2 Switch will be clearly marketed as such. Not sure if they will include or exclude its sales as part of the Switch Lifetime but that is i think a less important conversatioon.

To clarify what I mean here, and in my long and rambling post that followed (which is a good reminder to not start writing "quick" replies at 1am), I don't think Nintendo should take a mid-ground approach between a "pro" console and a full new generation, but I'm concerned that they will. I think Nintendo should just release a Switch 2, and literally just call it Switch 2. Have a gradual cross-gen period like we're seeing with the new Xbox and PS consoles, but release a device that's very clearly a new generation and very much a straight-forward update to the Switch.

Looking at the evidence we have, though, (which admittedly is sparse), and it doesn't really paint a picture of a company preparing a full on, straight-forward new generation console. If you go to the OP of this thread and look through the rumours, it seems pretty consistent from the reports from the like of NateDrake and Bloomberg that this is being positioned (at least to third party developers) as a 4K Switch, not a Switch 2. A more recent Twitter post from ZhugeEX suggested it's now more likely to be a successor, but that's only based on timing. I would also argue that a Samsung 8nm manufacturing process is more indicative of a "pro" model rather than a successor.

But it's too late for a "pro" model, particularly one which features big architectural jumps. So what I'm concerned about is that Nintendo is releasing a sort-of-successor, but isn't committing fully to just releasing a Switch 2, and in particular that they're not committed to continuing on with the Switch line as the future of their hardware endeavours. Nintendo haven't ever released a Console 2 before. Even aside from using numerical naming scheme, they haven't really released a new generation which is just a straight forward upgrade over the previous gen in their recent history. Every new generation has had some new "thing" that distinguishes it from the previous gen, and has had some new name to go along with that. The DS to 3DS is the closest they've got to a straight forward upgrade, but it's not really a DS2, and very much leaned on the 3D feature. You could also argue that NES>SNES>N64>Gamecube was a line of traditional upgrades (although none had BC, so it's not quite the same as the current situation), but that wasn't particularly successful for them, as sales dropped each generation.

So I'm concerned that Nintendo will do a bit of a reverse-DS. With the Nintendo DS they made a big point of calling it the "third pillar" when it was launched, in part to assuage owners of the GBA, which wasn't all that old, but also because it gave them an easy way out if the device failed (keeping in mind many were expecting Sony to dominate handhelds with the PSP). If the DS failed, they could just release a new handheld in the GameBoy line to compete more directly with the PSP, and sweep the DS under the rug, as it was only the "third pillar". I fear Nintendo could be trying to do the same thing in reverse, instead of releasing a very different device while giving themselves an "out" to go back to a more traditional approach, they'll release a more traditional device in a way that gives them an "out" to release something different.

While many people complain about Nintendo being conservative, in reality they're anything but conservative when it comes to hardware. Sony has a very conservative approach to hardware. They take what works, add more powerful internals, maybe update the controller, and increment the number on the end. Nintendo have never really taken this approach, or at least never so explicitly, and not finding so much success with it. Their most successful hardware of the past few years (and ever, for that matter), are the DS, Wii and Switch, all of which were very different to their predecessors. I'm sure there are many people in senior positions in Nintendo who still have this anti-conservative approach to new hardware, and don't want to release a simple generational update based on past experience. I think they've found themselves in an awkward compromise where they're going to release a new Switch-style device, but without fully committing to it as their new generation, and I think without committing to it they'll harm its chances to succeed.

IIRC the 3DS did not have a comfortable profit margin at $250, it was more or less sold at their typical minor profit. The price cut really harmed their financials for a few years. Whatever the reason they really did seem to overspend on the 3DS BoM.

But when it comes to whatever Dane is, I think the point I'm trying to make is a "pro" style upgrade and a generational upgrade don't have to be different concepts. Nintendo's primary underlying strategy with the Switch was to unify the software environment for all of their products, so you can play ~99% of the same library on everything they release. So far this has extended to the base Switch, the Lite and the OLED model.

There's no reason they can't then sell a 4k model that has the same software environment. It will be able to play games the others can't play but the hybrid is also capable of playing games the Lite can't play, so it fits with the current patterns.

It wouldn't be positioned as a pro, and it wouldn't be positioned as a Switch 2. It would be another member in the Switch family, one which would- in the long run- serve the purpose of a generational upgrade. Maybe 3-4 years after launch when no more base Switches are being made.

The reason they would want to do this is to have a much higher point of entry when it comes to revenue. I do believe they'd go for a comfortable profit margin for this concept, maybe as high as $450 or even $500.

The reason it works is because the current environment for consumer electronics is insatiable. People would buy a Switch that can play all your existing games in 4k plus new ones, especially early adopters who tend to be enthusiasts. Even at $500 I can't see it not being sold out in most major markets for months, at least if it is launched relatively soon (i.e. within the next ~15 months).

The long term goal then would be to slowly position it as a new base model that will be 2025's default Switch, for instance. This would involve lowering the price or releasing a slightly lower end model with the same Dane chip, but for example with less storage.

This is how they avert the increased risk that comes with having only one active platform- they don't deactivate it until the next platform has a userbase that can sustain itself. This is done by positioning not as the "next platform" but as just another member of the current platform family.


I also think this strategy does jive with the idea that it's using an 8nm process. It doesn't need to be cutting edge when it will still be enough of an upgrade to greatly improve existing Switch games which will be the primary driver early on.

According to two contemporary reports, the 3DS was estimated to cost about $100 in BoM at launch for its $250 retail price. Of course there's packaging, distribution, retailer margins, etc., but that still points to a comfortable profit. Particularly when the DSi BoM came to $75.58 at launch for a $170 retail price (same article). The DSi was a mid-gen refresh/pro model which I would definitely expect to be sold at a profit, but the 3DS had a higher margin, particularly when you consider that the packaging/distribution costs would have been about the same for both of them. Then the Switch was estimated at a $257 BoM for a $300 retail price. These are obviously all estimates, and could be wrong, but they're the most accurate info I can find.

The issue I see with the "current environment for consumer electronics is insatiable", is that when they were making the important design decisions on this, that wouldn't have been obvious at all. I'd guess that somewhere around late 2019 to early 2020 they would have had to come to an agreement with Nvidia on the high-level details of Dane (ie manufacturing process, CPU and CPU architectures, and numbers of cores, SMs, etc.) to give enough time for the chip to be designed, taped out and manufactured. At that point if they'd based their plans on "we can sell anything at $500 and people will just eat it up", they would have been laughed at.

I also don't think it's entirely true. Demand is high, but Switch is easily enough available and sales are starting to drop. Pro style consoles sold alongside the base model also seem to sell much less than the base model. If the current circumstances existed during the launch of the PS4 Pro and Xbox Series X, I can't see there being anywhere near the demand for the PS4 Pro there is now for the PS5. If positioned as a Switch 4K, I could definitely see there being a decent early-adopter demand, but no guarantee beyond that. I don't think playing the same games at higher resolution is that big of a draw for most people, particularly the market that is currently happy to pay $300 for the Switch.

Another possibility is Samsung's 5LPE process node or Samsung's 5LPP process node.
Although EUV lithography instead of DUV lithography is being used, one potential advantage of using Samsung's 5LPE process node or Samsung's 5LPP process node instead of using TSMC's N5 process node is that Nintendo and Nvidia don't need to completely redesign Dane from scratch since I imagine Nintendo and Nvidia would still be using Samsung's IP when designing Dane, assuming kopite7kimi's still correct about Dane being based on Samsung's 8N process node.
Another potential advantage is that Nvidia would have an easier time securing capacity for Samsung's 5LPE process node or Samsung's 5LPP process node in comparison to TSMC's N5 process node since I imagine Samsung's 5LPE process node and Samsung's 5LPP process node are not as high in demand as TSMC's N5 process node, although probably in much higher demand in comparison to Samsung's 8N process node, especially with the possibility that Nvidia decides to have mid-range and/or entry-level Lovelace GPUs fabricated using Samsung's 5LPE process node or Samsung's 5LPP process node.
Of course, one potentially serious caveat to using Samsung's 5LPE process node or Samsung's 5LPP process node is that the yields for Samsung's 5LPE process node or Samsung's 5LPP process node is probably not as good as the yields for TSMC's N5 process node, especially if there are grains of truth to Business Korea's report. And I'm inclined to believe there are large grains of truth to Business Korea's report, considering a Xiaomi executive mentioned that the reason the Snapdragon 780G was discontinued relatively quickly was that there was not enough capacity, which does seem to strongly validate Business Korea's report.

I'd have said Samsung's 7LPP process node's also a possibility. However, Andrei Frumusanu mentioned that the binning of the Exynos 990 units Anandtech received are apparently, with the binning of most Exynos 990 units out there being as bad, or worse. And of course, Samsung's 7LPP process node is 20-30% less power efficient in comparison to TSMC's N7 process node.
The only chips publicly revealed and rumoured to be fabricated using Samsung's 7LPP process node is the IBM Telum processor, IBM Power10 processor, and HW 4.0, respectively, which I imagine are relatively low volume chips as far as fabrication is concerned. And I could be wrong, but I imagine yields are not as much of a concern for relatively low volume chips (e.g. IBM Power10 processor, etc.) in comparison to relatively high volume chips (e.g. Exynos 990, etc.).
So I imagine Samsung's 7LPP process node having terrible yields is too much of a huge deal breaker to be a serious consideration, especially when considering that Dane probably has a high performant mobile GPU.

And although the yields for Samsung's 5LPE process node or Samsung's 5LPP process node is probably not as good as the yields for TSMC's N5 process node, I imagine the yields for Samsung's 5LPE process node or Samsung's 5LPP process node are better than the yields for Samsung's 7LPP process node, considering there are many products equipped with the Snapdragon 888 released.


I believe the Kirin 990 from Huawei is the only chip publicly revealed to be fabricated using TSMC's N7+ process node, which I don't think is surprising, considering TSMC's N7+ process node isn't IP compatible with TSMC's N7 process node. (And AMD mentioned Zen 3 isn't fabricated using TSMC's N7+ process node.)


Gabe Newell mentioned that selling the Steam Deck at the MSRP of $399.99 is painful. So Valve's definitely selling the Steam Deck at a loss. And nobody knows if Nintendo's willing to sell at a loss or not.

Yeah, Samsung's 7nm or 5nm processes would also be possibilities, although part of my thought process is what would have been a good choice if they were making the decision in late 2019/early 2020 as I mentioned above. TSMC's 7nm family would have been a known quantity by that point, with the A100 likely already taped out, and I think Nvidia and Nintendo could have been very confident about the performance, density and yields of future TSMC 7nm (or 6nm) processes at that stage. Samsung's 7LPP would have existed at the time, although Nvidia never designed anything for it, and to match (or at least get close to) TSMC's 7nm processes, they would have had to look at 5LPE or 5LPP, which would have been very much unknown quantities. So it's not so much that TSMC's processes would have better performance, density and yields (which they probably do), but TSMC's 7nm process would just be a much safer option, more of a known quantity.

the only way I can see Nvidia making a chip off 8nm is if TSMC suddenly has spare capacity that Nvidia buys up or Samsung gives out steep discounts for 4nm. the latter sounds more likely to me if this chip gets pushed out further to late 2023

There's no "suddenly" about it. The manufacturing process would have been chosen at the start of the SoC design process, which is years before hardware ships. Changing manufacturing process mid-design would be an extremely drastic move, resulting in significant costs and delays.

I’m out of step here… why are we considering that the Samsung 8nm process node is out?

I don't think Samsung 8nm is "out", it's been reported by Kopite7Kimi, who is a very reliable leaker on these things, and I have no reason to disbelieve him. The reason I brought it up is that I think it's an indication that they're not just going for a straight-forward Switch 2.

 

[Dekuman]

To clarify what I mean here, and in my long and rambling post that followed (which is a good reminder to not start writing "quick" replies at 1am), I don't think Nintendo should take a mid-ground approach between a "pro" console and a full new generation, but I'm concerned that they will. I think Nintendo should just release a Switch 2, and literally just call it Switch 2. Have a gradual cross-gen period like we're seeing with the new Xbox and PS consoles, but release a device that's very clearly a new generation and very much a straight-forward update to the Switch.

Looking at the evidence we have, though, (which admittedly is sparse), and it doesn't really paint a picture of a company preparing a full on, straight-forward new generation console. If you go to the OP of this thread and look through the rumours, it seems pretty consistent from the reports from the like of NateDrake and Bloomberg that this is being positioned (at least to third party developers) as a 4K Switch, not a Switch 2. A more recent Twitter post from ZhugeEX suggested it's now more likely to be a successor, but that's only based on timing. I would also argue that a Samsung 8nm manufacturing process is more indicative of a "pro" model rather than a successor.

But it's too late for a "pro" model, particularly one which features big architectural jumps. So what I'm concerned about is that Nintendo is releasing a sort-of-successor, but isn't committing fully to just releasing a Switch 2, and in particular that they're not committed to continuing on with the Switch line as the future of their hardware endeavours. Nintendo haven't ever released a Console 2 before. Even aside from using numerical naming scheme, they haven't really released a new generation which is just a straight forward upgrade over the previous gen in their recent history. Every new generation has had some new "thing" that distinguishes it from the previous gen, and has had some new name to go along with that. The DS to 3DS is the closest they've got to a straight forward upgrade, but it's not really a DS2, and very much leaned on the 3D feature. You could also argue that NES>SNES>N64>Gamecube was a line of traditional upgrades (although none had BC, so it's not quite the same as the current situation), but that wasn't particularly successful for them, as sales dropped each generation.

So I'm concerned that Nintendo will do a bit of a reverse-DS. With the Nintendo DS they made a big point of calling it the "third pillar" when it was launched, in part to assuage owners of the GBA, which wasn't all that old, but also because it gave them an easy way out if the device failed (keeping in mind many were expecting Sony to dominate handhelds with the PSP). If the DS failed, they could just release a new handheld in the GameBoy line to compete more directly with the PSP, and sweep the DS under the rug, as it was only the "third pillar". I fear Nintendo could be trying to do the same thing in reverse, instead of releasing a very different device while giving themselves an "out" to go back to a more traditional approach, they'll release a more traditional device in a way that gives them an "out" to release something different.

While many people complain about Nintendo being conservative, in reality they're anything but conservative when it comes to hardware. Sony has a very conservative approach to hardware. They take what works, add more powerful internals, maybe update the controller, and increment the number on the end. Nintendo have never really taken this approach, or at least never so explicitly, and not finding so much success with it. Their most successful hardware of the past few years (and ever, for that matter), are the DS, Wii and Switch, all of which were very different to their predecessors. I'm sure there are many people in senior positions in Nintendo who still have this anti-conservative approach to new hardware, and don't want to release a simple generational update based on past experience. I think they've found themselves in an awkward compromise where they're going to release a new Switch-style device, but without fully committing to it as their new generation, and I think without committing to it they'll harm its chances to succeed.

I think the biggest counterpoint here is they have said they expect Switch to last 7 years and we are indeed getting close to that end date, they now only have a single platform to support (so no more faffing around with quick and easy updates on a side platform) and Dane is architecturally different and at the low end will likely be 3X more powerful. That is usually the definition of a successor.

What they call it will be interesting, and they well could abandon the Switch branding, but that's not really supported by anything Iwata has said about NX or anything Nintendo has said since, just our concern about Nintendo's past, which I have to re-iterate is less and less relevant given Genyo Takeda and most of the old guard are retired, Iwata is gone and Switch was intended to be a fresh start with fresh faces helming its design and those people probably are still in charge given their prominence in every direct. We'd be looking more at what Shinya Takahashi, Kouizumi and Furukawa would do, but we simply don't know, or I don't.

I want to reference the 2017 Nintendo Switch hardware interview where they talk about the design philosophy behind it,m but I'm drawing a blank. Does anyone have a link to that interview? Only the Switch OLED interview is coming up when i search

Edit: found it. Here

Nintendo talks Switch - origins / creation, going with one screen, name, online play, January event, much more

A little while back, Famitsu published an interview with Nintendo’s Shinya Takahashi and Yoshiaki Koizumi all about Switch in one of its issues. It was extremely extensive, covering the system’s origins and early development (including some involvement from Satoru Iwata). The two developers also...

nintendoeverything.com

 

[PedroNavajas]

There's no "suddenly" about it. The manufacturing process would have been chosen at the start of the SoC design process, which is years before hardware ships. Changing manufacturing process mid-design would be an extremely drastic move, resulting in significant costs and delays.

But at the same time, chips redesigns to use newer nodes are not rare for the industry and even Nintendo did one for the current switch. Is true that a last time chip redesign would mean a significant delay, but the theory would be that the product is already delayed.

 

[Skittzo]

According to two contemporary reports, the 3DS was estimated to cost about $100 in BoM at launch for its $250 retail price. Of course there's packaging, distribution, retailer margins, etc., but that still points to a comfortable profit. Particularly when the DSi BoM came to $75.58 at launch for a $170 retail price (same article). The DSi was a mid-gen refresh/pro model which I would definitely expect to be sold at a profit, but the 3DS had a higher margin, particularly when you consider that the packaging/distribution costs would have been about the same for both of them. Then the Switch was estimated at a $257 BoM for a $300 retail price. These are obviously all estimates, and could be wrong, but they're the most accurate info I can find.

That's surprising to hear since they definitely were selling it at a loss after the $75 price cut. I guess shipping and packaging costs made up more of the price per unit than I thought.

The issue I see with the "current environment for consumer electronics is insatiable", is that when they were making the important design decisions on this, that wouldn't have been obvious at all. I'd guess that somewhere around late 2019 to early 2020 they would have had to come to an agreement with Nvidia on the high-level details of Dane (ie manufacturing process, CPU and CPU architectures, and numbers of cores, SMs, etc.) to give enough time for the chip to be designed, taped out and manufactured. At that point if they'd based their plans on "we can sell anything at $500 and people will just eat it up", they would have been laughed at.

Well yeah I don't think they designed it with a $500 launch price in mind, but I do think they're likely having discussions internally about driving up the price just based on current demand for any and all consumer electronics.

I also don't think it's entirely true. Demand is high, but Switch is easily enough available and sales are starting to drop. Pro style consoles sold alongside the base model also seem to sell much less than the base model. If the current circumstances existed during the launch of the PS4 Pro and Xbox Series X, I can't see there being anywhere near the demand for the PS4 Pro there is now for the PS5. If positioned as a Switch 4K, I could definitely see there being a decent early-adopter demand, but no guarantee beyond that. I don't think playing the same games at higher resolution is that big of a draw for most people, particularly the market that is currently happy to pay $300 for the Switch.

Switch isn't really all that available in most major markets, even now. In the US Amazon is backordered for every model but like a handful of the regular red/blue one. In Japan it's getting harder to find again, especially the OLED model.

Considering the OLED model has a price that serves as a premium entry point at this stage, the fact that it specifically is essentially sold out globally should indicate that a higher end 4k model that actually has more value in its upgrades would sell similarly well. PS4 Pro and XSX aren't exactly comparable specifically because we're not talking about a strictly "pro" style model here.

And I honestly disagree, I do think if the current climate existed during the launch of the PS4 Pro and X1X they would absolutely have sold out instantly just like the base models would have been doing.

 

[Dakhil]

Samsung's 7LPP would have existed at the time, although Nvidia never designed anything for it, and to match (or at least get close to) TSMC's 7nm processes, they would have had to look at 5LPE or 5LPP, which would have been very much unknown quantities.

Actually, Nvidia was originally supposed to use Samsung's 7LPP process node for the fabrication of the consumer Ampere GPUs according to Yoo Eung-joon, the chief of Nvidia Korea, according to the Korea Herald's report. So assuming the Korea Herald's report is accurate, Nvidia probably has initially designed the consumer Ampere GPUs for Samsung's 7LPP process node before ultimately deciding on using Samsung's 8N process node for consumer Ampere GPUs probably due to Samsung's 7LPP process node having yield issues.

 

[ReddDreadtheLead]

So, is Dane the next Nintendo Switch revision or successor? I don't believe Nintendo would waste a lot of efforts in creating 4K Switch revision when they can use that for the Switch successor.

It’s a successor, whether they position it to play the part of the PS5 (aka the PS4 Pro Pro, yes twice) remains to be seen.

I think the biggest counterpoint here is they have said they expect Switch to last 7 years and we are indeed getting close to that end date, they now only have a single platform to support (so no more faffing around with quick and easy updates on a side platform) and Dane is architecturally different and at the low end will likely be 3X more powerful. That is usually the definition of a successor.

Has Nintendo actually ever explicitly stated that they were looking for a 7 year lifespan or is that the inferred meaning of that? I keep seeing people mention 7 years but haven’t found anything that explicitly says 7 years. Just a longer lifespan.

And they’ve been saying “middle of lifecycle” not for a single year, but for a number of years now.

If we are to infer from this, it could be this: the beginning (growth, 3 years), the middle (the maturation, 3 years) and the end (the decline, 3 years), Nintendo could quite literally also be looking to have a, yes, 9 year lifespan for their platform before releasing a “successor”. Or whatever they want to call it. this would conveniently align with Atlan though

And not start (first half, 3 years), middle (half waypoint,~1year) and then end (the other half, 3 years). Like a perfect angle formed at the mid way point.

I don’t really think Nintendo views this as the latter, but as the former scenario. The former is usually how it’s run in a business for a product saturation, but the latter is more like a race which they don’t seem to be into.

But at the same time, chips redesigns to use newer nodes are not rare for the industry and even Nintendo did one for the current switch. Is true that a last time chip redesign would mean a significant delay, but the theory would be that the product is already delayed.

This isn’t likely. The foresight for this would be nearly future vision-esque where they knew of a chip shortage, a massive pandemic, etc. which, no one even expected to be this severe. The redesign would have been so costly….. oof.

The issue I see with the "current environment for consumer electronics is insatiable", is that when they were making the important design decisions on this, that wouldn't have been obvious at all. I'd guess that somewhere around late 2019 to early 2020 they would have had to come to an agreement with Nvidia on the high-level details of Dane (ie manufacturing process, CPU and CPU architectures, and numbers of cores, SMs, etc.) to give enough time for the chip to be designed, taped out and manufactured. At that point if they'd based their plans on "we can sell anything at $500 and people will just eat it up", they would have been laughed at.

I think it’s possible that, with the demand of the switch even before say, late 2019, which was pretty high, they could have designed the product as a 400 dollar product. The product that eventually came out, Aula, wasn’t designed it seems to be a 300 dollar product but a 350 despite not changing anything internally (drastically). Having the better screen being one of the features for it among the others. The dock was probably a last minute thing though, but the switch itself probably wasn’t.

And if course the packaging, the software, the retail, etc., are included in this.

 

[Dekuman]

Has Nintendo actually ever explicitly stated that they were looking for a 7 year lifespan or is that the inferred meaning of that? I keep seeing people mention 7 years but haven’t found anything that explicitly says 7 years. Just a longer lifespan.

And they’ve been saying “middle of lifecycle” not for a single year, but for a number of years now.

If we are to infer from this, it could be this: the beginning (growth, 3 years), the middle (the maturation, 3 years) and the end (the decline, 3 years), Nintendo could quite literally also be looking to have a, yes, 9 year lifespan for their platform before releasing a “successor”. Or whatever they want to call it. this would conveniently align with Atlan though

And not start (first half, 3 years), middle (half waypoint,~1year) and then end (the other half, 3 years). Like a perfect angle formed at the mid way point.

I don’t really think Nintendo views this as the latter, but as the former scenario. The former is usually how it’s run in a business for a product saturation, but the latter is more like a race which they don’t seem to be into.

Nintendo started mentioning Switch was in the middle of its lifespan in 2020, when it was entering its 4th year, that's where the 7 years came from. They did re-iterate it was in the middle of its lifespan last year (2021)

 

[Skittzo]

Nintendo started mentioning Switch was in the middle of its lifespan in 2020, when it was entering its 4th year, that's where the 7 years came from. They did re-iterate it was in the middle of its lifespan last year (2021)

They're saying it's still in the middle phase, not the growth phase or decline phase.

That says nothing whatsoever about how long they want it to last.

 

[Dekuman]

They're saying it's still in the middle phase, not the growth phase or decline phase.

That says nothing whatsoever about how long they want it to last.

From the translations, it doesn';t sound like he's talking about the product lifecycle (which is usually introduction/growth/maturity/decline) there's no reference to 'middle' anywhere but i'll grant you that's a possibility.

What they say this year will be telling.

 

[ILikeFeet]

I feel the "middle of the life span" stuff is uncommittal. there's no definition there so it can mean whatever it needs to mean

 

[Deleted member 1324]

Hopefully MVG hears more details that he can share during GDC 2022 at late March 2022.


Outside of whether the launch window has changed or remained the same, and maybe estimated performance metrics, I don't expect much information.

Caught in 4k*

MVG is @NateDrake s source!!

*4k rendering achieved with DLSS.

 

[fwd-bwd]

So what I'm concerned about is that Nintendo is releasing a sort-of-successor, but isn't committing fully to just releasing a Switch 2, and in particular that they're not committed to continuing on with the Switch line as the future of their hardware endeavours.

I don't think Samsung 8nm is "out", it's been reported by Kopite7Kimi, who is a very reliable leaker on these things, and I have no reason to disbelieve him. The reason I brought it up is that I think it's an indication that they're not just going for a straight-forward Switch 2.

You might be reading too much into the (rumored) decision of fabricating Dane on the Samsung 8nm. For better or worse, the Switch SOCs are tethered to Nvidia's automotive products*. It started with Drive PX (TX1). When Drive PX 2 (TX2 on 16nm) came along, the Switch SOC followed it to 16nm too. So if Dane indeed will be using the 8nm process, it may not be indicative of Nintendo's succession strategy but simply what Nvidia makes available to them.

* This could change if Nvidia decides to re-enter the mobile/Chromebook SOC market by forging a new partnership with, say, MediaTek.

 

[SiG]

I still feel we are putting too much faith into Kopite7Kimi's claims.

 

[Kenka]

Any chance that Dane would be a 'smaller' Switch 2, just like the Xbox Series S is the 'smaller' version of the Series X?

 

[Hermii]

I still feel we are putting too much faith into Kopite7Kimi's claims.

Because he is always wrong?

 

[PedroNavajas]

This isn’t likely. The foresight for this would be nearly future vision-esque where they knew of a chip shortage, a massive pandemic, etc. which, no one even expected to be this severe. The redesign would have been so costly….. oof.

They wouldn't need foresight. Mariko released 2 years after launch after all. If the next chip was designed to be out by 2021/2022 and Nintendo/Nvidia had a redesign scheduled for 2023-2024 launch (like they did with Mariko), then Nintendo could have just decided not to put the 8nm model into production and launch later with the second SoC.

I still feel we are putting too much faith into Kopite7Kimi's claims.

He might be right, but wrong in the end (Plans change).

 

[Shaoran Hyku]

Nintendo Cuts Switch Outlook Again on Supply, Logistics Jam

Nintendo Co. cut its Switch sales outlook for the second quarter in a row as console makers grapple with a chronic chip shortage that is likely to continue this year.

www.bloomberg.com

Quote from Furukawa:

“Switch is just in the middle of its lifecycle and the momentum going into this year is good,” Furukawa said on a call after the earnings report. “The Switch is ready to break a pattern of our past consoles that saw momentum weakening in their sixth year on the market and grow further.”

Let see you in 2027 I guess
(Not that they won't release Switch 2 until end of lifecycle, but is going to be interesting how late they will release it and how will affect the lifecycle of the original switch)

 

[Hermii]

Nintendo Cuts Switch Outlook Again on Supply, Logistics Jam

Nintendo Co. cut its Switch sales outlook for the second quarter in a row as console makers grapple with a chronic chip shortage that is likely to continue this year.

www.bloomberg.com

Quote from Furukawa:

Let see you in 2027 I guess
(Not that they won't release Switch 2 until end of lifecycle, but is going to be interesting how late they will release it and how will affect the lifecycle of the original switch)

He could be talking about software lifecycle and not hardware lifecycle. I expect them to release games for Switch, years after Dane Switch is out.

 

[ReddDreadtheLead]

They wouldn't need foresight. Mariko released 2 years after launch after all. If the next chip was designed to be out by 2021/2022 and Nintendo/Nvidia had a redesign scheduled for 2023-2024 launch (like they did with Mariko), then Nintendo could have just decided not to put the 8nm model into production and launch later with the second SoC.


He might be right, but wrong in the end (Plans change).

Going from the 22/20nm to 16/14nm Finfet is not the same as going from 8nm DUV to a 7-5nm EUV. If there was a delay and a redesign, you wouldn’t see that chip really for 2023/24. Probably 25/26 or later in the current times where it’s not really easy to do this.

Nintendo Cuts Switch Outlook Again on Supply, Logistics Jam

Nintendo Co. cut its Switch sales outlook for the second quarter in a row as console makers grapple with a chronic chip shortage that is likely to continue this year.

www.bloomberg.com

Quote from Furukawa:

Let see you in 2027 I guess
(Not that they won't release Switch 2 until end of lifecycle, but is going to be interesting how late they will release it and how will affect the lifecycle of the original switch)

Reading this…. It enforces that they are looking at it as a Beginning phase that was 2-3 years (growth), the middle phase which has been about 2-3 years I think? (Maturation) and they have a view or goal of the end phase which can be 2-3 years.


So the successor of the switch they aren’t planning for it to come out anytime soon, but probably like you said, in 2027 or around there.

And by successor, I’m referring to moving away from the switch concept and try something anew that differentiates itself from the switch.

Like Wii to Wii U, not DS to 3DS.

But let’s be honest with ourselves, does that really mean they won’t release a new piece of hardware in between? No, but this could mean they aren’t leaving the switch concept soon.

 

[SiG]

He might be right, but wrong in the end (Plans change).

This is basically where I'm at right now.

They will probably aim for a more "readily available" (and perhaps more contemporary) SoC, but everything does seem to point out that Nvidia doesn't any other gaming focused SoCs right now, and if they did, Nintendo probably already called dibs on it.
The situation of not announcing anything means that they coould easily get away with a non-commital stance on whatever chip they did have in the backburner. That, and the last thing we heard was it "not being taped out" means that they definitely might have to aim for a much later release schedule due to shortages. They wouldn't be forced to go along with a chip they can't even meet production for, and their project would probably be rolled over to an entirely new SoC.

 

[Baobab]

after this (repeted) statement my beleif in a new switch before autumn 2023 is fading ......

 

[karmitt]

Nintendo Cuts Switch Outlook Again on Supply, Logistics Jam

Nintendo Co. cut its Switch sales outlook for the second quarter in a row as console makers grapple with a chronic chip shortage that is likely to continue this year.

www.bloomberg.com

Quote from Furukawa:

Let see you in 2027 I guess
(Not that they won't release Switch 2 until end of lifecycle, but is going to be interesting how late they will release it and how will affect the lifecycle of the original switch)

No. It just means that this next device is a revision, and most/all Nintendo first party titles will still support the original Switch models first.

If anything this just solidifies what Nate and others said last year - it’s mostly for improved framerates and resolutions, and it’ll have select exclusives that take full advantage of it. I really fail to see how anything’s changed here.

 

[fwd-bwd]

From the analyst Q&A (note this is David Gibson's transcription not an official record, and could be inaccurate/incomplete):

 

[fwd-bwd]

Furukawa on metaverse (machine translated from Sankei Shimbun report):

As for entering the "metaverse" market, which is a virtual space that allows users to experience something similar to reality on the Internet, he said, "It is a field with great potential," but added, "If we can find a Nintendo approach that provides fresh surprises and fun, we will consider it, but at the moment it is difficult."

Respect