StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (Read the staff posts before commenting!)

[Pokemaniac]

I’m pretty sure that it was in 2019 that the news came out as it was two years prior to Aula, ie the OLED model releasing.

Switch OLED first appeared in public firmware builds in April 2020. It's plausible someone could have leaked it from developer builds a bit earlier, but I don't recall that happening.

 

[ReddDreadtheLead]

I don't see there being any good reason to think that this device won't still be a revision. That's the last we'd heard from Nate and others, and I don't think a year is going to change that.

For now I do think we're looking at a more ambitious GBC situation as you've phrased it.

I mean, let’s look at it this way, would the PS5 be considered a revision of the PS4? This “Dane” thing isn’t really fulfilling the PS4 Pro scenario most likely, and would follow the same traditional move that the Series had over the One S|X platform, no true next gen first party exclusives at the start for like two years or so, and then it will transition to having first party exclusives that run on the Series S|X and don’t run on the older platforms.

Even if Dane doesn’t have true exclusives for the first 2 years or so from Nintendo, it wouldn’t really be that different from what MS did with the Series and they place the Series as the successor, an iterative successor more or less, but a successor nonetheless. Not really a “revision” in the traditional console sense.

Define "we". As I understand it it only became pubic knowledge in April 2020 but I guess there may have been insider discussions about it beforehand?

I vaguely remember mention of other units (including a “better display” one) a few months before that.

I personally think it makes since for another revision “4K” switch. Get a couple more years our of the switch name. I’m not sure if their “next-gen” console will be in the “switch” family. I personally am expecting a brand new console with some sort of VR capabilities. That makes a lot of sense to me. We shall see though. They have big decisions to make.

Do you mean that they will ditch the switch concept and jump into VR? That’s very unlikely, the market for VR while growing is still very small relative to the overall console market that is +250M.

Maybe incorporation of VR into the “Switch” action/concept is what they will do if they decide to take VR more seriously, like a dockabke headset or something. The successor to LaboVR.

What did Grubb say?

He speculated, but the fact that he, someone who usually knows more than we do being in this industry, speculated on something that he commented on previously leads me to believe he doesn’t actually know much for a new console.

Team2023 will win this race! kek

I mean, I keep mentioning this but Nintendo has been iterating on adding on a new piece of hardware every two years that keeps player interest somehow regardless of the software. An iterative model can function differently than the way Apple has it, while still maintaining the same psychological aspects of an iterative model, it can be something that is new and fresh for the market to look into.

2017 OG switch, 2019 Switch Lite (plus better battery model), 2021 OLED model, 2022? Maybe but I’m not confident but it’s not out of the picture. 2023 seems more like what they will aim for.


And it’ll continue getting switch 1 games until ‘27 maybe?

With the possibility of Xenoblade 3 being potentially delayed into 2023 now on the table...

Q1 2023 Dane please

I wish. I’m moving my prediction from H1CY23 to H2CY23 at this point.

 

[ShadowFox08]

Fair enough. Valve mentioned that the 4 cores/8 threads Zen 2 CPU on the Steam Deck runs at a range of 2.4 - 3.5 GHz, which seems to suggest that the Steam Deck probably won't be running at 3.5 GHz for sustained periods of time.

I think the switch 2 model will be a tad bit above 1Ghz, but hopefully with 8 core CPU.

if switch 2 can get to a 3x power gap with x series s, then I kind of expect steam deck will be like half as fast as X box series s due to half as cores cores (but will one be dedicated to the OS?)

Well, if we're treating money as no object we may as well go with the latest and greatest. Slightly more seriously, though, I think some form of HBM could be a good solution for a (high-end) portable gaming device, but it would probably want to be a tweaked low-power variant. HBM2E is already pretty much the most power efficient memory out there, at around 4 pJ/bit (picoJoules per bit), but scaled up to 460.8GB/s, that's still just under 15W for the package. Much less than something like GDDR6 for the same bandwidth, and absolutely worthwhile for the high-end HPC GPUs it's used on, but far too much for a portable console.

HBM itself is inherently efficient for two main reasons. The first is physical, in that it's connected on package via an interposer rather than over PCB traces, and the second is, because of that interposer connection, it's able to use a very wide bus (1024 bit per package, vs 16-64 bit for other memory standards) at a lower clock, and wide and slow is more efficient than narrow and fast. However, the latest standards like HBM2E and HBM3 are basically wide and fast. HBM3 chips are going to hit 6.4Gb/s per pin, which is the same as LPDDR5, and faster than almost all currently available DDR5. Again, this makes sense for the kind of server/HPC applications that need this much bandwidth, and it's still no doubt the most power-efficient way to get it, but it's going in quite a different direction than a portable gaming device would require.

That said, if let's say Nvidia and Samsung were to work together to make the most efficient HBM possible, I'd say it could be very efficient indeed. Nvidia actually published a proposal for something called FGDRAM back in 2017 (PDF link), which is a variant of HBM which can significantly reduce power consumption, to around 2 pJ/bit. This is still targeting very high performance of around 1TB/s (basically it's an alternative to what would become HBM3), so I have no doubt you could improve efficiency further by bringing clocks down considerably and going back to a wide and slow approach.

Let's say they could hit 1.5 pJ/bit with FGDRAM and much-reduced clocks. As far as I'm aware the LPDDR4 in the original Switch consumed around 2W to provide 25.6GB/s, and with this memory you could hit around 165GB/s in the same 2W envelope. If you pushed the power envelope up to 5W (let's say for docked mode) you could get to 415GB/s, which is not bad at all for a device the same form-factor as a Switch. A device which would be vastly more expensive than the Switch, but the same form-factor nonetheless.

Just how much does that memory cost compared to lpddr4x and 5,

I don't see there being any good reason to think that this device won't still be a revision. That's the last we'd heard from Nate and others, and I don't think a year is going to change that.

For now I do think we're looking at a more ambitious GBC situation as you've phrased it.

We already got 3 revisions now. V2 switch, lite and OLED just recently.

GBC is just mainly a recoloring. Switch 2 with new ampere you tech that can match PS4 performance in docked mode as well as 720p PS4 in handheld with a new CPU clock that multiples times faster is not simply a revision.
And with DLSS thrown in, we can see Switch 2 bring 8-10x more GPU performant. This wide of a power gap would be a successor. Not a pro model or revision.

 

[ReddDreadtheLead]

Just how much does that memory cost compared to lpddr4x and 5

How much does the Switch Lite cost?

So, around that at the upper end. , some I remember seeing as 160 though….

But it also has to do with the complexity of the design.

 

[Dakhil]

Just how much does that memory cost compared to lpddr4x and 5,

16 GB of HBM2 with 4 stacked DRAM dies was estimated to be roughly $120 in December 2019. In comparison, 6 GB (48 Gb) of LPDDR4X seems to cost roughly $18 at the lowest and $20 at the highest, presumably per chip, in December 2019.

So I imagine HBM2 right now is still significantly expensive in comparison to LPDDR5.

 

[PedroNavajas]

The lack of 8 main cores on mobile devices, including the Deck, is a strong evidence that more than 4 cores is not desirable for the power profile. IMHO. Everything else being equal, a 4 cores at 2Ghz is much better than 8 cores at 1Ghz.

However, the "need" of the os taking 1 core makes this a bit harder. I know that orin has twelve A78's, but is possible for the custom Dane to have mixed cores, including a couple of low power ones for the OS?

 

[SiG]

Well, at this point, the only thing coming after Orin is a possible process node change for those chips (which wouldn't provide enough of a performance boost to matter that much) and the next generation Atlan chips, but those aren't going to be available until 2025 at the earliest according to Nvidia (and who knows what's going to happen now thanks to chip shortages).

So it's gonna be a Dane chip as we've already discovered or an Atlan variation for 2025/2026 release, take your pick.

it's whatever they could produce more of within the given timeframe. A preeminent process node change might end up being better for performance binning, and not to mention power consumption which Nintendo tends to prioritize.

 

[karmitt]

I mean, let’s look at it this way, would the PS5 be considered a revision of the PS4? This “Dane” thing isn’t really fulfilling the PS4 Pro scenario most likely, and would follow the same traditional move that the Series had over the One S|X platform, no true next gen first party exclusives at the start for like two years or so, and then it will transition to having first party exclusives that run on the Series S|X and don’t run on the older platforms.

The difference is that I don’t expect Nintendo to drop the Switch for another 4-5 years. When they say they want to redefine / extend the traditional lifecycle, I suspect they mean it - first party games will continue with few exceptions until 2026.

I could be way off base, but that’s my expectation. When murmurings came about early last year about the new hardware, it wasn’t just Nate reporting his understanding that it’d be a revision. Imran also mentioned that we should “expect this thing to be primarily for FPS boosts and resolutions, not a Switch 2. Think more Pro than Super.” I’ve not really changed my expectations since.

 

[ReddDreadtheLead]

The lack of 8 main cores on mobile devices, including the Deck, is a strong evidence that more than 4 cores is not desirable for the power profile. IMHO. Everything else being equal, a 4 cores at 2Ghz is much better than 8 cores at 4Ghz.

However, the "need" of the os taking 1 core makes this a bit harder. I know that orin has twelve A78's, but is possible for the custom Dane to have mixed cores, including a couple of low power ones for the OS?

Two things that have to be considered when discussing something such as this:

1. This is a console

The environment that the device would have makes use of a consistent level of performance, not like cellular devices and tablets which has the little cores for OS level tasks and the big cores for performance necessary tasks, but these being at high at all times consumes more energy, which is what the little cores can help with as they consume little energy by comparison and can aid the system in lowered power reduction.


The way the switch is has more similarities with a laptop than a mobile phone, even if ARM based, as they both only make use of performance cores, not efficiency cores.

And ARM doesn’t have a CPU that clocks to 4GHz



and it isn’t like they can’t have the little cores for this next whatever, it is just completely up to what they want out of it.

2. This is not x86


Steam deck has 4 cores, because valve chose that for the goal they have for the Steam deck where it can have a playable battery life and a bare minimum of performance that broadens the reach of STEAM to a wider audience. The Aya Neo Next has 8 cores and 16 threads, using the Rembrandt APU from AMD. x86 is not as efficient as ARM CPUs when discussing performance per watt, which the ARM CPUs greatly outclass the x86 CPUs.

The difference is that I don’t expect Nintendo to drop the Switch for another 4-5 years. When they say they want to redefine / extend the traditional lifecycle, I suspect they mean it - first party games will continue with few exceptions until 2026.

I could be way off base, but that’s my expectation. When murmurings came about early last year about the new hardware, it wasn’t just Nate reporting his understanding that it’d be a revision. Imran also mentioned that we should “expect this thing to be primarily for FPS boosts and resolutions, not a Switch 2. Think more Pro than Super.” I’ve not really changed my expectations since.

Right, that was for the time. Now for a device that isn’t even for certain releasing in 2022? What sense does it make to do that? Keep it as a device that does “res and FPS boost”?

And mind you, Imran isn’t exactly the most accurate, and is iffy at times with this type of information.

 

[Intoxicate]

Every Switch 2 now or next year wouldn't be a big step in my view. DLSS or not. I sold my PS4 since I didn't care for the "slightly" better graphics and was fine with the Switch. Look at the PS4 vs. PS5, it's the same stuff again. I was getting out of gaming with the Wii U after loving the Wii. It was more or less the same just in HD. I get why people want BotW with 60fps but then, for most people the games look and run fine and the OLED gives some more fidelity which to some is the same upgrade or even better then e.g. an overclocked Mariko SOC could offer. So I'm perfectly fine with them waiting until 23 or 24 when the manufacturing processes actually allows for a generational leap which is significant. I don't care for a "New" 3DS kinda upgrade.

 

[PedroNavajas]

Two things that have to be considered when discussing something such as this:

1. This is a console

The environment that the device would have makes use of a consistent level of performance, not like cellular devices and tablets which has the little cores for OS level tasks and the big cores for performance necessary tasks, but these being at high at all times consumes more energy, which is what the little cores can help with as they consume little energy by comparison and can aid the system in lowered power reduction.


The way the switch is has more similarities with a laptop than a mobile phone, even if ARM based, as they both only make use of performance cores, not efficiency cores.

And ARM doesn’t have a CPU that clocks to 4GHz



and it isn’t like they can’t have the little cores for this next whatever, it is just completely up to what they want out of it.

2. This is not x86


Steam deck has 4 cores, because valve chose that for the goal they have for the Steam deck where it can have a playable battery life and a bare minimum of performance that broadens the reach of STEAM to a wider audience. The Aya Neo Next has 8 cores and 16 threads, using the Rembrandt APU from AMD. x86 is not as efficient as ARM CPUs when discussing performance per watt, which the ARM CPUs greatly outclass the x86 CPUs.

WHile you might have a point with Phones, and even the Deck, Apple's M1 also has a 4 + 4 configuration.

 

[ReddDreadtheLead]

WHile you might have a point with Phones, and even the Deck, Apple's M1 also has a 4 + 4 configuration.

Apple’s M1 is the exception in laptops, not the norm. Most run on an AMD or Intel processor, granted that’s also because they haven’t moved to ARM and Microsoft hasn’t put the work to make it happen.

 

[karmitt]

Right, that was for the time. Now for a device that isn’t even for certain releasing in 2022? What sense does it make to do that? Keep it as a device that does “res and FPS boost”?

And mind you, Imran isn’t exactly the most accurate, and is iffy at times with this type of information.

He’s more likely to have spoken to people with dev kits than most. He also said he was surprised when Bloomberg came out with the article saying that the device was launching in 2021. I know this thread is a heck of a lot more knowledgeable about the capabilities and possibilities in tech, but I have zero reason to believe plans around position changed right now, even if the ceiling for tech they could be using is rising.

Is it really all that unreasonable that a device Nintendo releases in late 2022 has a primary aim of running first party Switch titles at 4K 30-60 ? This still sounds really substantial for Switch’s form factor and price.

 

[ShadowFox08]

Every Switch 2 now or next year wouldn't be a big step in my view. DLSS or not. I sold my PS4 since I didn't care for the "slightly" better graphics and was fine with the Switch. Look at the PS4 vs. PS5, it's the same stuff again. I was getting out of gaming with the Wii U after loving the Wii. It was more or less the same just in HD. I get why people want BotW with 60fps but then, for most people the games look and run fine and the OLED gives some more fidelity which to some is the same upgrade or even better then e.g. an overclocked Mariko SOC could offer. So I'm perfectly fine with them waiting until 23 or 24 when the manufacturing processes actually allows for a generational leap which is significant. I don't care for a "New" 3DS kinda upgrade.

It's gonna be a huge upgrade when you count performance differences with DLSS. Without it, it's 3x. But going from OF Switch to potentionally PS4 pro performance for GPU w/ DLSS is like 8-10x. And CPU will be at least 6x.
bandwidth will be 4x, not to mention more cache.

The lack of 8 main cores on mobile devices, including the Deck, is a strong evidence that more than 4 cores is not desirable for the power profile. IMHO. Everything else being equal, a 4 cores at 2Ghz is much better than 8 cores at 4Ghz.

However, the "need" of the os taking 1 core makes this a bit harder. I know that orin has twelve A78's, but is possible for the custom Dane to have mixed cores, including a couple of low power ones for the OS?

I really doubt Nintendo would ever go beyond 2Ghz for CPU. ARM isn't meant to go that high up to 4GHz anyway I think.

1-1.5Ghz 8 core CPU (at least at 1Ghz) will be more power efficient then 2 GHz 4 core.

 

[PedroNavajas]

It's gonna be a huge upgrade when you count performance differences with DLSS. Without it, it's 3x. But going from OF Switch to potentionally PS4 pro performance for GPU w/ DLSS is like 8-10x. And CPU will be at least 6x.
bandwidth will be 4x, not to mention more cache.


I really doubt Nintendo would ever go beyond 2Ghz for CPU. ARM isn't meant to go that high up to 4GHz anyway I think.

1-1.5Ghz 8 core CPU (at least at 1Ghz) will be more power efficient then 2 GHz 4 core.

I fixed my typo, sorry... I intended to say 4x2Ghz >8x1Ghz in performance.

 

[PedroNavajas]

1-1.5Ghz 8 core CPU (at least at 1Ghz) will be more power efficient then 2 GHz 4 core.

More power efficient, but less performant. Mutithreading is always going to be slower than single threading for the same on-paper performance. The reason why a pentium dual core + 750ti could run Alien Isolation at 60fps, while the PS4 couldn't.

 

[Z0m3le]

Is there a solid source for this, or just your instinct? I'd be so bloody happy if that happened, but feels like we'd know by now surely?

No, just speculation.

Couldn’t they just do a 25% overlock on the CPU and GPU and call it a day or did they honestly hoped to run at Max clocks?

They were looking for ~double the clock speed, at least from the GPU, and maybe 1.78GHz CPU (same as loading speed)

Am I reading this right that you yourself know of a third party working on a Dane game? Or are you referencing Nate's comments some weeks back.

Not asking you to divulge any more info or anything, just sniffing out any crumb of hopium. That Grubb comment the other day got me spooked haha

I talk to insiders, and was able to talk to someone directly working on a project that a known insider vetted to me. I don't know anything about the developer, I personally asked them not to give me any details like that. I don't want to hold anyone's job in my hands, but I do trust the insider, and the person I talked to definitely sounded legit enough.

 

[Z0m3le]

dp

 

[PedroNavajas]

They were looking for ~double the clock speed, at least from the GPU, and maybe 1.78GHz CPU (same as loading speed)

Sounds to me like they were aiming for running the docked mode while on Handheld. And it makes perfect sence, since it would have been an easy upgrade.

 

[ReddDreadtheLead]

He’s more likely to have spoken with people with dev kits than most. He also said when Bloomberg came out with the article saying that the device was launching in 2021 that he was surprised. I know this thread is a heck of a lot more knowledgeable about the capabilities and possibilities in tech, but I have zero reason to believe plans around position changed right now, even if the ceiling for tech they could be using is rising.

Is it really all that unreasonable that a device Nintendo releases in late 2022 has a primary aim of running first party Switch titles at 4K 30-60 ? This still sounds really substantial for Switch’s form factor and price.

@ the bolded: Wasn’t that Nate that said that?


Second part of post: that isn’t impossible, no I’m not really saying that, as we have worked out the math behind it and based on what is out there and available, however how likely is it to launch 1 year after the release of the NSW OLED and not 2 years after like the other models that have been introduced to the platform? If we are to credit this as a revision, wouldn’t it make sense to expect it in the same time difference of two years?

They’ve been playing their cards almost perfectly and have basically avoided a lot of overload for the product family, with very clear messaging on one over the other. This is only speculation, but the V2 model is very likely to get phased out over the course of the next fiscal year (FY23) to make room for the OLED as the next flagship and then a price drop by then to the OG model’s price range of 299.99, while the SUPER Nintendo Switch is priced at a rather expensive price of 399.99, but continuing the cross gen period for a while.

More power efficient, but less performant. Mutithreading is always going to be slower than single threading for the same on-paper performance. The reason why a pentium dual core + 750ti could run Alien Isolation at 60fps, while the PS4 couldn't.

I fixed my typo, sorry... I intended to say 4x2Ghz >8x1Ghz in performance.

While true, the 7 available cores at 1GHz would outperform the 3 available core at 2GHz in this scenario (1 core reserved to OS)

A good middle ground is also 6 Cores at 1.5-1.8(?)GHz + 2-4 little cores at 1.3GHz(?) for the OS, in this case it would outperform the both scenarios while keeping the low power profile and not dedicate a big core for the OS.

But it’s an odd design of possibly a 8-10 core CPU, something I don’t see them doing.





On another note: The weird thing about Ampere that I keep forgetting is that, they list the decompression from an SSD as a feature that the GPU can assist/do with and I’m accustomed to it taking CPU resources for that. The RTX I/O should not be ignored for this as a feature even if it’s their own rebranded form of RTX I/O (Mega Mushroom I/O? lol) to not use of a CPU for the same job.


Let’s assume it is 8 SMs, for conversation sake, can’t 1 SM be dedicated just for RTX I/O? Or is it something that is shared across all the SMS as a minor GPU resource?

PS5 and SX|S have a separate CPU core or 2 for this specific purpose I think?


and don’t you need DirectStorage for this?

 

[ILikeFeet]

@ the bolded: Wasn’t that Nate that said that?


Second part of post: that isn’t impossible, no I’m not really saying that, as we have worked out the math behind it and based on what is out there and available, however how likely is it to launch 1 year after the release of the NSW OLED and not 2 years after like the other models that have been introduced to the platform? If we are to credit this as a revision, wouldn’t it make sense to expect it in the same time difference of two years?

They’ve been playing their cards almost perfectly and have basically avoided a lot of overload for the product family, with very clear messaging on one over the other. This is only speculation, but the V2 model is very likely to get phased out over the course of the next fiscal year (FY23) to make room for the OLED as the next flagship and then a price drop by then to the OG model’s price range of 299.99, while the SUPER Nintendo Switch is priced at a rather expensive price of 399.99, but continuing the cross gen period for a while.



While true, the 7 available cores at 1GHz would outperform the 3 available core at 2GHz in this scenario (1 core reserved to OS)

A good middle ground is also 6 Cores at 1.5-1.8(?)GHz + 2-4 little cores at 1.3GHz(?) for the OS, in this case it would outperform the both scenarios while keeping the low power profile and not dedicate a big core for the OS.

But it’s an odd design of possibly a 8-10 core CPU, something I don’t see them doing.





On another note: The weird thing about Ampere that I keep forgetting is that, they list the decompression from an SSD as a feature that the GPU can assist/do with and I’m accustomed to it taking CPU resources for that. The RTX I/O should not be ignored for this as a feature even if it’s their own rebranded form of RTX I/O (Mega Mushroom I/O? lol) to not use of a CPU for the same job.


Let’s assume it is 8 SMs, for conversation sake, can’t 1 SM be dedicated just for RTX I/O? Or is it something that is shared across all the SMS as a minor GPU resource?

PS5 and SX|S have a separate CPU core or 2 for this specific purpose I think?


and don’t you need DirectStorage for this?

RTXIO and Directstorage are windows specific solutions. You can implement these things on other OSs (and Nvidia already has on linux) without the need for such branding

 

[daase ko]

No, just speculation.

They were looking for ~double the clock speed, at least from the GPU, and maybe 1.78GHz CPU (same as loading speed)

I talk to insiders, and was able to talk to someone directly working on a project that a known insider vetted to me. I don't know anything about the developer, I personally asked them not to give me any details like that. I don't want to hold anyone's job in my hands, but I do trust the insider, and the person I talked to definitely sounded legit enough.

Why is an insider sharing source? Sounds like problem.

 

[Deleted member 645]

Do you mean that they will ditch the switch concept and jump into VR? That’s very unlikely, the market for VR while growing is still very small relative to the overall console market that is +250M.

Nintendo is very unpredictable. I could see them having a console that is VR specific with some twist or gimmick attached to it. I personally think they struck gold with switch and should never abandon it. All they have to do is keep working with Nvidia on architecture and how they can make the most of the hardware. Approve joycons and try to eliminate drift. If they can do this and upgrade online and chat to industry standards then they could be market leaders for years and years to come. However, this is Nintendo so they seem to always want to leave us wanting for something.

 

[Z0m3le]

Why is an insider sharing source? Sounds like problem.

The person who talked to me, told me that the specific insider could vet them, so sort of backwards. I was just online discussing Nintendo's plans and this person was in the channel with me, started to DM me and an insider I know in that channel vetted them.

 

[PedroNavajas]

@ the bolded: Wasn’t that Nate that said that?


Second part of post: that isn’t impossible, no I’m not really saying that, as we have worked out the math behind it and based on what is out there and available, however how likely is it to launch 1 year after the release of the NSW OLED and not 2 years after like the other models that have been introduced to the platform? If we are to credit this as a revision, wouldn’t it make sense to expect it in the same time difference of two years?

They’ve been playing their cards almost perfectly and have basically avoided a lot of overload for the product family, with very clear messaging on one over the other. This is only speculation, but the V2 model is very likely to get phased out over the course of the next fiscal year (FY23) to make room for the OLED as the next flagship and then a price drop by then to the OG model’s price range of 299.99, while the SUPER Nintendo Switch is priced at a rather expensive price of 399.99, but continuing the cross gen period for a while.



While true, the 7 available cores at 1GHz would outperform the 3 available core at 2GHz in this scenario (1 core reserved to OS)

A good middle ground is also 6 Cores at 1.5-1.8(?)GHz + 2-4 little cores at 1.3GHz(?) for the OS, in this case it would outperform the both scenarios while keeping the low power profile and not dedicate a big core for the OS.

Yes, I was thinking if its possible to use 4xA78 along with 2xA55 little cores for the os. Instead of 6-8xA78 that are lower clocked.

 

[Dekuman]

@Z0m3le are you saying you think this Dane based Switch may not release at all? I'm a bit baffled by that twist. What's changed.

 

[Hosermess]

@Z0m3le are you saying you think this Dane based Switch may not release at all? I'm a bit baffled by that twist. What's changed.

Don't think it's a twist. Nate also commented that there's always a chance (however very low) that Nintendo cancels the project. Both, I assume, are just covering all the bases.

 

[Z0m3le]

@Z0m3le are you saying you think this Dane based Switch may not release at all? I'm a bit baffled by that twist. What's changed.

Nothing, I haven't even been able to confirm if Dane itself was "delayed" to early 2023. I'm only saying that Nintendo has a stacked year and I think it makes more business sense to launch a 3D Mario this holiday season than botw's sequel, simply because Zelda will sell just as well outside of the holidays, and that to me means Mario this calendar year with Dane and Zelda following in feb/mar next year.

It is also always possible for hardware to be canceled before it goes into production, and with the chip shortages, they might decide to ride out 2022 and 2023 with a ton of great software and reload for a launch of new hardware at the end of 2024. I think it's unlikely mind you, but possible.

 

[Skittzo]

Nothing, I haven't even been able to confirm if Dane itself was "delayed" to early 2023. I'm only saying that Nintendo has a stacked year and I think it makes more business sense to launch a 3D Mario this holiday season than botw's sequel, simply because Zelda will sell just as well outside of the holidays, and that to me means Mario this calendar year with Dane and Zelda following in feb/mar next year.

It is also always possible for hardware to be canceled before it goes into production, and with the chip shortages, they might decide to ride out 2022 and 2023 with a ton of great software and reload for a launch of new hardware at the end of 2024. I think it's unlikely mind you, but possible.

I think Zelda is always going to be something that works better for a hardware launch than any other series. If they need to delay Dane then yeah Zelda might get delayed with it. 3D Mario (or even 2D Mario) this holiday makes a ton of sense with the movie in December, though I could see whatever Mario game they have launching after the movie too, like January or February.

Either way it'll be a good year.

 

[daase ko]

Don't think it's a twist. Nate also commented that there's always a chance (however very low) that Nintendo cancels the project. Both, I assume, are just covering all the bases.

I don't think it's an idea worth entertaining at this point.
Probably wasn't either in 2020's end when we first heard of it, given developpers where already making games for it.

 

[ShadowFox08]

@ the bolded: Wasn’t that Nate that said that?


Second part of post: that isn’t impossible, no I’m not really saying that, as we have worked out the math behind it and based on what is out there and available, however how likely is it to launch 1 year after the release of the NSW OLED and not 2 years after like the other models that have been introduced to the platform? If we are to credit this as a revision, wouldn’t it make sense to expect it in the same time difference of two years?

They’ve been playing their cards almost perfectly and have basically avoided a lot of overload for the product family, with very clear messaging on one over the other. This is only speculation, but the V2 model is very likely to get phased out over the course of the next fiscal year (FY23) to make room for the OLED as the next flagship and then a price drop by then to the OG model’s price range of 299.99, while the SUPER Nintendo Switch is priced at a rather expensive price of 399.99, but continuing the cross gen period for a while.



While true, the 7 available cores at 1GHz would outperform the 3 available core at 2GHz in this scenario (1 core reserved to OS)

A good middle ground is also 6 Cores at 1.5-1.8(?)GHz + 2-4 little cores at 1.3GHz(?) for the OS, in this case it would outperform the both scenarios while keeping the low power profile and not dedicate a big core for the OS.

But it’s an odd design of possibly a 8-10 core CPU, something I don’t see them doing.





On another note: The weird thing about Ampere that I keep forgetting is that, they list the decompression from an SSD as a feature that the GPU can assist/do with and I’m accustomed to it taking CPU resources for that. The RTX I/O should not be ignored for this as a feature even if it’s their own rebranded form of RTX I/O (Mega Mushroom I/O? lol) to not use of a CPU for the same job.


Let’s assume it is 8 SMs, for conversation sake, can’t 1 SM be dedicated just for RTX I/O? Or is it something that is shared across all the SMS as a minor GPU resource?

PS5 and SX|S have a separate CPU core or 2 for this specific purpose I think?


and don’t you need DirectStorage for this?

What's so odd about it? The Orion NX has an 8 core cpu

 

[ReddDreadtheLead]

RTXIO and Directstorage are windows specific solutions. You can implement these things on other OSs (and Nvidia already has on linux) without the need for such branding

oh this is great news then, that resolved one worry with this! And I doubt Nintendo wouldn't want to have this for developers.

Nintendo is very unpredictable. I could see them having a console that is VR specific with some twist or gimmick attached to it. I personally think they struck gold with switch and should never abandon it. All they have to do is keep working with Nvidia on architecture and how they can make the most of the hardware. Approve joycons and try to eliminate drift. If they can do this and upgrade online and chat to industry standards then they could be market leaders for years and years to come. However, this is Nintendo so they seem to always want to leave us wanting for something.

I don't buy this, how often has Nintendo drastically changed the concept of a device down to it's DNA? Nintendo can be unpredictable, but the level of unpredictability that people see from them gets so overblown online to unrealistic levels.

They did it 3-ish times, across all the consoles they ever released. The GBA to the DS changed the format from single screen to double screen but still kept the portability aspect to it. The Wii changed the whole controller scheme compared to the predecessor, the GCN. The Wii U also changed the concept again from the Wii. These three consoles were the instances where they drastically altered the wya of play compared to their predecessor where it wasn't recognizable as the successor immediately.

Hell, I've read comments of expectations that it will be a purely stationary console, no portability, just a set-top box with the carts and pro controller. And no, not as the switch TV, as the Switch 2 with a whole new processor. I'm not saying its impossible, but the chance of them radically changing a format that works is really low, and only 2 out of the 3 times they did it actually worked well.

What's so odd about it? The Orion NX has an 8 core cpu

Orin NX has 8 homogeneous CPU cores them being 2 4A78AE clusters, not 8-10 heterogeneous core configuration that leans more into more performance heavy cores than the even split of little plus big or the layout of having more little cores and 2 big cores like some mobile devices. It would be a first to do that, that I know of.

6A78+2A55 or 6A78+4A55 hasn't been done yet.

M1Pro and Max actually follow the heterogeneous format that leans into more perf cores than having it split evenly, 2LITTLE+6-8big cores, depending on the model

 

[ReyVGM]

I think I developed a way of knowing if the next console is going to be a slightly more powerful iteration or a different, more powerful hardware: by looking at Nintendo's own potential crossgen announcements.

Bear with me.

Gamecube-Wii
As the Gamecube era was ending and before the Wii was shown, everyone thought the Wii was going to be on par with the tech of the era (remember the revolutionary NURBS?). Of course, the Wii ended up being just a more capable GC with very similar graphics capabilities. That's why we saw Zelda Twilight Princess released on both consoles looking pretty much the same. We also saw several other GC ports (not remasters), such as Metroid Prime 1&2, Pikmin 1&2, Resident Evil 1&4, among others.

WiiU-Switch
With the WiiU, none of that crossgen shenanigans happened. Since the hardware was so radically different and much more powerful than the Wii, you couldn't really do any crossgen titles unless it was for a remake or remaster.

Then came the Switch. I remember when "Zelda U" was announced to be coming to the Switch, everyone assumed it was going to look much better than the WiiU version, because obviously the Switch just had to be on par with the capabilities of a PS4. As we all now know, BotW doesn't look that much different from the WiiU version, because the Switch capabilities (graphically-wise) are very similar to the WiiU. And that's why we saw a Zelda-crossgen repeat, and obviously a TON of WiiU ports since.

More powerful nextgen console or Switch 1.5?

So, now comes the point of my post.

Since the Wii days, Nintendo has been notorious for not providing detailed console specs. When the new console games are first shown, we can only (visually) tell if the hardware is significantly more powerful or not by comparing its graphics to the graphics from their previous console.
Now, I don't know when the next hardware will come, but if we see BotW2, or Prime 4, or whatever big-budget game Nintendo has at that time also being announced for the next console, then expect the next hardware to only be a minor improvement over the current Switch*. If those titles stay exclusive to the Switch and the next console has fresh nextgen-exclusive titles announced, then expect the hardware to be significantly more powerful than the Switch** and any title that is brought from the Switch would be a remaster and not a port.

*I know other developers do crossgen titles and significantly improve their graphics on the more powerful hardware, but Nintendo, so far, doesn't have a habit of doing that.
**Significantly more powerful than the Switch does not mean as powerful as a PS5.

 

[Pokemaniac]

The lack of 8 main cores on mobile devices, including the Deck, is a strong evidence that more than 4 cores is not desirable for the power profile. IMHO. Everything else being equal, a 4 cores at 2Ghz is much better than 8 cores at 1Ghz.

However, the "need" of the os taking 1 core makes this a bit harder. I know that orin has twelve A78's, but is possible for the custom Dane to have mixed cores, including a couple of low power ones for the OS?

In terms of ARM, this probably largely comes down to how clusters work and the fact that the vast majority of ARM devices benefit a lot from the power consumption benefits of having little cores. For a device with relatively fixed clocks looking for very consistent performance like the Switch, more exotic configs like 6-8 A78C cores make a lot more sense.

The Steam Deck, on the other hand, isn't super comparable. Being an x86 design, the cores are likely quite a bit larger than the average ARM cores.

 

[Dekuman]

I think I developed a way of knowing if the next console is going to be a slightly more powerful iteration or a different, more powerful hardware: by looking at Nintendo's own potential crossgen announcements.

Bear with me.

Gamecube-Wii
As the Gamecube era was ending and before the Wii was shown, everyone thought the Wii was going to be on par with the tech of the era (remember the revolutionary NURBS?). Of course, the Wii ended up being just a more capable GC with very similar graphics capabilities. That's why we saw Zelda Twilight Princess released on both consoles looking pretty much the same. We also saw several other GC ports (not remasters), such as Metroid Prime 1&2, Pikmin 1&2, Resident Evil 1&4, among others.

WiiU-Switch
With the WiiU, none of that crossgen shenanigans happened. Since the hardware was so radically different and much more powerful than the Wii, you couldn't really do any crossgen titles unless it was for a remake or remaster.

Then came the Switch. I remember when "Zelda U" was announced to be coming to the Switch, everyone assumed it was going to look much better than the WiiU version, because obviously the Switch just had to be on par with the capabilities of a PS4. As we all now know, BotW doesn't look that much different from the WiiU version, because the Switch capabilities (graphically-wise) are very similar to the WiiU. And that's why we saw a Zelda-crossgen repeat, and obviously a TON of WiiU ports since.

More powerful nextgen console or Switch 1.5?

So, now comes the point of my post.

Since the Wii days, Nintendo has been notorious for not providing detailed console specs. When the new console games are first shown, we can only (visually) tell if the hardware is significantly more powerful or not by comparing its graphics to the graphics from their previous console.
Now, I don't know when the next hardware will come, but if we see BotW2, or Prime 4, or whatever big-budget game Nintendo has at that time also being announced for the next console, then expect the next hardware to only be a minor improvement over the current Switch*. If those titles stay exclusive to the Switch and the next console has fresh nextgen-exclusive titles announced, then expect the hardware to be significantly more powerful than the Switch** and any title that is brought from the Switch would be a remaster and not a port.

*I know other developers do crossgen titles and significantly improve their graphics on the more powerful hardware, but Nintendo, so far, doesn't have a habit of doing that.
**Significantly more powerful than the Switch does not mean as powerful as a PS5.

Wii by Nintendo's own admission was going off their own roadmaps and a pretty big reset. it was a once in a blue moon event, I would personally not consider it as a datapoint unless we are expecting the new Switch to be a reset again, would be pretty dumb IMHO when all they need to do is iterate on it.

 

[karmitt]

@ the bolded: Wasn’t that Nate that said that?

Nate was also surprised, but I was making a point to bring up more than just Nate having held similar info.

Hunted down the era post from Imran. He’s saying he’d expected 2022 here, and was ‘surprised and baffled’ by the date in Bloombergs reporting. Given the reporting ended up conflating two devices, I assume previous expectations remain unchanged - a revision in 2022(+), and mostly frame rate and resolution improvements.

 

[bloopland33]

I talk to insiders, and was able to talk to someone directly working on a project that a known insider vetted to me. I don't know anything about the developer, I personally asked them not to give me any details like that. I don't want to hold anyone's job in my hands, but I do trust the insider, and the person I talked to definitely sounded legit enough.

Very cool, thanks for the reply. When I read stuff like this I find it so hard to believe this thing is a late 2023/2024 device.

 

[Skittzo]

Nate was also surprised, but I was making a point to bring up more than just Nate having held similar info.

Hunted down the era post from Imran. He’s saying he’d expected 2022 here, and was ‘surprised and baffled’ by the date in Bloombergs reporting. Given the reporting ended up conflating two devices, I assume previous expectations remain unchanged - a revision in 2022(+), and mostly frame rate and resolution improvements.

Yup this is where I'm at. So many people were saying 2022 before Bloomberg made a whoopsie that it's very silly to just throw out all of those because Bloomberg was wrong.

 

[Hermii]

Nate was also surprised, but I was making a point to bring up more than just Nate having held similar info.

Hunted down the era post from Imran. He’s saying he’d expected 2022 here, and was ‘surprised and baffled’ by the date in Bloombergs reporting. Given the reporting ended up conflating two devices, I assume previous expectations remain unchanged - a revision in 2022(+), and mostly frame rate and resolution improvements.

I think mostly res and frame rate improvements is what you would get from the first wave of games, just like it has been for ps5 and Xbox.

 

[RennanNT]

oh this is great news then, that resolved one worry with this! And I doubt Nintendo wouldn't want to have this for developers.

I don't buy this, how often has Nintendo drastically changed the concept of a device down to it's DNA? Nintendo can be unpredictable, but the level of unpredictability that people see from them gets so overblown online to unrealistic levels.

They did it 3-ish times, across all the consoles they ever released. The GBA to the DS changed the format from single screen to double screen but still kept the portability aspect to it. The Wii changed the whole controller scheme compared to the predecessor, the GCN. The Wii U also changed the concept again from the Wii. These three consoles were the instances where they drastically altered the wya of play compared to their predecessor where it wasn't recognizable as the successor immediately.

Yep, in truth they have just been iterating over hardware, adding more forms of input and outputs.

Even the Wii U was just iterating over the DS concept rather than the Wii, which was surprising because people expected a Wii successor, but not as much once you consider how much of a monster the DS was, even comparing with the Wii.

What I really consider as exceptions are the Wii and the Switch, because they re-evaluated what was necessary and cut down multiple things, including some of the selling points of their predecessor, instead of simply iterating over.

 

[Look over there]

The lack of 8 main cores on mobile devices, including the Deck, is a strong evidence that more than 4 cores is not desirable for the power profile. IMHO. Everything else being equal, a 4 cores at 2Ghz is much better than 8 cores at 1Ghz.

However, the "need" of the os taking 1 core makes this a bit harder. I know that orin has twelve A78's, but is possible for the custom Dane to have mixed cores, including a couple of low power ones for the OS?

Off the shelf/reference-wise, a single cluster of the version of DynamIQ we're assuming (not the version introduced in 2021) allows: maximum 8 cores in total, no more than 4 being 'big'. So by default, you can do 4 big/0-4 little if you want.
(when we talk about 5 and more A78, we're referring specifically to the A78C; strictly speaking, when ARM announced that, they only gave examples of 6 and 8 A78 cores, with no A55s)
That said, don't worry too much about the distinction between 4 big and 6/8 big from the perspective of power. One of the features that DynamIQ introduced was the ability to split up the cores into different voltage/frequency domains. As long as you add the requisite voltage regulators (1 per domain) when designing the chip, you can have different groups of cores run at different frequencies. For example, given 8 A78C cores, you can theoretically split them into 1 (OS)/3 (match the OG Switch)/4. For games that would only need 3 cores, the OS core can run at lower clocks and the extra 4 cores can idle, I suppose. Of course, the actual concern here is the cost/complexity of doing this. And the die space.
As for other combinations? Ehhh, should be possible, but that's entering the territory of custom job by Nvidia's engineers.

 

[ReddDreadtheLead]

I think I developed a way of knowing if the next console is going to be a slightly more powerful iteration or a different, more powerful hardware: by looking at Nintendo's own potential crossgen announcements.

Bear with me.

Gamecube-Wii
As the Gamecube era was ending and before the Wii was shown, everyone thought the Wii was going to be on par with the tech of the era (remember the revolutionary NURBS?). Of course, the Wii ended up being just a more capable GC with very similar graphics capabilities. That's why we saw Zelda Twilight Princess released on both consoles looking pretty much the same. We also saw several other GC ports (not remasters), such as Metroid Prime 1&2, Pikmin 1&2, Resident Evil 1&4, among others.

WiiU-Switch
With the WiiU, none of that crossgen shenanigans happened. Since the hardware was so radically different and much more powerful than the Wii, you couldn't really do any crossgen titles unless it was for a remake or remaster.

Then came the Switch. I remember when "Zelda U" was announced to be coming to the Switch, everyone assumed it was going to look much better than the WiiU version, because obviously the Switch just had to be on par with the capabilities of a PS4. As we all now know, BotW doesn't look that much different from the WiiU version, because the Switch capabilities (graphically-wise) are very similar to the WiiU. And that's why we saw a Zelda-crossgen repeat, and obviously a TON of WiiU ports since.

More powerful nextgen console or Switch 1.5?

So, now comes the point of my post.

Since the Wii days, Nintendo has been notorious for not providing detailed console specs. When the new console games are first shown, we can only (visually) tell if the hardware is significantly more powerful or not by comparing its graphics to the graphics from their previous console.
Now, I don't know when the next hardware will come, but if we see BotW2, or Prime 4, or whatever big-budget game Nintendo has at that time also being announced for the next console, then expect the next hardware to only be a minor improvement over the current Switch*. If those titles stay exclusive to the Switch and the next console has fresh nextgen-exclusive titles announced, then expect the hardware to be significantly more powerful than the Switch** and any title that is brought from the Switch would be a remaster and not a port.

*I know other developers do crossgen titles and significantly improve their graphics on the more powerful hardware, but Nintendo, so far, doesn't have a habit of doing that.
**Significantly more powerful than the Switch does not mean as powerful as a PS5.

These don’t really make sense, with respect to the cross gen period, if you have the same uArch, you can have a hardware that is several magnitudes more powerful or just the same and just have games run on both until the new one picks up enough steam to drop support of the older platform.

Especially in the now where it’s left to the software to speak to the hardware, Nintendo hasn’t done to the metal for a long ass time in the home console front, and a backwards compatible route was more complicated where they just had the Wii as a virtual machine in the Wii U.

Unless I’m misunderstanding what you are saying/implying here.

 

[Thraktor]

Just how much does that memory cost compared to lpddr4x and 5,

A lot! See Dakhil's post:

16 GB of HBM2 with 4 stacked DRAM dies was estimated to be roughly $120 in December 2019. In comparison, 6 GB (48 Gb) of LPDDR4X seems to cost roughly $18 at the lowest and $20 at the highest, presumably per chip, in December 2019.

So I imagine HBM2 right now is still significantly expensive in comparison to LPDDR5.

The one caveat I'd make to this is that the article claims that in 2016, HBM cost about $120/GB, which can't be right, as AMD GPUs with 4GB of HBM were selling for $550 at retail in mid-2015, so wholesale prices for the RAM alone couldn't have been $480. In any case, the 2019 figure probably isn't that far off reality, and I'd say 2+ times the cost per GB of LPDDR5 is about right.

One important factor about HBM is that it's not just the cost of the HBM stacks themselves, you also have to use more advanced (read expensive) packaging solutions to package your logic die and HBM stack(s) together with thousands of interconnects between them. I believe most chips which use HBM (eg Nvidia and AMD's HPC GPUs) use TSMC's CoWoS packaging technology, which isn't cheap. A few years ago TSMC announced a packaging solution called InFO_MS, which would allow a cheaper alternative to CoWoS for smaller HBM use-cases, eg one logic die and a single HBM stack, but I'm not sure if anything's ever used it.

 

[JoshuaJSlone]

Do we have any reason to believe Mario is being made other than banking on a movie tie-in? That would be a very fun surprise.

From Galaxy through Odyssey was a string of major console 3D Marios every 3-4 years. We're a bit past 4 years now, so we're a bit overdue with just Bowser's Fury since. If it misses this year, it will be the biggest gap since 64->Sunshine.

 

[ika]

I'm usually one of those who thinks "the sooner the better", because Nintendo often delays games not weeks but months (sometimes near a whole year) and time is something that's not infinite to us humans, but I just got a new TV (LG OLED 65" C1) and while I was configuring all settings I started to search for 4K HDR content on Youtube to see what the panel is capable of... After some cinema trailers and the typical "amazing vistas" videos, I started looking for game footage...

Long story short, if they're working on a DLSS Switch that can get EPD and MonolithSoft to do similar things to these emulator videos below, I can gladly wait until late 2023 to get BOTW2 and Xenoblade 3...

Zelda Breath of the Wild


Xenoblade X


Xenoblade Definitive Edition


Xenoblade 2 Torna


The difference these videos makes on the LG OLED TV compared with my BENQ monitor, oh my... I couldn't believe it!

 

[maniacmaison]

Wow, Xenoblade 2 looks breathtaking.

 

[Dakhil]

Job Openings - Careers at Nintendo of America

Lead Device Driver Software Engineer (NTD), Software Development

careers.nintendo.com

Lead Device Driver Software Engineer (NTD)​

⋮

DESCRIPTION OF DUTIES

• Design and develop device drivers for current and future gaming consoles
• Implement robust embedded software using modern C++ constructs
• Drive architecture definition and create associated documentation
• Lead development activities on major software subsystems, providing hands-on assistance to junior team members

SUMMARY OF REQUIREMENTS

• 7+ years' experience developing device drivers for complex SOCs, and have a fundamental understanding of power and clock tree initialization
• 7+ years' experience developing multithreaded software
• Hands-on experience with USB technologies
• Proficient with fundamental computer science data structures and algorithms
• 3+ years' professional experience developing C++ software
• Hands-on experience with lower-level device driver topics such as DMA, memory hierarchy, cache coherency, use of memory barrier instructions
• Ability to use a logic analyzer or oscilloscope for debug and validation
• Degree in Computer Engineering, Computer Science or Electrical Engineering

PREFERRED REQUIREMENTS

• MS degree
• Experience using real time operating systems (RTOS)
• Experience developing USB software stacks

* This is not a remote opportunity and you will be required to work onsite in Redmond, WA

 

[lexony]

Zelda Breath of the Wild

I don‘t like that reshade. It makes everything to dark and the color palette of the game gets lost.

 

[Deleted member 1324]

The difference is that I don’t expect Nintendo to drop the Switch for another 4-5 years. When they say they want to redefine / extend the traditional lifecycle, I suspect they mean it - first party games will continue with few exceptions until 2026.

I could be way off base, but that’s my expectation. When murmurings came about early last year about the new hardware, it wasn’t just Nate reporting his understanding that it’d be a revision. Imran also mentioned that we should “expect this thing to be primarily for FPS boosts and resolutions, not a Switch 2. Think more Pro than Super.” I’ve not really changed my expectations since.

I think the same. I don’t think Nintendo are comfortable moving up to the next level of development costs if they can avoid it until around 2025 when a true next generation Nintendo system will arrive. WiiU / Switch level visuals at 4k using DLSS is absolutely fine for them considering their stylised visuals and the fact Switch is still selling amazingly well in the mass market despite it being derided online by enthusiasts for low resolutions or poor framerates.

A collection of platforms ranging from $199 to $399 (Lite, Standard, OLED, DLSS) that simply play the games at higher resolutions the higher you go on the price point is a good long term strategy for them in terms of their business imo.

 

[Pokemaniac]

I think the same. I don’t think Nintendo are comfortable moving up to the next level of development costs if they can avoid it until around 2025 when a true next generation Nintendo system will arrive. WiiU / Switch level visuals at 4k using DLSS is absolutely fine for them considering their stylised visuals and the fact Switch is still selling amazingly well in the mass market despite it being derided online by enthusiasts for low resolutions or poor framerates.

A collection of platforms ranging from $199 to $399 (Lite, Standard, OLED, DLSS) that simply play the games at higher resolutions the higher you go on the price point is a good long term strategy for them in terms of their business imo.

By any reasonable metric, Dane is already going to be a "true next generation Nintendo system" by itself. I don't doubt that first party releases for the existing Switch will continue for some time after it releases, but I just can't see that sustaining at anywhere near 100% for more than a year or two unless adoption stalls pretty badly. The system will spend some time being functionally a revision, like what's happened with the new PlayStation and Xbox, but it will almost certainly start pulling away as time goes on just like those two systems are only just beginning to do.

The way I see things, the system can't remain the premium model forever. Eventually it will become the baseline.