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

[Alovon11]

I'm just looking forward to the next Mochizuki scoop. What's the deal with the dev kits that are out there?

That is the main thing I feel people are ignoring atm XD

Mochi does get proper info, the "Pro 2021" thing was mistaking the SWOLED for it which he did get the fact hardware for 2021 was coming true.

So the devkit comment likely very much is true.

 

[manfromthedam]

Mario Kart 8 DLC could signify a longer wait for Nintendo Switch 2

Nintendo expects to sell around 4 million Switch systems during the three-month period ending March 31.

venturebeat.com

"Nintendo plans to roll them out in six waves comprised of eight courses each. These waves will begin hitting Switch in March and continue through the end of 2023. And that timing is key because it’s probably the earliest that fans could expect to see a full successor to the Switch".

The dreams — or maybe they were hallucinations — of the Switch Pro are dead. Nintendo is almost certainly instead looking to launch a follow-up to the Switch — although that will likely be hardware that maintains the formfactor and momentum of what is now Nintendo’s best-selling home console of all time".

"And that probably puts a Switch 2 launch in March 2024 at the earliest and probably closer to November 2024".



I think this is very interesting because people are saying that Jeff is one of sources that was saying that "Pro" is coming in 2022.
this clearly indicates that not only he doesn't believe that any more, but its seems he think that "Pro" is Switch 2 now and that will not be out before 2024.


I would love to hear in Nate have new infos.

I'd go further than that. If they're supporting Mario Kart 8 until the end of 2023 they want to reap the benefits from that extended shelf life beyond just 2024. I'm with @Skittzo in expecting Switch to stick around for a decade. That'd mean we're exactly half-way in the Switch's lifespan, and they're re-launching the Switch this year with Sports, Kart, Xenoblade, Zelda etc..

This picture could include a Switch Revision really well IMO.

 

[Hermii]

If Dane still plays MK, just in 4k I dont know how support for MK8 means anything really. It just means the original Switch still has years of software support left in it, which most of us assumed it would anyway. Regardless of new hardware launching or not.

 

[Alovon11]

I'd go further than that. If they're supporting Mario Kart 8 until the end of 2023 they want to reap the benefits from that extended shelf life beyond just 2024. I'm with @Skittzo in expecting Switch to stick around for a decade. That'd mean we're exactly half-way in the Switch's lifespan, and they're re-launching the Switch this year with Sports, Kart, Xenoblade, Zelda etc..

This picture could include a Switch Revision really well IMO.

like I've said before 2022 is very much seeming to be 2017-2.

The only things missing now pretty much are a Confirmed 3D Mario and a console launch XD

 

[PedroNavajas]

How did kopite7kimi got the information that dane is based on orin? It is a fair estimate, but we are under the assumption that she has insider information. What kind of information it would be?

 

[Alovon11]

How did kopite7kimi got the information that dane is based on orin? It is a fair estimate, but we are under the assumption that she has insider information. What kind of information it would be?

Kopite is the NVIIDA Leaker.

They leaked all of Ampere, more or less Orin, and a lot of other things with a very high accuracy rate.

 

[Simba1]

I'd go further than that. If they're supporting Mario Kart 8 until the end of 2023 they want to reap the benefits from that extended shelf life beyond just 2024. I'm with @Skittzo in expecting Switch to stick around for a decade. That'd mean we're exactly half-way in the Switch's lifespan, and they're re-launching the Switch this year with Sports, Kart, Xenoblade, Zelda etc..

This picture could include a Switch Revision really well IMO.

Maybe, but main point of article that there is no "Pro" and seems that Jeff thinks that we only have Switch 2 and normal revisions similar or like Switch Lite or Switch OLED (he even mention Switch Lite OLED like possible revision).

Personally, I always found strange to release more expansive OLED model if they intend to release stronger Switch model just year later,
dont make too much sense.

 

[PedroNavajas]

Kopite is the NVIIDA Leaker.

They leaked all of Ampere, more or less Orin, and a lot of other things with a very high accuracy rate.

So, the leak is from the chip design teams? Would it be possible for the leaks to come from the manufacturing or driver development side?

 

[Ozzie]

So...I'm kind of confused. Do we expect any form of revision anytime soon?

 

[Herb Alpert]

So...I'm kind of confused. Do we expect any form of revision anytime soon?

With the recent developments, my personal take is that no, they'll keep the big power boost for something they'll sell as a next gen switch, probably back compatible, and they won't release it until March 2024.
Just my opinion though.

 

[IamPeacock]

If I have to deal with blurry Switch games and bad framerates for another 3~ years I‘m going to pull my hair out, which will be extra painful since I shave my head.

I own a Switch, buy the games, but emulate most of them on PC. 4K and 60fps of Nintendo sweetness.

I'm in the same boat as you. I can't deal with those sub 30 fps anymore.

 

[Skittzo]

Jeff's conclusion there is extremely counterintuitive. How is longer support for the base Switch a sign that they're working on a Switch 2 rather than a pro? It should signal the exact opposite.

The fact of that matter is that 11 developers had 4k dev kits last year. That's not something that could possibly have been fabricated. If we're looking at no earlier than 2024 for a new gen then what the hell were those 4k dev kits?

It just doesn't add up.

 

[PedroNavajas]

Jeff's conclusion there is extremely counterintuitive. How is longer support for the base Switch a sign that they're working on a Switch 2 rather than a pro? It should signal the exact opposite.

The fact of that matter is that 11 developers had 4k dev kits last year. That's not something that could possibly have been fabricated. If we're looking at no earlier than 2024 for a new gen then what the hell were those 4k dev kits?

It just doesn't add up.

A change of plans, due to severe supply chain disruptions, is possible IMHO.

 

[Skittzo]

A change of plans, due to severe supply chain disruptions, is possible IMHO.

If there was a change of plans about any hardware that's not something these new software announcements would be able to reflect, which is more what I'm getting at. I'm basically wondering how this direct convinced Jeff that a new revision couldn't be coming, rather than convince him of the opposite.

 

[Thraktor]

Thanks for all this.

Is the a scenario where you think we get an 8SM Dane?

Is there a scenario where you think we don’t get a Dane/Orin SoC?

I think there are two scenarios where we'd gen an 8SM Dane. The first is where Nintendo is both far more ambitious with performance than I'd expect, and also far less concerned about battery life than I'd expect. The second is where they only run the full 8 SMs in docked mode, and disable several of them in portable mode. This is technically possible, but means they're paying for hardware that isn't in use a lot of the time, and they either clock the 8SMs very low in docked mode (defeating the purpose somewhat) or have a big performance differential between portable and docked modes. Neither of these scenarios seem very likely to me.

In theory we wouldn't get a Dane device if they just scrap the entire project, but that seems pretty unlikely. Unless they have another SoC design waiting in the wings, that would mean any new hardware (pro or successor) would be pushed back to 2025 at the earliest.

Not thraktor but will probably answer the latter one, scenario in which they don’t get an Orin/Dane SoC would be one in which Nintendo pulls all plans on it and ultimately shelves the idea of that in a similar case to what they did with the GameBoy successor that eventually came out as the GameBoyAdvance and had the GameBoy Color as the stopgap, though they wouldn’t have that as the stopgap in this case. They’d perhaps have the Switch LITE with an OLED display.

Reason to not use it (Dane) would be if they were unable to secure sufficient capacity for their needs. This extends beyond the SoC itself, such as components that pertain to the Memory, the Motherboard, the storage, the sensors, etc. if one of the crucial components is in short supply, it makes it very difficult to produce in mass quantities which is the EoS that Nintendo operates in, shifting 20M a year.

In the scenario they don’t use it, NVidia would probably rebrand it as part of the ORIN family for unique small case use dedicated for AI and probably disable some unnecessary feature.


This would of course sour the relationship between Nintendo and NVidia if it’s awaiting a tape out meaning it’s core design is complete, but some aspects of it aren’t up to Nintendo’s requirements…. And they decide to just not use it. Would sour relationship with Samsung as well as they would also be involved with this chip considering it’s their fab

The one thing I'd point out for this is that the crucial components of a Dane model would be sourced separately from the other Switch models, so in a supply constrained world if you want to produce as many Switches as possible (base or pro or successor or whatever), producing a Dane-based model alongside the existing models would allow them to increase overall production. Dane itself is manufactured on a different process at a different foundry to Mariko, it would be using LPDDR5 RAM instead of LPDDR4X, probably 128GB of flash instead of 32GB/64GB (and possibly UFS instead of eMMC), and quite a few of the smaller ICs would be different (eg USB-C chip, possibly WiFi/bluetooth chip, etc). Screen would still probably be OLED sourced from Samsung, but OLED is oversupplied at the moment, so that should't be a problem.

I also don't think it would sour the relationship between Nintendo and Nvidia that much. Whatever contract they have would almost certainly have clauses that make sure Nvidia aren't out-of-pocket if Nintendo cancel the project, and any cancellation would almost certainly be followed (or even preceded) by the start of a new project for an SoC for Nintendo's next console.

I have to push back a little on this and I think the way desktop Ampere is organized we don't have the entire picture...
Most of the assumptions we are making is going by kopite7kimi saying Dane is based off of Orin, if the next Switch deviates heavily from Orin's established design, it's not really based on that SoC anymore.

The Tegra X1, X2 and Xavier were all the largest forms of those SoC's, but in this case Nintendo would be coming to the table with an Nvidia design that's starting much larger than what they currently can realistically use at the moment(on 8nm that is).
Any cut down versions of Nvidia's Tegra SoC's are still the same footprint as the full chip was our main argument.

I think using some of this logic one might assume Nintendo could of halved the cores in the TX1 to achieve better TDP on 20nm in portable mode and to achieve the docked performance with higher clocks. Instead they chose to use the full TX1 design and underclock the SoC to meet whatever metrics they wanted, even if they are on the conservative side.

I don't think it's ever been a concern on whether the Switch is profitable enough for Nintendo and Nvidia to warrant making something completely customized. Just reading through their recent investors notes just proves they are constantly concerned with the ongoing chip shortages and how this effects future profits. We can't fully expect Nintendo to adopt things like OLED displays, increased RAM, fast UFS storage, premium build materials on top of an extremely custom SoC(vs adapting from something Nvidia already has planned) and for this device not to be priced at $450-500...

Dane being based on Orin only means whatever Nvidia and Nintendo need it to mean. There's a certain amount of shared requirements to the chips (ie they're SoCs releasing around the same time, they both require high ML inference performance, etc.) so Nvidia are going to share technology between them as much as makes sense, but they're not going to restrict either chip for arbitrary reasons. For example, both chips require high ML inference performance (Orin for automotive use-cases, Dane for DLSS), but neither needs training performance, so we've got a tensor core setup which operates at double the rate compared to desktop Ampere, but only supports lower precision modes, dropping the TF32 support that's only really used for training.

Nvidia have never released both big and small SoC dies in the same generation before, so trying to extrapolate from a sample size of zero isn't much help to us. The fact that previous "cut down" versions of Nvidia SoCs are the same footprint isn't relevant, as they're literally just binned versions of the same die, so of course they're the same size. Dane is clearly a separate die from Orin (they're not putting a 400mm2+ die in a device like the Switch), so what Nvidia have done or not done with binned SoC dies in the past doesn't really have any bearing.

IMO (and I believe I deviate from the mainstream with this) I don't believe they'll want their next "true" platform to simply be a Switch 2. They'll want it to revolve around a new concept, like they have been doing ever since the DS came out. I have no idea what that could be, but I don't see it as something that will need a big jump in processing capability to differentiate itself.

Yeah, this is what I'm leaning towards now. A "pro" Switch (obviously not called Switch Pro, but whatever it's called) with Dane late this year/early next year, and then a new console somewhere around late 2025. I don't think it will be a Switch 2, and will likely at least have some kind of new concept or features to differentiate it from the Switch.

One advantage of the Switch is that it actually gives them quite a bit of freedom to build a new console and retain a very good level of backwards compatibility. They would need an ARM SoC (probably from Nvidia), but they're quite free in form-factor, so long as there is some support for traditional control schemes.

 

[ITSMILNER]

I fail to see how releasing a Switch lite OLED makes any sense, a lot of people will be interested in Switch Sports this year which a lite would not support. Would many current lite owners upgrade to an OLED version?

 

[LoneRanger]

I still believe in our french poster who disappeared some months ago, he said IIRC:

• Xbox One S raw GPU performance (~1,4 TFlops, x3,5 OG Switch on docked)
• DLSS in both configurations, confirming Tensor Cores
• CPU will be A78C, don’t said number of cores

Add 8GB RAM with enough bandwidth and you have a new model enough to run every PS4 game without effort and some downports of PS5 games.

Its 100% confirm COVID has delayed all Nintendo plans.

 

[PedroNavajas]

I think there are two scenarios where we'd gen an 8SM Dane. The first is where Nintendo is both far more ambitious with performance than I'd expect, and also far less concerned about battery life than I'd expect. The second is where they only run the full 8 SMs in docked mode, and disable several of them in portable mode. This is technically possible, but means they're paying for hardware that isn't in use a lot of the time, and they either clock the 8SMs very low in docked mode (defeating the purpose somewhat) or have a big performance differential between portable and docked modes. Neither of these scenarios seem very likely to me.

What would you consider "very low clock"? Some could ague that the <1/3 of X1 on Switch is already a very aggressive downclock. Do you believe that an 8SM chip would need to be significantly lower than that?

Now, I think that if Switch 2 is really aiming for a 4k output, having a large differential with the handheld mode might be warranted, even with DLSS. However, I think that memory bandwidth is a bigger issue if aiming for a native 1440p output.

 

[Ozzie]

With the recent developments, my personal take is that no, they'll keep the big power boost for something they'll sell as a next gen switch, probably back compatible, and they won't release it until March 2024.
Just my opinion though.

I always thought the Switch would get an upgrade revision similar to New 3ds. So in other words, we won't get any upgrade until 2 years?

 

[Dakhil]

Switch Pro could still happen (although I doubt that personally), but the Switch 2 isn't happening for a long time.

Who says the DLSS model* has to strictly be a mid-gen refresh or a "true" successor? Why couldn't the DLSS model* be an iterative successor in the sense where developers are completely free to release exclusive games, but the cross-gen period's for an iterative successor is likely much longer than for a "true" successor? (I imagine the PlayStation 5 and/or Xbox Series X|S is the exception rather than the norm as far as the cross-gen period's concerned due to the global chip shortage making hardware demand outstrip supply.)

Nintendo's emphasis on maintaining and expanding the 100 million annual playing users for the next hardware platform when answering question #4 of the Q&A seems to strongly imply the DLSS model* is an iterative successor.

September 30 along side with xenoblade 3. One only can hope

I imagine that's contingent on Dane being taped out by the end of March 2022 at the absolute latest.

 

[NineTailSage]

I think there are two scenarios where we'd gen an 8SM Dane. The first is where Nintendo is both far more ambitious with performance than I'd expect, and also far less concerned about battery life than I'd expect. The second is where they only run the full 8 SMs in docked mode, and disable several of them in portable mode. This is technically possible, but means they're paying for hardware that isn't in use a lot of the time, and they either clock the 8SMs very low in docked mode (defeating the purpose somewhat) or have a big performance differential between portable and docked modes. Neither of these scenarios seem very likely to me.

In theory we wouldn't get a Dane device if they just scrap the entire project, but that seems pretty unlikely. Unless they have another SoC design waiting in the wings, that would mean any new hardware (pro or successor) would be pushed back to 2025 at the earliest.



The one thing I'd point out for this is that the crucial components of a Dane model would be sourced separately from the other Switch models, so in a supply constrained world if you want to produce as many Switches as possible (base or pro or successor or whatever), producing a Dane-based model alongside the existing models would allow them to increase overall production. Dane itself is manufactured on a different process at a different foundry to Mariko, it would be using LPDDR5 RAM instead of LPDDR4X, probably 128GB of flash instead of 32GB/64GB (and possibly UFS instead of eMMC), and quite a few of the smaller ICs would be different (eg USB-C chip, possibly WiFi/bluetooth chip, etc). Screen would still probably be OLED sourced from Samsung, but OLED is oversupplied at the moment, so that should't be a problem.

I also don't think it would sour the relationship between Nintendo and Nvidia that much. Whatever contract they have would almost certainly have clauses that make sure Nvidia aren't out-of-pocket if Nintendo cancel the project, and any cancellation would almost certainly be followed (or even preceded) by the start of a new project for an SoC for Nintendo's next console.



Dane being based on Orin only means whatever Nvidia and Nintendo need it to mean. There's a certain amount of shared requirements to the chips (ie they're SoCs releasing around the same time, they both require high ML inference performance, etc.) so Nvidia are going to share technology between them as much as makes sense, but they're not going to restrict either chip for arbitrary reasons. For example, both chips require high ML inference performance (Orin for automotive use-cases, Dane for DLSS), but neither needs training performance, so we've got a tensor core setup which operates at double the rate compared to desktop Ampere, but only supports lower precision modes, dropping the TF32 support that's only really used for training.

Nvidia have never released both big and small SoC dies in the same generation before, so trying to extrapolate from a sample size of zero isn't much help to us. The fact that previous "cut down" versions of Nvidia SoCs are the same footprint isn't relevant, as they're literally just binned versions of the same die, so of course they're the same size. Dane is clearly a separate die from Orin (they're not putting a 400mm2+ die in a device like the Switch), so what Nvidia have done or not done with binned SoC dies in the past doesn't really have any bearing.

There's a lot of unknowns of just how scalable Ampere is and the performance metrics there in, but we should all be mindful that if this device is to have DLSS capabilities and possibly some RT functionality as well. Then a baseline has to be established of where that starts at realistically, because anything else below that line really doesn't make much sense to discuss really...

 

[Skittzo]

I fail to see how releasing a Switch lite OLED makes any sense, a lot of people will be interested in Switch Sports this year which a lite would not support. Would many current lite owners upgrade to an OLED version?

I'm with you, I see no business appeal to that product whatsoever. I can see it being popular with a niche audience- people who don't want docking functionality but also don't mind paying a bit extra for a nicer screen- but again that's gotta be quite a niche audience.

The OLED hybrid is already positioned as a premium handheld device, plus it has the ability to dock giving it added value over the regular Lite. It's already the model designed to get Lite owners to upgrade.

 

[Simba1]

I always thought the Switch would get an upgrade revision similar to New 3ds. So in other words, we won't get any upgrade until 2 years?

Me too, 1st I thought V2 model will be that, than that OLED will be type of New 3DS upgrade,
at this point I don't see at all upgraded model just full next gen (Switch 2).

 

[Z0m3le]

I think if plans changed for Dane, these games that were targeting it for release, would likely also see a delay, the direct this week has multiple months with multiple first party releases, which leads me to believe they have plenty of software in the pipeline where a delay just wouldn't make sense with this release schedule.

Posted this in beyond3D:

For backwards compatibility, couldn't they just use half the shaders at twice the clock? Tegra X1 is 2B transistors, but the GPU is much smaller than that, for BC, if they can use half the shader count, it could be as small as 300M-500M transistors, when considering Tegra X1 is made on 20nm with a ~17M Transistors per mm, Samsung's 8nm is ~3x as dense, meaning they could fit ~6B transistors in the same space, even losing out on 500M transistors would still net them 3 times+ the performance jump thanks to "Ampere+" architecture (Orin uses much more internal cache and "Dane" is rumored to be a much smaller sister die). As for DLSS, MX570 will support it, and the even lower end Nvidia A2 with just 1280 shaders and 40 tensor cores should be capable of DLSS, so I'd expect Dane to have around half the shaders of MX570 to fit a similar power envelope of the original Switch, giving it 1024 shaders, 32 tensor cores and possibly 8 RT cores (again half of MX570).

Even if the clock was 1GHz when docked, the GPU would push 2TFLOPs Ampere"+" and offer DLSS, when in portable mode, I'd expect them to use 600MHz or so, pushing 1.2TFLOPs Ampere, which should match up well with PS4 even without DLSS, though considering the speed at which the tensor cores could process 720p, DLSS could be a very realistic option for portable mode, which would see it exceed Steam Deck in visual performance, as it could render the games at 480p-540p and use DLSS to reach 720p, the benefit here is AA is applied via DLSS and the smaller screen would help hide any imperfections of the image.

A78C would allow a huge jump in CPU performance over the current Switch's A57 cores at 1GHz, I'd expect a ~50% increase in clock and IPC is about 3 times more performant over A57, and A78C can be an 8 core cluster. That could offer as much as 10 times the performance of the CPU found in Switch (for gaming, as I'm counting 3 cores vs 7 cores here).

Lastly, Nintendo currently uses LPDDR4X in their Switch, but I'd expect them to use either LPDDR5 or LPDDR5X, which at 128bits is 102GB/s or ~133GB/s. Switch has 25.6GB/s, so the upgrade is again huge. This is based on GPU architecture 2 years old right now, and Dane which if real, started in late 2019 after Mariko and before Orin's announcement in December 2019, should be ready this year or next IMO.

This is mostly what I've come to after the last few years of speculation, I also don't think the die needs to be limited to ~120mm^2, as Wii U's GPU alone was ~160mm^2 and was built on a more expensive MCM solution. We should actually know if Dane is coming this calendar year around the end of April when Nintendo releases their earnings and announce projected hardware sales, they are expecting to sell 23M units this FY, but if next FY there is 25M+ well that would mean they have entirely different units with different chips in them. I fully expect Nintendo to treat a new model like this as a "4K" revision, there is also a lot of talk from Nintendo's leaders about pursuing cloud technology for game streaming, which could mean they are building a "Dane" powered network to run their first party titles that target Dane onto the current Switch units, I don't expect Nintendo to launch exclusive Dane software in a capacity exceeding New 3DS had, at least not for a couple years, when they could have their cloud ready. Ultimately they could sell you a Switch "2" game on your Switch and when you play it, it launches a cloud version of the game, but when you play it on a Switch 2, it runs locally, wouldn't matter if you had a digital or physical version of the game either, sort of the best hope for an everlasting platform and moving away from hardware that Iwata talked about originally, and solves the problem of maintaining the 100M+ owners they currently have, that the president of Nintendo just talked about in their Q&A.

 

[ITSMILNER]

I'm with you, I see no business appeal to that product whatsoever. I can see it being popular with a niche audience- people who don't want docking functionality but also don't mind paying a bit extra for a nicer screen- but again that's gotta be quite a niche audience.

The OLED hybrid is already positioned as a premium handheld device, plus it has the ability to dock giving it added value over the regular Lite. It's already the model designed to get Lite owners to upgrade.

Exactly, you have your standard Switch, your Switch Lite then the OLED, that's a model for every potential customer (Not including a home console only model but then that's not what the Switch is about imo)

Another thing that I thought about, we are 6 years into Switch coming up and we are just getting rumblings of games like Batman Arkham series and Mass Effect coming to the console. To me (and I may be reading too much into it) but the timing of them plus the rumours of DLSS devkits being out there make me think these projects were only started when the new devkits for a more powerful system became available. The games could be developed on the new kits and a scaled back version for the OG model could be released and then patched/unlocked for Dane systems when available.

Wouldn't Nintendo piss a lot of their partners off if they don't actually release a console to match the kits they have passed to devs to work on?

 

[Skittzo]

I think if plans changed for Dane, these games that were targeting it for release, would likely also see a delay, the direct this week has multiple months with multiple first party releases, which leads me to believe they have plenty of software in the pipeline where a delay just wouldn't make sense with this release schedule.

Yeah I still have seen no evidence or compelling argument that suggests Dane is delayed or cancelled. Of course it could happen, but honestly that direct convinced me more that Dane is on track than anything else.

Wouldn't Nintendo piss a lot of their partners off if they don't actually release a console to match the kits they have passed to devs to work on?

Yeah, definitely. Not that pissing them off would be the end of the world or something, but yeah they'd want to avoid doing that if possible.

 

[Z0m3le]

Exactly, you have your standard Switch, your Switch Lite then the OLED, that's a model for every potential customer (Not including a home console only model but then that's not what the Switch is about imo)

Another thing that I thought about, we are 6 years into Switch coming up and we are just getting rumblings of games like Batman Arkham series and Mass Effect coming to the console. To me (and I may be reading too much into it) but the timing of them plus the rumours of DLSS devkits being out there make me think these projects were only started when the new devkits for a more powerful system became available. The games could be developed on the new kits and a scaled back version for the OG model could be released and then patched/unlocked for Dane systems when available.

Wouldn't Nintendo piss a lot of their partners off if they don't actually release a console to match the kits they have passed to devs to work on?

I actually think this has more to do with Switch having sold 100M+ units, we are going to see a lot more random ports and new games going forward because Nintendo is flagging to publishers that Switch isn't going away and has ~100M active gamers wanting more games.

 

[Vash_the_Stampede]

I think if plans changed for Dane, these games that were targeting it for release, would likely also see a delay, the direct this week has multiple months with multiple first party releases, which leads me to believe they have plenty of software in the pipeline where a delay just wouldn't make sense with this release schedule.

Posted this in beyond3D:

For backwards compatibility, couldn't they just use half the shaders at twice the clock? Tegra X1 is 2B transistors, but the GPU is much smaller than that, for BC, if they can use half the shader count, it could be as small as 300M-500M transistors, when considering Tegra X1 is made on 20nm with a ~17M Transistors per mm, Samsung's 8nm is ~3x as dense, meaning they could fit ~6B transistors in the same space, even losing out on 500M transistors would still net them 3 times+ the performance jump thanks to "Ampere+" architecture (Orin uses much more internal cache and "Dane" is rumored to be a much smaller sister die). As for DLSS, MX570 will support it, and the even lower end Nvidia A2 with just 1280 shaders and 40 tensor cores should be capable of DLSS, so I'd expect Dane to have around half the shaders of MX570 to fit a similar power envelope of the original Switch, giving it 1024 shaders, 32 tensor cores and possibly 8 RT cores (again half of MX570).

Even if the clock was 1GHz when docked, the GPU would push 2TFLOPs Ampere"+" and offer DLSS, when in portable mode, I'd expect them to use 600MHz or so, pushing 1.2TFLOPs Ampere, which should match up well with PS4 even without DLSS, though considering the speed at which the tensor cores could process 720p, DLSS could be a very realistic option for portable mode, which would see it exceed Steam Deck in visual performance, as it could render the games at 480p-540p and use DLSS to reach 720p, the benefit here is AA is applied via DLSS and the smaller screen would help hide any imperfections of the image.

A78C would allow a huge jump in CPU performance over the current Switch's A57 cores at 1GHz, I'd expect a ~50% increase in clock and IPC is about 3 times more performant over A57, and A78C can be an 8 core cluster. That could offer as much as 10 times the performance of the CPU found in Switch (for gaming, as I'm counting 3 cores vs 7 cores here).

Lastly, Nintendo currently uses LPDDR4X in their Switch, but I'd expect them to use either LPDDR5 or LPDDR5X, which at 128bits is 102GB/s or ~133GB/s. Switch has 25.6GB/s, so the upgrade is again huge. This is based on GPU architecture 2 years old right now, and Dane which if real, started in late 2019 after Mariko and before Orin's announcement in December 2019, should be ready this year or next IMO.

This is mostly what I've come to after the last few years of speculation, I also don't think the die needs to be limited to ~120mm^2, as Wii U's GPU alone was ~160mm^2 and was built on a more expensive MCM solution. We should actually know if Dane is coming this calendar year around the end of April when Nintendo releases their earnings and announce projected hardware sales, they are expecting to sell 23M units this FY, but if next FY there is 25M+ well that would mean they have entirely different units with different chips in them. I fully expect Nintendo to treat a new model like this as a "4K" revision, there is also a lot of talk from Nintendo's leaders about pursuing cloud technology for game streaming, which could mean they are building a "Dane" powered network to run their first party titles that target Dane onto the current Switch units, I don't expect Nintendo to launch exclusive Dane software in a capacity exceeding New 3DS had, at least not for a couple years, when they could have their cloud ready. Ultimately they could sell you a Switch "2" game on your Switch and when you play it, it launches a cloud version of the game, but when you play it on a Switch 2, it runs locally, wouldn't matter if you had a digital or physical version of the game either, sort of the best hope for an everlasting platform and moving away from hardware that Iwata talked about originally, and solves the problem of maintaining the 100M+ owners they currently have, that the president of Nintendo just talked about in their Q&A.

Hey thanks for this - big fan of your posts from ResetEra. A few clarifying points

1. What is your guess on timing, this year or next?

2. Are you in the Pro camp or successor camp?

3. Said differently, are you in the @Thraktor camp of a 4-6SM and 4-6 A78s unit that’s a nice bump up in specs or in the @Alovon11 camp of 8SMs and 8 A78, which could be a Series S competitor in docked mode.

Thanks!

 

[Z0m3le]

Hey thanks for this - big fan of your posts from ResetEra. A few clarifying points

1. What is your guess on timing, this year or next?

2. Are you in the Pro camp or successor camp?

3. Said differently, are you in the @Thraktor camp of a 4-6SM and 4-6 A78s unit that’s a nice bump up in specs or in the @Alovon11 camp of 8SMs and 8 A78, which could be a Series S competitor in docked mode.

Thanks!

I think because of the Mario Movie's "delay" to the holidays, that a mainline Mario, that I have a gut feeling exists, is what they would want to release this holiday instead of Zelda. Remember they said they were 'targeting a 2022 release' for botw2, and I think if you are moving Zelda to March next year, that is when you release Dane.

I believe it will be a Gameboy Color type revision, specifically, I think they will call it "Switch 4K model" and I believe they are pursuing a cloud gaming service that I went over in the post you quoted.

I believe it will be 6-8 A78 cores and 6-8SMs, I think ultimately the count doesn't matter, we know that Orin has a 15W setting that offers ~36TOPs, because Dane uses the same architecture, we know it should be capable of very similar performance at that power consumption, which leaves me with the range I talk about in that post again.

 

[PedroNavajas]

Looks like I was wrong.

 

[Skittzo]

I think because of the Mario Movie's "delay" to the holidays, that a mainline Mario, that I have a gut feeling exists, is what they would want to release this holiday instead of Zelda. Remember they said they were 'targeting a 2022 release' for botw2, and I think if you are moving Zelda to March next year, that is when you release Dane.

I think they would've announced a Mario game at this Direct if there was one coming this holiday. The fact that they left the holiday period conspicuously open along with the fact that they didn't announce any other Zelda games for this year (like WWHD/TPHD) heavily, heavily suggests they're confident in BOTW2 being the holiday game this year.

Them not mentioning Zelda at all at the last Direct sets up everybody's expectations to see it at their summer Direct, it would be really poor marketing to announce a delay there.

I believe it will be 6-8 A78 cores and 6-8SMs, I think ultimately the count doesn't matter, we know that Orin has a 15W setting that offers ~36TOPs, because Dane uses the same architecture, we know it should be capable of very similar performance at that power consumption, which leaves me with the range I talk about in that post again.

Isn't that 15W setting for the SoC, not the whole device itself? Dane needs to be closer to like, 8-10W I thought. What does Erista top out at in the Switch?

 

[Dakhil]

• Nintendo used X1 as it is, so if Nintendo doesn't want to spend much on custom design, 8SM seems like the cheapest design.

I don't think the Tegra X1 is necessarily a good indicator of how Nintendo and Nvidia could approach designing Dane since I imagine going from the Wii U to the Nintendo Switch is a very different situation from going from the Nintendo Switch to the DLSS model*.

What does Erista top out at in the Switch?

The Nintendo Switch (2017) can exceed 15 W when turned on and discharged regardless of which mode (TV mode or handheld mode).

 

[Z0m3le]

The two sites has merits and is contradictory in surprising ways.

8 SM Arguments:

• Dane was leaked to be based on Orin and Orin is 8SM.
• Nintendo used X1 as it is, so if Nintendo doesn't want to spend much on custom design, 8SM seems like the cheapest design.
• If Nintendo also cheapened out, they can share the fab work already done for Orin.
• Nintendo can just downclock or not use all the SM if the power consumption is too much.

4-6 SM:

• 8 SM is too power hungry.
• 8 SM is too large
• Downclocking/disabling SM is inefficient, and Nintendo would be stuck producing larger silicons, which is more expensive over time.

I think all those arguments are true. What's more, thinking that Nintendo is a cheap-ass doesn't solve the issue. If Nintendo cheapened on design and other fab work, then 8SM it is. If Nintendo want's to release a device that will be cheap to produce in the future, 4SM seems like the right choice.

Is not an easy answer.

Orin is 16SM (2048 shaders divided by 128 shaders per SM is 16).
Dane is custom, we are talking about a custom chip and 8SM should still fix in a die similar in size to Tegra X1 on 8nm.
Fab work for Orin is done, but Dane's fab work continues.
Power consumption is better with more SM on a smaller clock than less SM on a higher clock, power consumption would lend itself to a bigger GPU than a "faster" one.

As for the rest, as I mention above Orin has a 100TOPs mode that is 45 watts and a 36TOPs mode that is 15 watts, that means half the GPU with a lower clock, should fit the power consumption we see with Orin.

 

[Skittzo]

The Nintendo Switch (2017) can exceed 15 W when turned on and discharged regardless of which mode (TV mode or handheld mode).

I mean Erista the SoC specifically, not the entire unit. That was my point, 15W for the SoC alone seems a bit too high based on what we know about the base Switch.

 

[Vash_the_Stampede]

I think because of the Mario Movie's "delay" to the holidays, that a mainline Mario, that I have a gut feeling exists, is what they would want to release this holiday instead of Zelda. Remember they said they were 'targeting a 2022 release' for botw2, and I think if you are moving Zelda to March next year, that is when you release Dane.

I believe it will be a Gameboy Color type revision, specifically, I think they will call it "Switch 4K model" and I believe they are pursuing a cloud gaming service that I went over in the post you quoted.

I believe it will be 6-8 A78 cores and 6-8SMs, I think ultimately the count doesn't matter, we know that Orin has a 15W setting that offers ~36TOPs, because Dane uses the same architecture, we know it should be capable of very similar performance at that power consumption, which leaves me with the range I talk about in that post again.

Last one: what kinda performance do you think we can get on 6-8SMs and 6-8 A78.

It seems clear that’s CPU would be ~10x vs. Mariko with 6 gaming CPUs; how does that compare with Series S?

I’m less sure on the pre/post DLSS math for 6-8 SMs at your assumed 15w power envelop.

I’m hopeful they prioritize performance this time around bc it’s easier to get more battery life with better batteries (3%-5% per year improvement) or future revisions.

 

[PedroNavajas]

Orin is 16SM (2048 shaders divided by 128 shaders per SM is 16).

You are right. I think I got mixed with the textures units and previous architectures.

 

[Skittzo]

Last one: what kinda performance do you think we can get on 6-8SMs and 6-8 A78.

It seems clear that’s CPU would be ~10x vs. Mariko with 6 gaming CPUs; how does that compare with Series S?

I’m less sure on the pre/post DLSS math for 6-8 SMs at your assumed 15w power envelop.

I’m hopeful they prioritize performance this time around bc it’s easier to get more battery life with better batteries (3%-5% per year improvement) or future revisions.

10x for the CPU seems way too high for 6 cores based on what I remember of the IPC discussion we had months/year? ago. I think the realistic range for CPU improvements tops out at 9x, and is likely going to be lower than that since they won't use extravagant clock speeds.

And IIRC it won't quite reach the area of Series S, but it'll definitely be a fair bit closer to that level than the Switch was to the XB1/PS4 CPUs when it launched.


I think it's probably a good idea to assume they'll go conservative on most of the clocks specifically to ensure better yields and therefore supply/availability, especially in the current climate.

 

[Z0m3le]

I think they would've announced a Mario game at this Direct if there was one coming this holiday. The fact that they left the holiday period conspicuously open along with the fact that they didn't announce any other Zelda games for this year (like WWHD/TPHD) heavily, heavily suggests they're confident in BOTW2 being the holiday game this year.

Them not mentioning Zelda at all at the last Direct sets up everybody's expectations to see it at their summer Direct, it would be really poor marketing to announce a delay there.

Isn't that 15W setting for the SoC, not the whole device itself? Dane needs to be closer to like, 8-10W I thought. What does Erista top out at in the Switch?

Just a gut feeling, I think E3 will have the Mario movie trailer, so announcing a Mario game could happen then, 3D World was announced at E3 before its holiday release and DKTF was delayed at the same time. Zelda will sell as well in any month, but Mario would benefit from the holiday release as we saw with odyssey overtaking MK8 throughout 2018 until fall that year.

As for the power consumption, I'm not saying Dane's SoC would consume 15 watts, I'm saying that a SoC with half the GPU at a lower clock is what fits that range, Nintendo would lower the clock further from that point, just as we see in TX1's 10watt/15watt design.

 

[Z0m3le]

10x for the CPU seems way too high for 6 cores based on what I remember of the IPC discussion we had months/year? ago. I think the realistic range for CPU improvements tops out at 9x, and is likely going to be lower than that since they won't use extravagant clock speeds.

And IIRC it won't quite reach the area of Series S, but it'll definitely be a fair bit closer to that level than the Switch was to the XB1/PS4 CPUs when it launched.


I think it's probably a good idea to assume they'll go conservative on most of the clocks specifically to ensure better yields and therefore supply/availability, especially in the current climate.

It's 7 cores for games and 1 for OS.

You are right. I think I got mixed with the textures units and previous architectures.

Yep, just trying to get everyone knowledge'd up

 

[ReddDreadtheLead]

Everything we've discussed up until now is all speculation and kopite7kimi is the only figure of authority to pull from on Dane.
Even them removing all of the automotive specific hardware is just speculation, for all we know Nintendo were playing around with Xavier seeing all of the things they can do in development. I've made the argument for the possibilities of that dedicated hardware being used in a Head-mounted VR solution in the future for head and controller tracking via cameras and sensors.

Correct that it is just speculation, my point is that if we are discussing the possibility of removing something like the DLA we can’t exclude the possibility of removing some SMS from the GPC. The conversation of the speculation would be a contradiction where we choose what they can and can’t do and apply it to what they can’t and can do. If removing the PVA is on the table or swapping CPU cores is on the table, so it removing SMs off the thing.

The one thing I'd point out for this is that the crucial components of a Dane model would be sourced separately from the other Switch models, so in a supply constrained world if you want to produce as many Switches as possible (base or pro or successor or whatever), producing a Dane-based model alongside the existing models would allow them to increase overall production. Dane itself is manufactured on a different process at a different foundry to Mariko, it would be using LPDDR5 RAM instead of LPDDR4X, probably 128GB of flash instead of 32GB/64GB (and possibly UFS instead of eMMC), and quite a few of the smaller ICs would be different (eg USB-C chip, possibly WiFi/bluetooth chip, etc). Screen would still probably be OLED sourced from Samsung, but OLED is oversupplied at the moment, so that should't be a problem.

I also don't think it would sour the relationship between Nintendo and Nvidia that much. Whatever contract they have would almost certainly have clauses that make sure Nvidia aren't out-of-pocket if Nintendo cancel the project, and any cancellation would almost certainly be followed (or even preceded) by the start of a new project for an SoC for Nintendo's next console.

My point with respect to “situation of them not using it” is just looking at the Dane unit in a vacuum. What affects the Mariko unit wouldn’t really have a direct effect on the Dane unit except for possibly the screen. If they couldn’t procure components that aren’t directly related to the SoC then it becomes difficult to produce at the economy of scale they sit at.

As for souring, of course they would take the payment as Nintendo would pay for it, but I don’t think the decision to just cull it would be a cheap one at all even if they move on to the next plans.

 

[PedroNavajas]

It's 7 cores for games and 1 for OS.

You see the 4 + 2 configuration as unlikely?

 

[Skittzo]

Just a gut feeling, I think E3 will have the Mario movie trailer, so announcing a Mario game could happen then, 3D World was announced at E3 before its holiday release and DKTF was delayed at the same time. Zelda will sell as well in any month, but Mario would benefit from the holiday release as we saw with odyssey overtaking MK8 throughout 2018 until fall that year.

Well for starters it's looking like E3 isn't a thing this year. Also I don't think they'll debut the Mario movie trailer at their summer direct, it's probably going to be a Mar10 day thing. If they were smart they would've booked a Superbowl spot for it too but it looks like that didn't happen.

I dunno, I just don't see them doing a 5 month marketing cycle for a brand new 3D Mario game when they're not in panic mode. And again the expectations they set for Zelda at this direct wouldn't be a good idea if they planned to delay it at their next direct.

As for the power consumption, I'm not saying Dane's SoC would consume 15 watts, I'm saying that a SoC with half the GPU at a lower clock is what fits that range, Nintendo would lower the clock further from that point, just as we see in TX1's 10watt/15watt design.

Ah okay, I see. That makes sense.

 

[Vash_the_Stampede]

10x for the CPU seems way too high for 6 cores based on what I remember of the IPC discussion we had months/year? ago. I think the realistic range for CPU improvements tops out at 9x, and is likely going to be lower than that since they won't use extravagant clock speeds.

And IIRC it won't quite reach the area of Series S, but it'll definitely be a fair bit closer to that level than the Switch was to the XB1/PS4 CPUs when it launched.


I think it's probably a good idea to assume they'll go conservative on most of the clocks specifically to ensure better yields and therefore supply/availability, especially in the current climate.

9x vs. 10x pretty close!

My mental math was >2x the cores * 3.0-3.5x IPC * 1.5x GHz

 

[Skittzo]

9x vs. 10x pretty close!

My mental math was >2x the cores * 3.0-3.5x IPC * 1.5x GHz

Well from what I remember 9x was said to be the top of the range for 8 A78 cores, not 6, which is what you had said for 10x.

Whatever the case I don't think we're looking at them hitting the top of the range in any category.

 

[ReddDreadtheLead]

9x vs. 10x pretty close!

My mental math was >2x the cores * 3.0-3.5x IPC * 1.5x GHz

This is a very optimistic clock frequency, and nothing is guaranteeing that they will actually have 7 cores for games. It could quite easily be 4 for games clocked high and 4 for games clocked low for the OS or other features.

Or only 6 cores in total with 5 for games.

Nothing is set, it’s best to draw your own conclusions.

 

[Z0m3le]

This is a very optimistic clock frequency, and nothing is guaranteeing that they will actually have 7 cores for games. It could quite easily be 4 for games clocked high and 4 for games clocked low for the OS or other features.

Or only 6 cores in total with 5 for games.

Nothing is set, it’s best to draw your own conclusions.

The discussion is upto, not is. I agree with most of what you are saying but they wouldn't waste those types of resources in a gaming device. Infact little cores don't make sense for game consoles, because you don't have idle performances or background tasks to worry about, that is why ARM designed A78C specifically for dedicated gaming devices.

 

[Vash_the_Stampede]

Just so I better understand

There is a AGX mode with 12x A78 and a Jetson Orin NX which is only 8x A78

The NX is offered in a little as 10w and 15w. Presumably this board comes with stuff the Nintendo won’t need.

Do we think Dane is half the AGX or a full NX?

I haven’t been able to find tflop specs for either at the lower power profiles. However, looks like AGX SoC consumes 50w at 4 tflops vs. the 3050’s 4.3 tflops consuming 80w. Both are on 8nm. Why such a big difference?

 

[Dekuman]

I fail to see how releasing a Switch lite OLED makes any sense, a lot of people will be interested in Switch Sports this year which a lite would not support. Would many current lite owners upgrade to an OLED version?

I can see a soft upgrade of Switch Lites with OLED screens near the Switch's end of life if perhaps they have spare panels left they could use, like the GBA SP with the brighter DS Lite screens. But that said OLED screens are actually quite expensive and if Furukawa is to be believed Switch OLED margins are less than regular Switches despite the $50 price increase, so I think the price factor alone may prevent them from doing an OLED Lite because Switch Lite has always been positioned as the cheap alternative to the hybrid version and upgrading it with an OLED screen would probably also push the price up.

 

[Mercury_Sagit]

Just so I better understand

There is a AGX mode with 12x A78 and a Jetson Orin NX which is only 8x A78

The NX is offered in a little as 10w and 15w. Presumably this board comes with stuff the Nintendo won’t need.

Do we think Dane is half the AGX or a full NX?

I haven’t been able to find tflop specs for either at the lower power profiles. However, looks like AGX SoC consumes 50w at 4 tflops vs. the 3050’s 4.3 tflops consuming 80w. Both are on 8nm. Why such a big difference?

Per Techpowerup's GPU DB, RTX 3050's theoretical computational throughput is actually around 8TF FP32, not 4.3. It also has more CUDA cores - 2304 vs 2048 (AGX) vs 1024 (NX) - and much higher GPU clocks - 1545 Mhz vs 1024 Mhz (AGX/NX). That can easily explains the different in power consumption.

 

[Vash_the_Stampede]

Per Techpowerup's GPU DB, RTX 3050's theoretical computational throughput is actually around 8TF FP32, not 4.3. It also has more CUDA cores - 2304 vs 2048 (AGX) vs 1024 (NX) - and much higher GPU clocks - 1545 Mhz vs 1024 Mhz (AGX/NX). That can easily explains the different in power consumption.

Sorry, I was referring to 3050 laptop version which has half the tflops.

NVIDIA GeForce RTX 3050 Mobile Specs

NVIDIA GA107, 1343 MHz, 2048 Cores, 64 TMUs, 32 ROPs, 4096 MB GDDR6, 1500 MHz, 128 bit

www.techpowerup.com