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

[Hero of Hyrule]

I believe in an announcement/tease/reveal in the September Direct, LFG

 

[SiG]

it also needs to make third parties happy. if the gap is just too large, then you can't conduct business

And yet the Switch still had numerous impossible ports.

I think we're forgetting the moneyhat side of the equation as well/buisness politics, and instead simply relying on the same reasoning of "if it can't match the other two in terms of horsepower, it's bust"...which was what I was complaining about initially.

 

[ShadowFox08]

So is it possible the next Switch could be a little thicker? I ask because really the current Switch's joy cons are a bit on the small side for me. Also a thicker Switch would probably allow for slightly better cooling, would it not?

There's a decent chance the system will be. I can see a slight redesign with joycons, which will likely have to fit in current switch hybrid models.

Remember, the OS cpu core can be downclocked while in game, so 6 or 8 cores makes more sense, especially because they don't add much to the cost as they fit in one package thanks to A78C, which was designed for games specifically.

Also, minimum portable clock is likely 460MHz just to cover any issues from Switch compatibility, you'd be looking at a very small power hit to support the additional 40mhz. That's 1.41tflops.

I'm a bit hesitant and not entirely convinced we are getting that much GPU power in handheld mode. Not in 8nm Samsung at least.

The way I interpret a PS4/+ portable is not a 1.4 TFLOP machine in handheld, but half or 60% of that. Something like 700-1000 gflops played at 720p. Really, this could be what Polygon's friend meant and might be more feasible on 8nm Samsung.

I look at steam deck and see it can run just over 25 watts running at 1.3Ghz GPU and 3Ghz CPU for like 80 minutes. Yes the RAM and storage power draw adds up as well.

1.4 TFLOPs in handheld mode seems like it would take up a lot of power draw on 8nm Samsung. Certainly going above 10 watts standard of what Nintendo Especially when we have to factor in other parts of the SOC. CPU (up to 8 cores), RAM (12GB DDR5), SSD, any Drake tech, all which could take more power draw than the switch Soc as well. And there is heat dissipation also.

5nm is more likely to get those raw clockspeeds though.

The most confident thing I feel about all this is that I don't think we will reach above 80% of Drake's total power in CPU and GPU with an acceptable power draw on 8nm, due to the handheld limitations.

Your English is perfectly understandable.

I think TSMC's N3 process node and/or TSMC's N3E process node is/are completely out of the question, considering that Apple's most definitely paying huge amounts of money to have temporary, exclusive access to TSMC's N3 process node and/or TSMC's N3E process node before other companies do.
And yes, there was a rumour from DigiTimes about Nvidia securing too much capacity for TSMC's 4N process node. And I think there are grains of truth to that rumour since TSMC has warned customers of excessive capacity.

And I don't believe I've heard any rumours about Nintendo not being happy with Nintendo's new hardware.

Agreed. Maybe for a revision in 2025 or later.

And yet the Switch still had numerous impossible ports.

I think we're forgetting the moneyhat side of the equation as well/buisness politics, and instead simply relying on the same reasoning of "if it can't match the other two in terms of horsepower, it's bust"...which was what I was complaining about initially.

We missed out on a lot more ports than the seemingly impossible ports that we got that were funded by money hatting. Others couldn't be ran at acceptable performance that devs invisioned. Also, the cost of carts vs discs were a major issue as well and can continue to be. Lots of AAA western and Japanese ports we missed out on.
Cod, assassin's Creed, resident evil games and others to name a few.

In order for the Drake to have the same CPU power gap as Switch vs PS4, it needs to have minimum of eight core A78s CPUs running at 1.0 GHz (or six at 1.4..). Anything less and the gap will widen. 4 core CPUs will not enough. Keep in mind that 1 core is sacrificed for the OS of course, and PS4's jaguar cores are 1.6 GHz, while current gen PS5 are 2.2x higher in clock speed.

 

[kvetcha]

I look at steam deck and see it can run just over 25 watts running at 1.3Ghz GPU and 3Ghz CPU for like 80 minutes. Yes the RAM and storage power draw adds up as well.

x86 vs ARM certainly comes into play here. An Apple M1 Mac Mini pulls 29W under max load for 4x 3.2ghz cores, 4x 2ghz efficiency cores, and a 2.6TF integrated GPU on TSMC 5nm.

 

[Look over there]

While I'd prefer N5 to maximize perf/watt, N7/N6 can still deliver quite an acceptable device. Sufficient power savings to get our GPU config in portable into a credible area of course, but there was also a nice leap in SRAM density between the 10nm and 7nm generations.
By default, the A78C is listed to allow for a maximum of 8 MB of L3 cache. This is a very low chance, super long shot, but I hope that Nvidia goes custom on that aspect and bumps the L3 cache up to like, 16 MB or so.
Referencing this to grab urls... Looking at this, on N7, we see that the chunk of 32 MB cache on a Zen 3 CCD takes up ~27 mm^2. The 64 MB 3D V-cache is ~36 mm^2. So there's a difference of about 9 mm^2 for that extra 32 MB. Divide by 4 to get about 2.25 mm^2 per 8 MB. So, I'm guessing that if you were to take a pre-existing chunk of cache (like 8 MB of L3) and enlarge it by 8 MB (to hit 16 MB), it's costing you 2-3 mm^2. I think that the bang for the buck is there.
And this gives us 31.8 Mib/mm^2 on N5, so I'm guessing about 2 mm^2 to expand 8 MB to 16 MB as a best case scenario.
(although funny enough, I'm less confident on the bang for the buck of going above 16 MB)

 

[Hermii]

Hey everyone. Didn’t get much info as we weren’t alone a lot but I did ask about the drive. He said that Drake uses the same kind of drive as the current Switch but that it could change as there’s still a while before release. That’s the last I’m going to ask as he seemed annoyed when I brought it up so I’m going to leave it there.

I hope you’re all having a good weekend!

Its kind of dissapointing, but should have expected it I suppose.

Should still have much shorter load times cause of no cpu bottleneck.

 

[Simba1]

If they can run on these older machines Desktop machines that have competent CPUs but aren’t as strong as the PS5/XBS, they can probably run good enough on a system like Drake (assuming 8 cores).



I’m not necessarily saying that you are wrong but more that the comparison is very flawed

Those older machines (XB1/PS4) don't have desktop CPUs but CPU (Jaguar) thats made for notebooks, but now they have full desktop CPU with high clocks.
Like wrote, 90% of current games are still cross gen, PS5/XSX hardware yet needs to be pushed and used only for next gen games without looking at last gen hardware when they developing games.

I disagree that comparison is flawed.

This “it’s mobile so it can’t compare” line of reasoning isn’t the whole truth, imo. It’s a hybrid after all. Hence the very name. They can take it beyond “mobile” when docked, it does have a fan. I realize there are limitations because it’s not using desktop hardware but that gap is not the same, especially when it’s custom developed for it’s purpose this time, and it’s not an off the shelf mobile chip. Markio was just a node change right? This is the first time they are developing a new chip for a successful platform that features the hybrid nature and optimizing for that.

ARM CPU (in this case most likely A78s) is mobile CPU, LPDDR memory is also low power RAM memory for mobile devices.
point that you can use it in docked mode or maybe even in desktop hardware doesnt make it hybrid or desktop CPU/RAM because its made to be low power and used in mobile devices.

 

[ILikeFeet]

And yet the Switch still had numerous impossible ports.

I think we're forgetting the moneyhat side of the equation as well/buisness politics, and instead simply relying on the same reasoning of "if it can't match the other two in terms of horsepower, it's bust"...which was what I was complaining about initially.

I'm talking in context of the Switch compared to the PS4/XBO. I honestly think 4 cores would be a worse situation than the Switch was unless they clocked those cores high, all over 3GHz or something. and by that point, we're breeching the power budget.

don't lump me in with that "it needs to match horsepower" reasoning. I never believed that and have repeatedly said that I don't think cpu-limited games will be all that plentiful, which would be a boon for drake. but with excess clock and 8 cores, you bet your ass devs will make us of that when they can. having 4 cores is even limiting PC and they have higher power limits

 

[Mr Swine]

Can we compare Drakes GPU specs to the Nvidia MX570 and RTX 2050 Laptop since they both are Ampere and have 1024 and 2048 Cuda cores? They are on Samsung 8nm mode but run at much higher frequencies (570 has 64bit bus while 2050 has 128)

 

[Deleted member 887]

I'm a bit hesitant and not entirely convinced we are getting that much GPU power in handheld mode. Not in 8nm Samsung at least.

The way I interpret a PS4/+ portable is not a 1.4 TFLOP machine in handheld, but half or 60% of that. Something like 700-1000 gflops played at 720p. Really, this could be what Polygon's friend meant and might be more feasible on 8nm Samsung.

I agree with you on the limitations of Samsung 8nm, but 1TFLOP in handheld is the absolute bare minimum. OG Switch's handheld GPU clock was 302Mhz. Drake has 12SMs. 12SMs @302MHz is 1TFLOP. You can't clock under that for a couple reasons.

First, backwards compat - you cannot reliably emulate (either in hardware OR software) any system at a lower clock speed, even if you have more cores than the OG system. If Drake has hardware comaptibility, then 302Mhz is the minimum, if there is a software solution involved on the GPU side, then it will almost definitely need to be higher.

Second, folks have pointed out that Ampere's power curve breaks down at those clocks. Meaning clocking under wouldn't save you power consumption.

As for "Ps4 Level in handheld" - Nate has used similar language to talk about the performance of the device, not just Polygon. The PS4 wasn't a 1.4 TFLOP machine, though, it was 1.82. So the 1.3 number I used in my estimates was broadly reinterpreting "PS4" as "last gen" and using the Xbox One as my baseline. As many games failed to hit the 1080p goal of the PS4, I don't think we can go lower on the basis of the 720p screen.

I look at steam deck and see it can run just over 25 watts running at 1.3Ghz GPU and 3Ghz CPU for like 80 minutes. Yes the RAM and storage power draw adds up as well.

1.4 TFLOPs in handheld mode seems like it would take up a lot of power draw on 8nm Samsung. Certainly going above 10 watts standard of what Nintendo Especially when we have to factor in other parts of the SOC. CPU (up to 8 cores), RAM (12GB DDR5), SSD, any Drake tech, all which could take more power draw than the switch Soc as well. And there is heat dissipation also.

Yeah, I totally agree. My initial exercise was to see if I could use the existing docs on Orin to tease out some rough power draw for the TPCs in Ampere on Samsung 8nm.

I thought what would come out would be a version of Drake that ran 8 CPUs and 12SM, but something close to current clock speeds for the Switch. That would track with a modest upgrade that matched all our minimum performance expectations without being a next-gen blowout device that we sometimes start anticipating. What I found was I just couldn't make that happen. More than that, I'm not sure that you can run 12SMs, no CPUs at all for under 8W.

The most optimistic power draw numbers for the GPU, with the most pessimistic clocks, gives the Drake GPU consuming 8.2W. If we assume that the OS CPU can clock itself down to nothing, and run 3 AMD ARM CPUs at 1Ghz - basically the current Switch config, 0 upgrades on the CPU side - then we get an exact 10W for power draw. That is such an extremely CPU limited machine I'm not sure it's actually possible to use the GPU, because the CPU won't be able to keep it fed.

This is all back of the envelope math, maybe there is some power savings on the table I'm not seeing. Orin swaps out half of the RT cores and runs the tensor cores instead at double rate. My assumption has been that this tradeoff has no power implications, but perhaps I'm wrong. Maybe there is more than 10% IO overhead, or leakage due to unusued PVA/DLA silicon that accounts for more savings moving from Orin to Drake. If I get really optimistic, I can get 6 CPU cores in there at 1GHz, but that's starting to get into fantasy land to result in a device that still kinda sucks.

TL;DR - Drake GPU on Samsung 8nm might be actively impossible. Even if it isn't, it seems completely incompatible with a CPU upgrade in anyway, which is likely a non-starter. I'm ready to let 8nm go.

 

[Deleted member 887]

Can we compare Drakes GPU specs to the Nvidia MX570 and RTX 2050 Laptop since they both are Ampere and have 1024 and 2048 Cuda cores? They are on Samsung 8nm mode but run at much higher frequencies (570 has 64bit bus while 2050 has 128)

Not meaningfully without knowing clock speed and amount of RAM- we know Drake has 1536 CUDA cores, if that means anything to you

 

[undefined]

Can we compare Drakes GPU specs to the Nvidia MX570 and RTX 2050 Laptop since they both are Ampere and have 1024 and 2048 Cuda cores? They are on Samsung 8nm mode but run at much higher frequencies (570 has 64bit bus while 2050 has 128)

I thought they separated their series by architecture but seems like 2050 is actually Ampere while the rest of the RTX 20xx series is Turing

 

[ILikeFeet]

I thought they separated their series by architecture but seems like 2050 is actually Ampere while the rest of the RTX 20xx series is Turing

it's a very weird product release. they could have just called it the 3040 or some shit

 

[ReddDreadtheLead]

Those older machines (XB1/PS4) don't have desktop CPUs

I never said the XB1 and PS4, I said older desktop machines.

I disagree that comparison is flawed.

You can disagree, but that doesn’t make you correct. One architecture aims to be very performant at very low power draw the other one isn’t meant for low power draw, it is meant to give server class level performance and scales up from that desktop where power draw isn’t a concern (yet). They have completely different goals in mind here.

Hell, you can have an arm cpu match an x86 cpu with a fraction of the TDP. It isn’t even funny.




And on another note, the CPUs in the other consoles cannot draw even close to 100W at all.

Like at all.

The majority of the TDP in those systems is going to the GPU, the GDDR6 memory, the Solid State Storage, etc.

The CPUs should draw no more than ~50-65W at the clockspeeds of the systems. 65W is pushing it.

Well, the 4700S at 4GHz draws like 75W, but it doesn’t operate at that, it operates at 3.5GHz in its console counterpart.


There are even other Zen 2 CPUs with similar configs that are rated for 35-54W!

Those were just slightly overclocked from base.

And 3 A78 at 2GHz would only be about XB1X/PS4Pro territory.


I know you said 4, but only 3 would be available for games.

 

[Pokemaniac]

I've long generally assumed that Drake was 8nm, because I didn't see a reason to assume otherwise, and that we would see very low clocks to compensate power-wise. My assumption has also been that this would be positioned as a "Pro" device and that Nintendo would be able to squeeze a "Successor" device out of the same investment by some combination of die shrinking Drake and/or riding other efficiency gains to let them up the clocks. But I think I have to finally concede that doesn't track.

I'm not sure this idea really makes sense regardless of any power draw concerns. It's kind of backwards compared to the typical revision strategy for consoles. From a hardware perspective, typically the shift in architecture is what yields the bigger boosts, and from a software perspective, there's only so much change you can tolerate before you just have to functionally treat something as a new platform anyway.

This is why I've been very confident it's a successor since we first heard it was Ampere. That isn't to say there won't or shouldn't be a fair bit of role compression going on, but over the long term, I expect it to be clear that this is the second generation Switch.

 

[mjayer]

ARM CPU (in this case most likely A78s) is mobile CPU, LPDDR memory is also low power RAM memory for mobile devices.
point that you can use it in docked mode or maybe even in desktop hardware doesnt make it hybrid or desktop CPU/RAM because its made to be low power and used in mobile devices.

Tell that to Apple, who is now putting custom Arm CPU's in desktop machines.

FWIW I'm typing this right now on a M1 MAX laptop that is quite a beast. My point is this new chip from Nvidia is still, really, the first one developed from the ground up to be a Switch engine. That's exciting.

 

[ILikeFeet]

ARM CPU (in this case most likely A78s) is mobile CPU, LPDDR memory is also low power RAM memory for mobile devices.

calling ARM cpus "mobile" is missing the forest for the trees, methinks. ARM is designed from the bottom of the power consumption spectrum and moves up. x86, on the other hand, starts from the top and moves down. there's nothing about ARM that can't be a desktop cpu. and those exist even. would you call Neoverse a "mobile" cpu when it's powering large server farms? what about intel Atoms that were in mobile phones?

 

[PedroNavajas]

it's a very weird product release. they could have just called it the 3040 or some shit

Why, when they can later release the 3040 for double the money!?

 

[Terrell]

Thank you for the clarification.

Basically, devkits for Nintendo's new hardware are using eMMC 5.1 for the internal flash storage, which is the same type of internal flash storage used on the Nintendo Switch, Nintendo Switch Lite, and the OLED model, according to rumours.

@Polygon, appreciate the info.

That’s disappointing, but also not a death knell for something better. If early dev kits aren’t focused on optimizing data access, no sense loading them up with eUFS just yet. I was just hoping against hope that there are dev kits out there that had this pinned down.

 

[ILikeFeet]

Why, when they can later release the 3040 for double the money!?

well there wouldn't be anything stopping them from releasing them as 3040s as dgpus. the Radeon 6400 and 6500 are both laptop gpus after all

@Polygon, appreciate the info.

That’s disappointing, but also not a death knell for something better. If early dev kits aren’t focused on optimizing data access, no sense loading them up with eUFS just yet. I was just hoping against hope that there are dev kits out there that had this pinned down.

I'm curious if emmc is that much cheaper than ufs. lower end phones still use it over ufs, so there's probably some truth to it. the steam deck doesn't seem to be too impaired by the choice to use emmc, but we haven't seen many full current gen games to guage

 

[Dakhil]

I do wonder where Nintendo could use Macronix's 48-layer 3D NAND, assuming the report from Taipei Times about Nintendo being the first customer of Macronix's 48-layer 3D NAND is true.

 

[ILikeFeet]

I do wonder where Nintendo could use Macronix's 48-layer 3D NAND, assuming the report from Taipei Times about Nintendo being the first customer of Macronix's 48-layer 3D NAND is true.

given the timeline, I wonder if that was delayed due to covid. also wonder if it was used for game cards, but without increases to speeds/file size

 

[Deleted member 887]

I'm not sure this idea really makes sense regardless of any power draw concerns. It's kind of backwards compared to the typical revision strategy for consoles.

I don't think it's that unprecedented - but I also realize I misspoke. What I meant was it's totally reasonable to have a "next gen" device have more power via the arch than just by increasing clocks. This is totally normal outside of consoles. Drake is so fat that a 130% power boost is possible while keeping clocks low, which is convenient because they're power constrained. And again, because they're power constrained, a die shrink offers a chance for a substantial boost that it doesn't offer Sony/Microsoft - they're already pushing the power/perf envelope with their APU.

I think whatever Nintendo does with Drake, something is about to be unprecedented in the industry.

1. If it's launched as a successor, they'll be doing so with very little warning
2. If it's launched as a pro, they'll be the first to make their "revision" a completely new arch
3. If it's launched as a pro, and slowly overtakes it's predecessor, before something similar happens to it, it'll be the first time a console maker has done that period.

If Nintendo goes for option 3, I can totally imagine a Switch '26 which is a die shrunk, clock boosted version of Switch '23, which would give Drake 6-7 years to make back it's design costs before repeating the process. If Nintendo goes for option 2, then I expect the battery life revision ala Mariko, but I wouldn't be utterly surprised by a "pro Drake".

It'll be interesting to see Nintendo's BC strategy here. If they can do something in hardware that isn't super cost/perf inefficient, then they can ride that out a long time. If they can't and they need some level of software emulation, then Nintendo will need to ride out Ampere until there is such a big leap in power available they can reliably emulate it.

 

[Dakhil]

given the timeline, I wonder if that was delayed due to covid. also wonder if it was used for game cards, but without increases to speeds/file size

Takashi Mochizuki when he was working for the Wall Street Journal has heard that Nintendo delayed 64 GB Nintendo Switch Game Cards until 2019.

And Anton Shilov on Anandtech has speculated that Macronix's 48-layer 3D NAND could be used for 64 GB Nintendo Switch Game Cards.

 

[PashArcade]

I think whatever Nintendo does with Drake, something is about to be unprecedented in the industry.

1. If it's launched as a successor, they'll be doing so with very little warning
2. If it's launched as a pro, they'll be the first to make their "revision" a completely new arch
3. If it's launched as a pro, and slowly overtakes it's predecessor, before something similar happens to it, it'll be the first time a console maker has done that period.

If Nintendo goes for option 3, I can totally imagine a Switch '26 which is a die shrunk, clock boosted version of Switch '23, which would give Drake 6-7 years to make back it's design costs before repeating the process. If Nintendo goes for option 2, then I expect the battery life revision ala Mariko, but I wouldn't be utterly surprised by a "pro Drake".

It'll be interesting to see Nintendo's BC strategy here. If they can do something in hardware that isn't super cost/perf inefficient, then they can ride that out a long time. If they can't and they need some level of software emulation, then Nintendo will need to ride out Ampere until there is such a big leap in power available they can reliably emulate it.

And this is what I don't like, imagine buying an expensive system, and knowing it's immediately getting replaced by another

Consoles aren't phones

Btw, where is the "6-7 years come from? By my math 2023 to 2026 only gives Drake a sad 3 years life cycle, not to mention the possible $500 price point

I can't find a reason to justify the purchase, and we're better off waiting for Switch 26

 

[NintenDuvo]

Have we had any rumblings about this device going in full production?

 

[kvetcha]

And this is what I don't like, imagine buying an expensive system, and knowing it's immediately getting replaced by another

Consoles aren't phones

Btw, where is the "6-7 years come from? By my math 2023 to 2026 only gives Drake a sad 3 years life cycle, not to mention the possible $500 price point

I can't find a reason to justify the purchase, and we're better off waiting for Switch 26

Mariko was a die shrunk, clock boosted version of Erista but we don't say that it reset the Switch's lifecycle.

 

[Dakhil]

Have we had any rumblings about this device going in full production?

As far as I know, no.

 

[NintenDuvo]

As far as I know, no.

Darn thats not a good sign for early 2023 :/

 

[Deleted member 887]

And this is what I don't like, imagine buying an expensive system, and knowing it's immediately getting replaced by another
Consoles aren't phones

I’m sorry you don’t like it! I get that.

But the play station already did it. It’s just now trying to get away from it - and failing.

Btw, where is the "6-7 years come from? By my math 2023 to 2026 only gives Drake a sad 3 years life cycle, not to mention the possible $500 price point

Again I’m imagining that Drake gets games for 6 years even if a “Drake Pro” comes along in 3.

I can't find a reason to justify the purchase, and we're better off waiting for Switch 26

 

[Serif]

1. If it's launched as a successor, they'll be doing so with very little warning
2. If it's launched as a pro, they'll be the first to make their "revision" a completely new arch
3. If it's launched as a pro, and slowly overtakes it's predecessor, before something similar happens to it, it'll be the first time a console maker has done that period.

I am wondering, how much of a 'warning' should Nintendo give when announcing a next-generation Switch? This isn't a new brand-spanking idea like the Wii -> Wii U -> Switch transition, nor is this like the Wii U where they need to alleviate fears by announcing they're working on new hardware early. I guess what I'm asking is - who is the warning intended for? I am assuming that with the number of third parties that have already received devkits, that they'll be showing off their titles or adding Switch 'Ultra' branding to games that have already been announced. So my assumption is that key partners won't be thrown off once this is revealed. I'm sure the backwards compatibility helps, projects currently in progress will not be nullified once this hardware is announced. But this isn't my wheelhouse so I'm curious if there's specific way they have to go about this.

Anyways I am anticipating option 3, the successor 'soft-launch', and I imagine the announcement kind of like this -> "Introducing Switch Ultra, the newest addition to the Switch family of consoles. Compatible with all Nintendo Switch games and select Nintendo Switch Ultra exclusive titles. Plays select Nintendo Switch and Nintendo Switch Ultra titles up to 4K and 60 FPS.". Shown alongside games like Elden Ring and Jedi Survivor (I hope!) And all games Nintendo releases from then on have 'playable on Switch Ultra' branding on them. I want Nintendo to treat this as their Pro / enhanced cross-gen for the first year or so and squeeze out as many Drake patches as humanly possible for their evergreens, and at the same time, I want third parties to go full steam-ahead with this as successor hardware and bring out the big guns.

With Drake guts crammed into potentially the Switch OLED shell, this almost feels like a relaunch of the Switch, I'm pretty excited.

 

[PashArcade]

Again I’m imagining that Drake gets games for 6 years even if a “Drake Pro” comes along in 3.

Ah, that's a relief, as long as Drake gets its own big exclusive games, I wouldn't mind its position, games are what truly matters after all

 

[PashArcade]

Anyways I am anticipating option 3, the successor 'soft-launch', and I imagine the announcement kind of like this -> "Introducing Switch Ultra, the newest addition to the Switch family of consoles. Compatible with all Nintendo Switch games and select Nintendo Switch Ultra exclusive titles. Plays select Nintendo Switch and Nintendo Switch Ultra titles up to 4K and 60 FPS.". Shown alongside games like Elden Ring and Jedi Survivor (I hope!) And all games Nintendo releases from then on have 'playable on Switch Ultra' branding on them. I want Nintendo to treat this as their Pro / enhanced cross-gen for the first year or so and squeeze out as many Drake patches as humanly possible for their evergreens, and at the same time, I want third parties to go full steam-ahead with this as successor hardware and bring out the big guns.

Me too, hopefully with all the hardware prowess we can finally have more JRPGs on the Nintendo ecosystem

1. Scarlet Nexus

2. Soul Hackers 2

3. Star Ocean: The Divine Force

4. Armed Fantasia: To The End of The Wilderness (The game has been successfully funded, congrats to them )

5. Kingdom Hearts IV

 

[PashArcade]

Anyways I am anticipating option 3, the successor 'soft-launch', and I imagine the announcement kind of like this -> "Introducing Switch Ultra, the newest addition to the Switch family of consoles. Compatible with all Nintendo Switch games and select Nintendo Switch Ultra exclusive titles. Plays select Nintendo Switch and Nintendo Switch Ultra titles up to 4K and 60 FPS.". Shown alongside games like Elden Ring and Jedi Survivor (I hope!) And all games Nintendo releases from then on have 'playable on Switch Ultra' branding on them. I want Nintendo to treat this as their Pro / enhanced cross-gen for the first year or so and squeeze out as many Drake patches as humanly possible for their evergreens, and at the same time, I want third parties to go full steam-ahead with this as successor hardware and bring out the big guns.

Me too, hopefully with all the hardware prowess we can finally have more JRPGs on the Nintendo ecosystem

1. Scarlet Nexus

2. Soul Hackers 2

3. Star Ocean: The Divine Force

4. Armed Fantasia: To The End of The Wilderness (The game has been successfully funded, congrats to them )

5. Kingdom Hearts IV

 

[Terrell]

I'm curious if emmc is that much cheaper than ufs. lower end phones still use it over ufs, so there's probably some truth to it. the steam deck doesn't seem to be too impaired by the choice to use emmc, but we haven't seen many full current gen games to guage

Yes, but that's primarily because of its ubiquity outside the smartphone market, for devices that don't demand the kind of access speeds or capacities of even the most budget smartphone. But eUFS is making headway on price parity, since eUFS 2.1 has made its way into smartphones in the $300 range as far back as 2019. And especially since the Nvidia architecture this hardware is being built on is already built with eUFS in mind, I/O lanes and all, it'd be a shame to backslide without a much more significant cost savings.
But in an early dev kit? They might only need to include 32GB (128GB at MOST) for testing and, if they're not concerned with load time optimizations until the final hardware is pinned down and producible, they can cheap out pretty substantially, even using the eMMC harvested off of faulty Switch motherboards.

given the timeline, I wonder if that was delayed due to covid. also wonder if it was used for game cards, but without increases to speeds/file size

Takashi Mochizuki when he was working for the Wall Street Journal has heard that Nintendo delayed 64 GB Nintendo Switch Game Cards until 2019.

And Anton Shilov on Anandtech has speculated that Macronix's 48-layer 3D NAND could be used for 64 GB Nintendo Switch Game Cards.

I wouldn't look at it too deeply. The speculation runs counter to the established Game Card tech, which is ASIC ROM. Moving to a volatile storage method increases read speed. My feeling is that the report is either mistaken (which is more likely) or the 3D NAND is being used to achieve something else, though what that might be is a mystery to me.

Speaking of which, Dakhil, do we know what process node is being used to make the current ASICs in Game Cards? Since most ROM tech hasn't ventured out of the safe pre-FinFET purely planar nodes (I think 28nm or larger?), I'm assuming we'd be looking at something in the 28-40nm range.

I ask because TSMC is actually looking to add production capacity of 28nm, thinking that the greater number of chips per wafer will offset the increased cost to their clients who can't get on a 32nm or higher production line due to no capacity increases, and it's likely other fabs will follow their lead. If we're looking at greater capacities on Game Cards, a shrink to a smaller planar node would be the ticket, assuming it's not already at that size.

 

[lucapfstr]

I'm not sold on the idea of a soft launch, just because it seems like a marketing nightmare. It's inconvenient to market things like "this is a switch 2 game but it's also compatible with the switch but only the pro version." I think plenty of people would get it, like those who got a Wii U despite it's terrible marketing. But In my eyes it's simply better to rip off the band-aid that yes, this is the next-gen switch 2 or whatever. Just say that you plan on officially supporting the original switch for _ more years, and once that time passes, release the die shrink. Rinse and repeat!

 

[Dakhil]

Speaking of which, Dakhil, do we know what process node is being used to make the current ASICs in Game Cards? Since most ROM tech hasn't ventured out of the safe pre-FinFET purely planar nodes (I think 28nm or larger?), I'm assuming we'd be looking at something in the 28-40nm range.

I don't know how recent this is, but a data sheet from Macronix mentioned 45 nm** is used for XtraROM (p. 5).

** → a marketing nomenclature used by all foundry companies

 

[ReddDreadtheLead]

I'm not sold on the idea of a soft launch, just because it seems like a marketing nightmare. It's inconvenient to market things like "this is a switch 2 game but it's also compatible with the switch but only the pro version." I think plenty of people would get it, like those who got a Wii U despite it's terrible marketing. But In my eyes it's simply better to rip off the band-aid that yes, this is the next-gen switch 2 or whatever. Just say that you plan on officially supporting the original switch for _ more years, and once that time passes, release the die shrink. Rinse and repeat!

It seemed to have worked for Microsoft with the Series lol

 

[Terrell]

I don't know how recent this is, but a data sheet from Macronix mentioned 45 nm** is used for XtraROM (p. 5).

** → a marketing nomenclature used by all foundry companies

OOF. As a general rule, ROM chips don't typically see a die shrink unless it's absolutely necessary, so even if this is dated for 2014, it's still probably accurate. (EDIT: Yep, website still lists 48nm in the spec, now that I understand what that number in the one column represents)
Since it's a planar node, we should be able to confirm just by looking at the dimensions (one of the benefits of planar nodes, as I understand it), but let's just take that at face value. That'd mean a die shrink to the smallest possible planar node would means you could get double the ROM chips from the same amount of wafer. There's what your holdup on the 64GB Game Cards has likely been, having to engineer the ROM to a new node and get production going.

 

[Pokemaniac]

I'm not sold on the idea of a soft launch, just because it seems like a marketing nightmare. It's inconvenient to market things like "this is a switch 2 game but it's also compatible with the switch but only the pro version." I think plenty of people would get it, like those who got a Wii U despite it's terrible marketing. But In my eyes it's simply better to rip off the band-aid that yes, this is the next-gen switch 2 or whatever. Just say that you plan on officially supporting the original switch for _ more years, and once that time passes, release the die shrink. Rinse and repeat!

It's not that complicated. All they have to say is it's a game that you can play on both Switch and Switch 2, or just call it a Switch game and start throwing "also playable on Switch 2" stickers on the box like they used to do for 2DS.

 

[Dakhil]

OOF. As a general rule, ROM chips don't typically see a die shrink unless it's absolutely necessary, so even if this is dated for 2014, it's still probably accurate. (EDIT: Yep, website still lists 48nm in the spec, now that I understand what that number in the one column represents)
Since it's a planar node, we should be able to confirm just by looking at the dimensions (one of the benefits of planar nodes, as I understand it), but let's just take that at face value. That'd mean a die shrink to the smallest possible planar node would means you could get double the ROM chips from the same amount of wafer. There's what your holdup on the 64GB Game Cards has likely been, having to engineer the ROM to a new node and get production going.

The information from the website and the data sheet could be very old since Macronix's 2019 annual report mentions "Jun. (2014) ‧ Mass production of 32nm XtraROM® products." (p. 5)

 

[lucapfstr]

It's not that complicated. All they have to say is it's a game that you can play on both Switch and Switch 2, or just call it a Switch game and start throwing "also playable on Switch 2" stickers on the box like they used to do for 2DS.

Yeah, that works. I meant it more in the sense of calling this 2023 model a ‘pro’ and transitioning that same architecture to be called ‘switch 2’ later on. It’s late enough in the life cycle that they could release a console called Switch 2 already, and phase out the original architecture a couple of years in.

 

[Terrell]

The information from the website and the data sheet could be very old since Macronix's 2019 annual report mentions "Jun. (2014) ‧ Mass production of 32nm XtraROM® products." (p. 5)

Hmmm... dang, looks like we'd need to dissect and measure to get a clear idea of the nodes in use after all.

Here's a look at an original Macronix ROM chip from BotW, if anyone is able to figure out the planar node in use from it:

Games that required larger capacity were around double the height of this one, attached to a circuit board that ran the full length of the Game Card casing.

 

[ShaunSwitch]

I don't think it's that unprecedented - but I also realize I misspoke. What I meant was it's totally reasonable to have a "next gen" device have more power via the arch than just by increasing clocks. This is totally normal outside of consoles. Drake is so fat that a 130% power boost is possible while keeping clocks low, which is convenient because they're power constrained. And again, because they're power constrained, a die shrink offers a chance for a substantial boost that it doesn't offer Sony/Microsoft - they're already pushing the power/perf envelope with their APU.

I think whatever Nintendo does with Drake, something is about to be unprecedented in the industry.

1. If it's launched as a successor, they'll be doing so with very little warning
2. If it's launched as a pro, they'll be the first to make their "revision" a completely new arch
3. If it's launched as a pro, and slowly overtakes it's predecessor, before something similar happens to it, it'll be the first time a console maker has done that period.

If Nintendo goes for option 3, I can totally imagine a Switch '26 which is a die shrunk, clock boosted version of Switch '23, which would give Drake 6-7 years to make back it's design costs before repeating the process. If Nintendo goes for option 2, then I expect the battery life revision ala Mariko, but I wouldn't be utterly surprised by a "pro Drake".

It'll be interesting to see Nintendo's BC strategy here. If they can do something in hardware that isn't super cost/perf inefficient, then they can ride that out a long time. If they can't and they need some level of software emulation, then Nintendo will need to ride out Ampere until there is such a big leap in power available they can reliably emulate it.

There is also the option that they launch Drake on an advanced node like TSMC 4nm/5nm and don't die shrink it for its entire 6 year lifespan.

This would be in the instance that the display is the same 720p OLED display in the OLED model. They could use very reserved clocks in handheld mode, let's say the switch 460mhz profile and then just have a bigger gap between handheld and docked with a 1.3GHZ docked GPU clock. It would be 1.4 TFLOP in handheld and 4 TFLOPS docked, kind of lines up with what the rumours have said. Aiming for somewhere between OG and Mariko battery life, something like 4.5 hours min battery life for demanding games.

The advantage here is they would save money down the line doing a redesign and they could launch a lite model of the switch earlier by using binned Drake chips that can't sustainably hit the 1.3GHZ docked GPU clock. Using this approach would probably net them really good volume of usable chips per wafer, I don't see many chips not being able to hit a 460mhz clock sustainably. Chip would be small as well, less than 80mm^2 I would wager.

This would also ride on Nintendo being offered a good deal on Drake being on a more advanced node to help nvidia avoid a TSMC fine. In the instance the rumoured TSMC fine is false then I see TSMC 6nm more likely and a die shrink akin to the Mariko shrink 2 years into the consoles lifespan.

 

[Hajeman]

1. If it's launched as a successor, they'll be doing so with very little warning
2. If it's launched as a pro, they'll be the first to make their "revision" a completely new arch
3. If it's launched as a pro, and slowly overtakes it's predecessor, before something similar happens to it, it'll be the first time a console maker has done that period.

For me, if a Switch Drake is 3 - 4x more powerful than the current Switch, better promote it as a revision. But, if a Switch Drake is 10x or more powerful than the current one, better advertising it as a successor. Please keep it simple (don't make it half revision half successor) because general gamers don't see what we see here

 

[ShaunSwitch]

For me, if a Switch Drake is 3 - 4x more powerful than the current Switch, better promote it as a revision. But, if a Switch Drake is 10x or more powerful than the current one, better advertising it as a successor. Please keep it simple (don't make it half revision half successor) because general gamers don't see what we see here

I think asking for a ten times leap is a bit of a lofty goal. We know it has six times the GPU core count and is unlikely to run those at lower clocks than the original switch so a six times jump at minimum seems reasonable.

Coincidently a six times leap is about the same as the jump between the ps4 and PS5.

 

[ReddDreadtheLead]

Maybe I’m too optimistic, but if people care about it:

A jump from Switch to Switch Drake will have a bigger perceived jump than from XB1 to Series X or PS4 to PS5.


Even if it’s about the same, people may see Drake as a bigger jump than the PS5 and Series X to their respective consoles.

 

[Hajeman]

I think asking for a ten times leap is a bit of a lofty goal. We know it has six times the GPU core count and is unlikely to run those at lower clocks than the original switch so a six times jump at minimum seems reasonable.

Coincidently a six times leap is about the same as the jump between the ps4 and PS5.

My comment based on the reply here said Drake in TV mode is around PS4 (1.4 - 1.8tflops) and some said around PS4 Pro or Xbox Series S (4tflops). If around PS4, it's a revision. If around PS4 Pro/XSS it's a successor. It's not board well if it's a successor but only 3 - 4x powerful & if it's 8 - 10x powerful why calling it's a revision?

 

[Hermii]

Maybe I’m too optimistic, but if people care about it:

A jump from Switch to Switch Drake will have a bigger perceived jump than from XB1 to Series X or PS4 to PS5.


Even if it’s about the same, people may see Drake as a bigger jump than the PS5 and Series X to their respective consoles.

Kinda depends how much of a gamechanger dlss really is on this device.

And if storage is fast enough to run stuff like Nanite without too much compromise.

 

[ReddDreadtheLead]

Don’t use FLOPs!