-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
Based on how tech has progressed since the Switch, how powerful could a Switch Pro in early 2021 actually be?
- Thread starter ItWasMeantToBe19
- Start date
- Status
Well technically precedence would dictate 5nm.
PS4/XBO came out on 28nm, then a few years later Switch came on 20nm.
In the same way PS5/XSX are on 7nm now, in a few years it should be 5nm.
So are people either underestimating or overestimating specs while arguing if it's best to have a "pro version" sooner vs having a more powerful "sequel" later?
Current discussion is basically completely about a sequel system. No discussion has been happening regarding a 2021 Pro system for the latter half of the thread, I believe.
mass production for 5nm starts this year for Apple and Huawei products
not optimized for games, but, in theory, could possibly be used as a preliminary test platform. a proof of concept, if you will, for stuff like DLSS. has more than enough power to overcome differences including the caramel cpu cores
It sounds like 'Pro' is not going to happen and we're full bore towards a Switch 2. Though with current events being what they are, who knows when it will appear. 2022?
Most here seem to expect 2023. And most speculation here is based on that expectation as well.
those have large metal bodies that act as a heat spreader and large graphite pads. switch 2 may need more than that due to the smaller body
I don't see switch selling like bananas in it's current forms going away for at least 2 years so this makes sense to me too.
I could see Nintendo still doing basically a throw away pro model to milk its hardcore fans one more time before dropping the switch 2 like 1 year later though. Complete with the n3ds strategy of having like 2 exclusive games for it.
I actually think a $199 home console only version is the most likely iteration we might see.
Or just a price drop to $199 for the full model and $150 for the lite.
If it had full backwards compatibility and ran current software with better frame rates and resolution, those two exclusives would be more than enough to convince me to jump in.
I don't see anything new coming from Nintendo in the hardware space for two years or more, personally. The Switch is lighting up the markets as it is.
I think what they meant was Switch used 20nm when PS4 Slim/One S had moved to 16nm. Idk if there is a major difference but I personally think Nintendo will stay a node behind for costs. So if 5nm+ is 2021, then maybe 5nm in 2022-23 for switch 2 imo
I don't expect a Switch 2 until 2023 earliest, I mean why go into a full fledged sequels this early into the Switch success?
Would a mid-gen upgrade be more sensible to continue momentum?
Depends on when it releases I'd say. If 5nm is the new standard in 2021 then a 2023 Switch using 5nm makes sense, just like a 2017 Switch used 2015 tech.
I think most of us have given up on the idea of a pro. Basically all the current speculation is about a Switch 2.
Nintendo is pretty comfortable with 5 year gens and even a mid gen refresh will be way behind for third party ports because it will be hamstrung by the og switch. They'll still likely break 100 mil units in a 5-6 year timeframe, they'll be comfortable with moving on and even their own internal devs will be pushing for it by the end of next year imo. So a fall 2022 switch 2 launch makes sense. 2023 is probably more likely though.
Well, the scenario you sketch matches up with what Onix555 probably meant: Switch 2 is likely to use a node more advanced than the consoles do that are releasing 3 years earlier, but it's entirely that it won't have as advanced a node as a revision of those consoles released 3 years later and roughly concurrently with the Switch 2. That would point to 5nm instead of XSX/PS5's 7nm+ node, and a 5nm+ or 3nm node for the revision of the XSX/PS5.
Well Apple I believe gets 5nm this year, AMD gets it next year for Zen 4, graphics cards come in 5nm early 2022 and 3nm chips ship out later that year, with Nintendo Switch buying up freed 5nm production a year later. Time line is a year slower than the switch using 20nm actually, so 5nm isn't the optimistic choice, it's actually right about where it makes the most economical sense.
When did 20nm mass production start? I find articles going back to 2011 but also as recently as 2014 stating that Samsung started mass 20nm production lol
Whatever the timeline appears to be now I think we'll have a better idea in 2021 when products start launching. It'll be easier to make good educated guesses then, assuming they use a similar timeline as they did for Switch 1.
Assuming they even make a Switch 2 and not something completely different again.
I think we might start hearing rumors of a switch successor next year. very early dev kits might start going out to field developer opinions and what not.
2014 for Apple.
Nikkei reported that a next gen Switch was already in production early last year.
Im thinking you might hear very light rumours of it. But 2022 is when it'll really heat up; probably get a codename announcement by summer.
Good point. Yeah I never actually thought they'd abandon the concept, it was kinda tongue in cheek.
So here's a topic I don't think we've discussed as it relates to the Switch 2: in the XSX/PS5, much ado has been generated regarding the choice between more CUs and higher frequencies. But that discussion was held in the context of two boxes that are a bit more malleable in terms of power requirements, whereas the Switch 2 has a comparatively strict limit on size and power draw (for handheld mode at least), as well as price. So, what makes more sense? Let's start from the assumption that we have a power limit for the GPU that we do not want to exceed, otherwise the battery life becomes too short to be considered acceptable. In that case, let's say we have a GPU configuration that has only used 83% of the power budget we "want" it to use. What do we do: increase the SM by 20% (as power draw scales linearly with SM count), or increase frequencies by 10% (as power scales quadratically with frequency)? As I see it, the following are the benefits and drawbacks of the two approaches (please feel free to add on to or correct the following list!):
20% higher SM count:
+ More ALU compute power than the frequency boost : The ALU compute power scales linearly with the SM count, and also linearly with frequency. As a result, the SM count increase adds 20% to the ALU compute power (this is the common TFLOPS number you always see thrown around), as opposed to a 10% increase with the frequency boost. This means more compute power for scene rendering and for post-processing (note that post-processing does not include DLSS!).
+ More tensor cores : I've checked on the tensor core and SM counts for the RTX cards, and each of them have a factor of 8 between them (# tensor cores = 8 * #SM). This leads me to believe that inside each SM 8 tensor cores are packed. As a consequence, increasing the SM count means increase the tensor core count by the same amount. This means that our tensor FLOPS count increase by 20%, whereas it would only increase by 10% in the case of a frequency boost. The added tensor core computational power means that performing the DLSS algorithm happens faster, and as a result we have more time for the ALUs to do their job within the time allotted for rendering a frame (16 ms for 60 fps, 33 ms for 30 fps, etc.). In essence, improving the tensor core performance number gives a small boost to effective computational capacity of the ALU as well - but only if we use DLSS. The benefit of more tensor compute power seems useful, therefore. I'm almost certain that the computations performed by the tensor cores are embarrassingly parallel, so little to no parallel computing overhead should impact this advantage, either.
+ RAM is closer in terms of cycles : This is actually the one disadvantage Cerny mustered when discussing his choice for high frequencies rather than more CUs. What it means is that when memory in the GPU caches needs to be refreshed, the RAM must be addressed. If the GPU runs at higher frequencies, then when the caches must be refreshed, this takes a finite time which we can count in terms of cycles (real time multiplied by frequency). With a higher frequency, the number of cycles it takes to refresh is higher, even if the actual time is the same. So, why does it matter if the real time is the same? Well, because the higher frequency has fewer SMs, it must make the most of its sequential speed to balance its lower parallelisation capability. This means that each SM must do more in the same period of time. If the lower SM, higher frequency configuration is made to wait for the same amount of real time as the higher SM, lower frequency configuration, this means in practice that it wastes more cycles, whereas it needs to maintain a steady advantage in active cycles to balance its deficiency in SM count. This is therefore an advantage for the 20% higher SM count configuration.
+/- More RT (ray tracing) cores : Ah, the holy grail of the next gen. I think it's important for Switch 2 to support ray tracing to a decent degree, because games during the next gen will be using it more and more, and I am doubtful that devs can just swap out a lighting system that uses ray tracing for a system based on rasterisation without a substantial redesign, which would mean that porting would be straight up expensive if Switch 2 doesn't support ray tracing. Again, I checked the RT core count against the SM count in RTX cards, and it seems that 1 RT cores is included in each SM. Therefore, we once again have a 20% increase in RT compute power, whereas a frequency boost of 10% would only give a 10% boost to RT compute power. I am uncertain whether this is also embarrassingly parallel, though: RT cores evaluate the BVH structure, which requires accesses to memory (at least, I would think so) and concurrent access to the same resource. Therefore, I am unsure whether the raw RT compute power is practically achievable. In fact, there may be an argument to be made that fewer RT cores at a higher frequency might perform as well or better because of fewer memory accesses - but I don't know that with any degree of certainty.
- Lower cache bandwidth : Because the frequency of the GPU is lower, the caches also act at a lower frequency (and therefore at a lower bandwidth), which might cause slight delays in delivering data to the CUDA cores from the caches.
(?) - Slower rasterisation : Cerny quoted this as a disadvantage of lower frequencies, but I'm not completely sure why tbh. Perhaps someone else can explain this?
- More die space : With more SMs comes more die space, since the SM size does not depend on its frequency (duh). Because Switch 2 has a limited form factor, die space is a premium. A 20% higher SM count means that the GPU takes up 20% more space.
- More expensive : The more silicon you need, the more expensive the chip. A bigger GPU is therefore more expensive.
------------
There you have it: a higher SM count has a lot of advantages for ALU/Tensor core performance, and furthermore puts slightly less pressure on the memory bandwidth requirements. RT performance remains to be seen, depending on how parallelisable it is. The big disadvantage is the die space required.
So, I pass the question on to all of you: which one is the better choice for a practical Switch 2 design? 10% higher frequency or 20% higher SM count? To me, the question comes down to whether the die size is already bursting at the seams as is. If only a small amount of the die is occupied by the GPU, and there is space available for the GPU on a reasonably-sized die, then adding the extra SMs seems like a better idea.
20% higher SM count:
+ More ALU compute power than the frequency boost : The ALU compute power scales linearly with the SM count, and also linearly with frequency. As a result, the SM count increase adds 20% to the ALU compute power (this is the common TFLOPS number you always see thrown around), as opposed to a 10% increase with the frequency boost. This means more compute power for scene rendering and for post-processing (note that post-processing does not include DLSS!).
+ More tensor cores : I've checked on the tensor core and SM counts for the RTX cards, and each of them have a factor of 8 between them (# tensor cores = 8 * #SM). This leads me to believe that inside each SM 8 tensor cores are packed. As a consequence, increasing the SM count means increase the tensor core count by the same amount. This means that our tensor FLOPS count increase by 20%, whereas it would only increase by 10% in the case of a frequency boost. The added tensor core computational power means that performing the DLSS algorithm happens faster, and as a result we have more time for the ALUs to do their job within the time allotted for rendering a frame (16 ms for 60 fps, 33 ms for 30 fps, etc.). In essence, improving the tensor core performance number gives a small boost to effective computational capacity of the ALU as well - but only if we use DLSS. The benefit of more tensor compute power seems useful, therefore. I'm almost certain that the computations performed by the tensor cores are embarrassingly parallel, so little to no parallel computing overhead should impact this advantage, either.
+ RAM is closer in terms of cycles : This is actually the one disadvantage Cerny mustered when discussing his choice for high frequencies rather than more CUs. What it means is that when memory in the GPU caches needs to be refreshed, the RAM must be addressed. If the GPU runs at higher frequencies, then when the caches must be refreshed, this takes a finite time which we can count in terms of cycles (real time multiplied by frequency). With a higher frequency, the number of cycles it takes to refresh is higher, even if the actual time is the same. So, why does it matter if the real time is the same? Well, because the higher frequency has fewer SMs, it must make the most of its sequential speed to balance its lower parallelisation capability. This means that each SM must do more in the same period of time. If the lower SM, higher frequency configuration is made to wait for the same amount of real time as the higher SM, lower frequency configuration, this means in practice that it wastes more cycles, whereas it needs to maintain a steady advantage in active cycles to balance its deficiency in SM count. This is therefore an advantage for the 20% higher SM count configuration.
+/- More RT (ray tracing) cores : Ah, the holy grail of the next gen. I think it's important for Switch 2 to support ray tracing to a decent degree, because games during the next gen will be using it more and more, and I am doubtful that devs can just swap out a lighting system that uses ray tracing for a system based on rasterisation without a substantial redesign, which would mean that porting would be straight up expensive if Switch 2 doesn't support ray tracing. Again, I checked the RT core count against the SM count in RTX cards, and it seems that 1 RT cores is included in each SM. Therefore, we once again have a 20% increase in RT compute power, whereas a frequency boost of 10% would only give a 10% boost to RT compute power. I am uncertain whether this is also embarrassingly parallel, though: RT cores evaluate the BVH structure, which requires accesses to memory (at least, I would think so) and concurrent access to the same resource. Therefore, I am unsure whether the raw RT compute power is practically achievable. In fact, there may be an argument to be made that fewer RT cores at a higher frequency might perform as well or better because of fewer memory accesses - but I don't know that with any degree of certainty.
- Lower cache bandwidth : Because the frequency of the GPU is lower, the caches also act at a lower frequency (and therefore at a lower bandwidth), which might cause slight delays in delivering data to the CUDA cores from the caches.
(?) - Slower rasterisation : Cerny quoted this as a disadvantage of lower frequencies, but I'm not completely sure why tbh. Perhaps someone else can explain this?
- More die space : With more SMs comes more die space, since the SM size does not depend on its frequency (duh). Because Switch 2 has a limited form factor, die space is a premium. A 20% higher SM count means that the GPU takes up 20% more space.
- More expensive : The more silicon you need, the more expensive the chip. A bigger GPU is therefore more expensive.
------------
There you have it: a higher SM count has a lot of advantages for ALU/Tensor core performance, and furthermore puts slightly less pressure on the memory bandwidth requirements. RT performance remains to be seen, depending on how parallelisable it is. The big disadvantage is the die space required.
So, I pass the question on to all of you: which one is the better choice for a practical Switch 2 design? 10% higher frequency or 20% higher SM count? To me, the question comes down to whether the die size is already bursting at the seams as is. If only a small amount of the die is occupied by the GPU, and there is space available for the GPU on a reasonably-sized die, then adding the extra SMs seems like a better idea.
Wider design is a better option all around, for Switch or anything else.
when you're talking about power consumption as strict as a mobile device, wider is better every time. there's only so large a fan you can stick on the thing and even that sucks battery. and more die space, while more costly, also gives you more surface area to dissipate heat, meaning less need of a fan. as mentioned higher up on this page, the Ipad and Surface Pro are laptop-class devices (so more power usage) but has no fan because of all the surface area that a gaming device won't have
Digital Foundry made a dlss 2 video and it delivered a perfect 1080 output from 540 internal resolution. I think switch 2 with Dlss is a must. Maybe that way they can get next gen ports.
I still think a Pro (Plus) is happening, maybe as soon as next year. Nintendo also said they wanted Switch to have a longer cycle than usual so the actual successor could end up 2024 or later.
Looking at their track record:
Game Boy: April 1989
Game Boy Color: October 1998 (9.5 years later)
Nintendo DS: November 2004
Nintendo DSi: November 2008 (4 years later)
Nintendo 3DS: February 2011
New Nintendo 3DS: October 2014 (3.5 years later)
Looking at their track record:
Game Boy: April 1989
Game Boy Color: October 1998 (9.5 years later)
Nintendo DS: November 2004
Nintendo DSi: November 2008 (4 years later)
Nintendo 3DS: February 2011
New Nintendo 3DS: October 2014 (3.5 years later)
I think they may as well just ride it out until 2022 with the current Switch at this point.
A Pro was a neat idea, but really looking at it how games would use it given that a Switch 2 probably has to be on market by 2023 at the latest. 2020 is kind of a "washed" year as it were with COVID19 to boot anyway, so the current Switch really only has to make it through 2021.
They could maybe make a Switch XL with a larger screen that cuts off more of that fat bezel and has some moderate overclocking of the CPU/GPU, but I've shifted my thought process on this and I don't think a Pro is needed.
Developers are shifting their focus to the next-generation now, Nintendo is better off just giving them a true next-gen Switch successor to work with sooner than some half step that still won't be good enough to run next-gen games. You need a full generation leap forward + DLSS integration.
The sooner Switch 2 dev kits are in dev hands, the sooner devs can start making games for it and the sooner the system can get in on next-gen projects like a Monster Hunter World 2 or Witcher 4. The sooner devs can integrate a Switch 2 into their next-gen planning, the better. When you come in too close to mid-cycle I think it creates a problem for devs because their development pipeline is totally set up without you included, the more Nintendo can lessen this the better.
A Pro was a neat idea, but really looking at it how games would use it given that a Switch 2 probably has to be on market by 2023 at the latest. 2020 is kind of a "washed" year as it were with COVID19 to boot anyway, so the current Switch really only has to make it through 2021.
They could maybe make a Switch XL with a larger screen that cuts off more of that fat bezel and has some moderate overclocking of the CPU/GPU, but I've shifted my thought process on this and I don't think a Pro is needed.
Developers are shifting their focus to the next-generation now, Nintendo is better off just giving them a true next-gen Switch successor to work with sooner than some half step that still won't be good enough to run next-gen games. You need a full generation leap forward + DLSS integration.
The sooner Switch 2 dev kits are in dev hands, the sooner devs can start making games for it and the sooner the system can get in on next-gen projects like a Monster Hunter World 2 or Witcher 4. The sooner devs can integrate a Switch 2 into their next-gen planning, the better. When you come in too close to mid-cycle I think it creates a problem for devs because their development pipeline is totally set up without you included, the more Nintendo can lessen this the better.
Last edited:
This year does seem to be a wash as you say. Originally BotW2 was heavily rumored to be this year's holiday game. Covid19 has delayed software development and hardware manufacturing world wide now for months, it's not close to being over yet, and that delay is affecting next gen consoles, not just possibly their launch date, but also the software devs behind next generation games.
Developers who have been told to work from home will have already seen a 6 month delay to their work if the game is still over half a year off. This will increase cross gen's length, or the operational time of Switch current gen ports.
The economy will also suffer a depression, which is a bad time historically to launch new consoles, in the 7th gen this happened 2 to 3 years into the generation when the housing market crashed, but this time it is happening this holiday. If holiday shopping is affected by covid19 because of a lack of funds from not working all year, or worse case, it is still lingering around and vaccines have not happened or still rolling out, the market will see a huge contraction.
Next gen consoles will still likely sell out at launch, their quantity might be limited but there are plenty of people in the market with the disposal income to buy up 3 to 4 Million XSX and 3 to 4 Million PS5 at launch, however the impact to the holiday would be revealed in total dollars spent and the shock to the economy would slow any recovery, meaning the depression will last through next year too, with 2021's holiday being a big turn around vs 2020's, and leading to a more confident market for 2022.
Developers will have recovered quite a bit in 2022 as well, setting up 2023 as the year where cross gen ends, next gen really starts, and disposable income becomes more stable. If there is a 2023 Switch 2.0 launch, it also means there is devkits for a switch 2 with major 3rd parties in 2021 and widely available in 2022 while cross gen is winding down. 2023 also gives Nintendo a chance to launch with a new Zelda, as plans changed and while a botw2 is close to being finished right now, the team should have already begun some plans on a next gen Zelda, which will solidify next year as development starts.
A 2024 launch isn't out of the question, especially if there is a next generation console delay until 2021, in a normal year, 2 years after next gen has launched is a good time, but considering everything around covid19 and this year being the launch year for next gen consoles, I think it has pushed everyone's plans for the next 18 months back, which should make 2023 the ideal year for a next generation Switch, it also happens to be the year Nintendo would likely use 5nm, and because of the delay to next gen consoles, 2024 is probably the year they would look at releasing something like a PS5 Pro or Slim.
Last edited:
While Nintendo has strong IP no doubt I think that roster becomes thin at a point, I don't think they can make it through all of 2023 with a down turn in sales at some point.
They've already released all the A-tier franchises for Switch already now that Animal Crossing is out, so likely no other game they can release going forward can give them the same boost going forward. Even BOTW2 will likely mainly be selling to people who already bought BOTW1.
I think BOTW2 probably does arrive this holiday, Nintendo has not been shut down
I think they launch Switch 2 in fall 2022 with Mario Kart 9 ... that is the perfect time for Mario Kart 9, but Switch 1 still receives new games through 2023. 3DS was still getting new games after Switch launched, no reason to not continue that, and Switch 2 should be able to run all those games.
2020: BOTW2, Bayonetta 3, Great Detective Pikachu 2, Mario All-Stars 3D
2021: New 3D Mario, DKC, Ring Fit 2, Metroid Prime 4, Mario Party Next
2022: New 2D Mario for both Switch 1/2, New Pokemon for both Switch 1/2. Switch 2 fall launch w/Mario Kart 9 + Splatoon 3.
Is somewhat along the lines of where I think they could go.
Last edited:
Nintendo isn't releasing all those Mario games, a new paper Mario, and a new Zelda all in from June-December. Zelda is 100% next year
Paper Mario is not really a system seller, so their big holiday game would be ... a collection of old Mario games?
Don't see that working out too well for them. 3D All-Stars is a nice support type of title but it's not the kind of project you bank a 50+ million selling console's important fourth holiday season on.
I don't see it happening, but we'll see. I'd be over the moon if it did
I mean the 4th holiday season is pretty important for any console but Nintendo in particular seems to emphasize it.
SNES - Donkey Kong Country (1994)
N64 - Donkey Kong 64 + Funtastic models (1999)
GameCube - Metroid Prime 2 (2004)
Wii U - Super Smash Bros. U (2014)
All pretty substantial holiday releases.
Switch - Mario 3D Collection (?)
Ehhhh ... that's not that a great way to maintain sales momentum. There needs to be a marquee holiday game this year, a platform only gets one prime period in its lifecycle you need to make hay with it.
SNES - Donkey Kong Country (1994)
N64 - Donkey Kong 64 + Funtastic models (1999)
GameCube - Metroid Prime 2 (2004)
Wii U - Super Smash Bros. U (2014)
All pretty substantial holiday releases.
Switch - Mario 3D Collection (?)
Ehhhh ... that's not that a great way to maintain sales momentum. There needs to be a marquee holiday game this year, a platform only gets one prime period in its lifecycle you need to make hay with it.
Uhh Wii U's 4th holiday season was 2015. Yoshi and Mario Maker I think? After Splatoon in summer, and Amiibo was at peak sales
EDIT: You're right, granted Wii U is not exactly the greatest success for Nintendo, lol. Mario Maker is still bigger than a 3D collection.
With no BOTW2, the year would be really mediocre post Animal Crossing
Xenoblade Remake
Mario 3D Collection
Paper Mario (?)
Maybe Bayonetta 3 (though if BOTW2 is delayed why wouldn't this?).
That's definitely one way to grind your sales momentum to a halt. The pressure is really on Nintendo always because they can't really go "well this year we'll just let Call of Duty and EA and NBA2K do the work" like Sony and MS can.
Xenoblade Remake
Mario 3D Collection
Paper Mario (?)
Maybe Bayonetta 3 (though if BOTW2 is delayed why wouldn't this?).
That's definitely one way to grind your sales momentum to a halt. The pressure is really on Nintendo always because they can't really go "well this year we'll just let Call of Duty and EA and NBA2K do the work" like Sony and MS can.
You're kidding yourself if you don't think a Mario collection would sell a shit ton.
The mario all stars 2 and new super Mario 3d world are coming this holiday, with paper Mario possibly ending the year in December. Nintendo has been dealing with stock issues for the last couple months, and they can treat this year like they did with Mario's 25th birthday on the Wii.
I know botw 2 sounds great for this year, but because of the virus, next year is probably the peak year for the switch and they also have the 35th anniversary for the legend of Zelda series to think about, just like they did With skyward sword.
If a new Mario doesn't come this year and isn't coming next year, they can release it after the Mario movie in summer 2022, so whatever comes after odyssey is launching to close out Nintendo Switch's last big holiday, that just leaves the first half of 2023, which could see another Yoshi and maybe Xenoblade X, there isn't a ton of big software needed in the launch year of another platform, because that platform needs the spot light.
The Mario team is probably remastering the original games for this year, adding levels to 3d world and then finishing off the system with a big new Mario game in 2022. Mario Kart 9 is definitely something they could launch this year, and have Mario Kart 10 ready for Switch 2 in 2023. This is what the rumor going around was from a couple months ago.
A Pro model requires an silicon production process update. They used the latest available one to improve the battery.
Looking back at the PS4/PS4 Pro I thing the Pro version was pointless. Same games with slight improvements for the Pro version that could have achieved using the resources that were used to do 2 "versions".
95% of the Switch games are fine with the available power. The games that could benefit most would be even better suited for a Switch 2.
i have given up on the performance Pro model, but think a premium model with a better screen, maybe metal enclosure could come before the Switch 2. But Corona might have an impact on any cosmetic Pro model.
7nm with rtx... Turing cores in the dock for DLSS2.0 when displaying to a TV.... Could be really good.
5nm seems more likely since mass production starts this year
But would Nvidia be ready? They haven't released a 7nm part yet.
by 2022/2023? yea they would be. an Ampere tegra might even be the only form of Ampere on 5nm as they'll probably move to Hopper
It would, as in so is a game like NSMBU, but it's not really the type of game that should be the marquee holiday game for a high selling system in probably its peak (or 2nd year peak) year. Not if you want to maintain sales momentum anyway.
I agree in a normal year, but there were rumors about a Mario Kart game this year, which could be the big holiday game.
Your argument about botw2 is that just botw people would buy it, but it should be the marquee title? I think the 35th anniversary of Mario is a bigger deal for Nintendo this year, and Zelda is a bigger deal next year with its own botw2 and 35th anniversary collection, which is already almost done as they would just need to update skyward sword and make sure the textures are good for ocarina and Majora's Mask from 3DS, as they created those in HD anyways.
Also metroid could be the same thing for 2021, trilogy is already done or so the rumor goes, metroid other m is the only 3d metroid without a HD upgrade, which could be happening via the metroid team.
Putting out that monolith fantasy adventure rpg next year is also probably happening. Pikmin 4 and the next big Mario game for 2022 would work well too.
A Pro model wouldn't only entail a process change like the TX1+ models did, it'd be an actual architecture revision. And unlike PS4Pro it'd likely have some exclusive software. I don't see much standing in the way of it honestly.
Mario Kart 8 DX is selling way too well, Nintendo knows that will cover MK players this gen.
I think knowing that they've ear marked MK9 and are holding onto that on purpose as a Switch 2 day 1 release. Probably smart.
I think Nintendogs 2 which was rumoured for years was kinda the same deal. Nintendo held onto that purposely until 3DS launched because they knew the DS didn't need it, the original game just kept selling and selling. Now unfortunately for them in that scenario, the casual gaming market had really started to fall apart by 2011 because of the smartphone and they got caught with their pants down as a result for the 3DS launch. They really banked hard on Nintendogs + cats being able to carry the system and they saw it was a massive seller on DS.
Pikmin, Metroid ... just aren't big enough sellers.
- Status