• Hey everyone, staff have documented a list of banned content and subject matter that we feel are not consistent with site values, and don't make sense to host discussion of on Famiboards. This list (and the relevant reasoning per item) is viewable here.

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

Remember when nobody Nvidia said there would be a certain amount of games with rtx at launch and there was only like two💀
I do. It was expected that there would be more in 2023, but we can’t understate the impact of the pandemic on development across the industry. Those games will still come during the course of the generation, so, it’s important that this successor is well-placed to receive ports and become the lead platform as much as possible.
 
0
Small button on below left on new JoyCon integrate + & - functions likes menu & map
So then it just can't play current Switch games? 🤣

Because those buttons do distinctly different things on Switch games, you can't just make a single button and go "well it does both!" I'm playing Origami King right now and + and - pull up very different things. To say nothing of.. like, most other games as well. I wonder if the person who designed this thought that through. 😅
 
Ray tracing is a lock-in for the successor, not something we’re saying hopefully. It was revealed in the leaks, but truthfully, this was pretty much known four years ago. To believe it’s “hopeful” requires you to believe that Nvidia would speak of its importance, then proceed to omit this feature from future hardware for one of their biggest partners, having tried to secure a deal with them well before the Switch. It’s hard to reconcile those points.

AMD had stolen a march on them with PS/XBox, THEN Samsung/AMD brought the first RT phone GPU to a retail product when the Xclipse 920 was in the Exynos 2200 SoC for the S22 line of phones, and there is also Steam Deck in the portable arena, and more recently, ROG Ally. So, let’s put to bed the idea that RT is this precious thing which only PS/XBox and high-end PCs will do now, and which Nintendo will follow in a generation’s time, playing catch-up with the industry. They might well have the best RT on consoles, considering the consensus view is that Nvidia’s RT is better than AMD’s RT. There is something called the “Common Endeavour”, or the “Great Determination” - Ultimately, one has to ask, do you believe that Nintendo aligned themselves with the leading graphics processing company in the world, only to make backwards-thinking, unambitious hardware? That would be a colossal failure to learn the lessons of the 3DS, and leave them susceptible to a worse-than-Wii U fate. Do you also believe that Nvidia courted them for years, only to enter a long-term partnership which involved creating hardware that left them behind? Surely, as the leading graphics processing company, they would want to act the part and uphold their reputation?

It’s a myth that Nintendo is somehow averse to the idea of higher performance - In fact, from a developer’s standpoint, it’s very much in their interests to have higher performance. It means they can spend less time on nip-and-tuck solutions to mask hardware limitations, AND increase their productivity - Something they’ll need to do because the well of past-gen ports WILL run dry, won’t grow future sales in meaningful ways, and the prolific software momentum has to be sustained as much as possible. For Nvidia, it’s getting DLSS tech into 140m+ pairs of hands, making their RT the first choice for developers, and growth in the low-power consumption, high performance arena. Those points combined form the Common Endeavour/Great Determination.

But please, don’t take my word for it - I mean, I was confident about this as far back as 2019. Nvidia stated that "their leadership had turned ray tracing into a standard", and announced partnerships with Adobe, Autodesk, Unity, Epic, EA, and Microsoft, among others - Seeing as this list includes Nintendo partners, too, a thing called "Reasonable Deduction" told me that it will surely be a part of their next system. Later that year, Morgan McGuire of Nvidia predicted that the first game to require it would ship in 2023, and every gaming platform would offer accelerated ray tracing by then - There is a pandemic, so, the timing could be a little off, BUT we are still talking about the same time window for a prospective successor to the Switch (i.e., from 2024 onwards). This reinforces a past point I made about ambition, and why I continue to push back against the narrative about "curbed expectations" in the Nintendo community. It's not misty-eyed fanboyism or dreamy fanfic hope, it's reasonable deduction and receipts. Finally, one has to reconcile all of this with the "Do Or Die" rhetoric of the leadership, because serious questions would need asking if they didn't deliver on these industry requirements. It would represent collective internal failures for Nintendo, Nvidia, and their partners.
I don't get it, I thought current Switch horsepower was perfectly adequate, PS5 Pro is a waste of resources, more powerful consoles are falsely advertised, "fake powerhouses" etc. yet now RT on Switch 2 is important, guaranteed etc.

It's almost as if horsepower is a bad thing on other consoles but good on Nintendo consoles.

They might well have the best RT on consoles, considering the consensus view is that Nvidia’s RT is better than AMD’s RT

Of course, I can imagine that little Switch 2 will have the best RT on consoles, the enormously more powerful PS5 Pro which will possibly release later won't have a chance because AMD are stinky-bums.
 
I wonder if Nintendo have the balls to be this whimsy anymore

1fb.gif


... Something tells me ... they have. They have.

And they're bigger than ever before.
 

Just found this tweet earlier and I wanted to get your thoughts on it. Honestly, Nintendo Switch Next does sound like what Nintendo might go with instead of Switch 2, though Super Switch sounds better imo. I don’t think they’ll go with a 120 Hz screen, they might stick with a 720p 60 Hz OLED screen for handheld then have it be upscaled to 1440p or 4K when docked. The extra power that this console would have (assuming it’s somewhere between a base PS4 and a PS4 Pro) should allow most games to run at a solid 60 fps, especially 1st party games. Some more demanding 3rd party games might be limited to 30 fps, but that’ll be fine as long as it’s stable. Other things I want to see are an option for different background themes, better online functionality, more systems added to Switch online (DS seems most likely to come next), better battery life (if possible considering how powerful it’ll be), and above all else, backwards compatibility with current Switch games. I’m thinking we’ll see an announcement before the end of this year (either at Gamescom, Tokyo Games Show, or in October like what they did with the OG Switch in 2016) and it’ll be released in the first half of 2024 (maybe March like OG Switch).

Here comes the Nintendo Switcherland
 
I'll be real with both you and @Hermii, cutting out all those A53 CPU cores amounts to far more than pennies, silicon wafers ain't cheap and the SoC is almost always the single most expensive component in most handheld consumer electronics, with RAM a close second. This is the whole reason companies make chips on a smaller process node over time, you either save a LOT of money getting more chips produced per wafer or can use the same amount of the silicon wafer to get a more performant chip. So reducing the die size by taking unused components out would make sense under the same cost-saving principle. Even going from 20nm to 16nm with the TX1 could likely have saved Nintendo a low double-digit number of US dollars in terms of SoC cost.

However, while cutting out the A53s would be preferable, I don't think they actually did, because it would still be cheaper (and, just as importantly, definitely faster) to just shrink the current design as is than fully redesign the die without the unused A53s. The ideal time to cut the A53s out of the design would have been before the 20nm Erista version was taped out and sent to production.
This doesn’t really have anything to do with my comment. I’m talking about the “Pennies saved is what they go for” mentality, and to be blunt they aren’t going to bother saving that if it’s a measly pennies especially at the scale they operate at and their worth of being a multi billion dollar company. They are a Large-Cap after all.

If it was like say…. those that make the GPDwins or the AyaNeos, they would be more keen on making a profit (they already have high PM). But for Nintendo across a span of 7 years? 500-600k isn’t really much for their scale.
 
This doesn’t really have anything to do with my comment. I’m talking about the “Pennies saved is what they go for” mentality, and to be blunt they aren’t going to bother saving that if it’s a measly pennies especially at the scale they operate at and their worth of being a multi billion dollar company. They are a Large-Cap after all.

If it was like say…. those that make the GPDwins or the AyaNeos, they would be more keen on making a profit (they already have high PM). But for Nintendo across a span of 7 years? 500-600k isn’t really much for their scale.
I don't know how much silicon or how much price the OFA or a53s for that matter would add. But the fact is that every little thing adds up when you're selling the same very purpose built product over a 7+'year period.
 
This doesn’t really have anything to do with my comment. I’m talking about the “Pennies saved is what they go for” mentality, and to be blunt they aren’t going to bother saving that if it’s a measly pennies especially at the scale they operate at and their worth of being a multi billion dollar company. They are a Large-Cap after all.

If it was like say…. those that make the GPDwins or the AyaNeos, they would be more keen on making a profit (they already have high PM). But for Nintendo across a span of 7 years? 500-600k isn’t really much for their scale.
How much was saved with the Gamecube Digital AV Out port removal?
 
I think we're starting to see the impact of the significant drop in prices of flash and RAM over the past year. As a point of comparison, last year's version of the tablet (which used the older Snapdragon 870), reportedly had the following two configurations:

8 GB RAM + 128 GB UFS: 2,699 yuan
12 GB RAM + 256 GB UFS: 2,999 yuan

The new tablet has three configurations:

12 GB RAM + 256 GB UFS: 2,399 yuan
12 GB RAM + 512 GB UFS: 2,599 yuan
16 GB RAM + 512 GB UFS: 2,799 yuan

The 12GB RAM & 512 GB UFS model is actually cheaper than last year's 8GB RAM & 128GB UFS model, and the 16GB RAM & 512 GB UFS model is only slightly more expensive than it, which are pretty big jumps in RAM and flash for a single year.
So whats the probability of a 16gb setup? also does it even make sense to use that much?
 
So whats the probability of a 16gb setup? also does it even make sense to use that much?
It's feasible. Likely? Don't think so. More ram is always good but you also need the transfer speed to fill it. For a handheld with limited cache, there's a practical limit and I think 16GB is just on the cusp of breaching that
 
This doesn’t really have anything to do with my comment. I’m talking about the “Pennies saved is what they go for” mentality, and to be blunt they aren’t going to bother saving that if it’s a measly pennies especially at the scale they operate at and their worth of being a multi billion dollar company. They are a Large-Cap after all.

If it was like say…. those that make the GPDwins or the AyaNeos, they would be more keen on making a profit (they already have high PM). But for Nintendo across a span of 7 years? 500-600k isn’t really much for their scale.
Except they don't trim just pennies from a single component, is the thing. They're often trimming pennies across multiple components, every chance they can get, because that's what capitalist production efficiency is all about. 10 pennies might not mean much, but hundreds of pennies across the entire product is meaningful. Hermii just happened to bring up a single component is all.

But in the case of discussing the SoC, it's irrelevant because it's not the best jumping-off point for such a discussion, as you're often actually trimming whole dollars when you more efficiently produce an SoC, which was my point.
 
Does it matter if the Switch 2 will have 8 or 12 GB of RAM? I mean, obviously 12 GB will be better no doubt and i'd prefer it just like everyone else would, but realistically speaking, are 8 GB not enough anymore for PS4 level visuals these days? This decade old system still got some AAA games this year somehow and they still hold up better than they should (see RE4 Remake), especially considering how outdated the hardware is by now.

I might be misremembering but i recall reading about the PS4 when it released that about 3 of its 8 GB of RAM were reserved for the OS, meaning only 5 GB were actually usable for games. If the Switch 2's OS will only use up 1 GB, that will still leave it with more RAM than even the PS4 Pro.
And by looking at PS4 Pro versions, some might recall that a lot of games only managed to hit 1296p/1440p native or 1800p/4K checkerboarded due to the PS4 Pro still having only 8 GB of RAM. It only got a larger GPU with some overclocks as well as a small overclock to the CPU, but that's about it. The Switch 2 on the other hand could have upwards of 2 GB more RAM reserved for games than the PS4 Pro simply by using a more optimized and limited OS. (In terms of functions & features) This should be enough to get its games DLSS'ed up from 1080p to 4K with an equal level of detail for game assets.

Given how well Nintendo optimizes their 1st party games, 8 GB will be plenty for their dev teams and they'll make sure to create even better looking games than what we already got on PS4 / PS4 Pro.

Feel free to let me know what you think of this analysis!
 
Does it matter if the Switch 2 will have 8 or 12 GB of RAM? I mean, obviously 12 GB will be better no doubt and i'd prefer it just like everyone else would, but realistically speaking, are 8 GB not enough anymore for PS4 level visuals these days? This decade old system still got some AAA games this year somehow and they still hold up better than they should (see RE4 Remake), especially considering how outdated the hardware is by now.

I might be misremembering but i recall reading about the PS4 when it released that about 3 of its 8 GB of RAM were reserved for the OS, meaning only 5 GB were actually usable for games. If the Switch 2's OS will only use up 1 GB, that will still leave it with more RAM than even the PS4 Pro.
And by looking at PS4 Pro versions, some might recall that a lot of games only managed to hit 1296p/1440p native or 1800p/4K checkerboarded due to the PS4 Pro still having only 8 GB of RAM. It only got a larger GPU with some overclocks as well as a small overclock to the CPU, but that's about it. The Switch 2 on the other hand could have upwards of 2 GB more RAM reserved for games than the PS4 Pro simply by using a more optimized and limited OS. (In terms of functions & features) This should be enough to get its games DLSS'ed up from 1080p to 4K with an equal level of detail for game assets.

Given how well Nintendo optimizes their 1st party games, 8 GB will be plenty for their dev teams and they'll make sure to create even better looking games than what we already got on PS4 / PS4 Pro.

Feel free to let me know what you think of this analysis!
the ps4 is also 10 years ago. Drake won't be just be getting PS4 games (which would eventually end) but the occasional PS5/Series game as well. those are made with a different design language thanks to the hardware. Drake has to keep up with them, not be on par with 10 year old consoles.

in the recent Q&A to shareholders, Takahashi talked about how Nintendo's own developers want more hardware resources. when adding 12GB of ram doesn't add to costs thanks to tumbling dram prices, why not add more that your developers want?
 
I don't get it, I thought current Switch horsepower was perfectly adequate, PS5 Pro is a waste of resources, more powerful consoles are falsely advertised, "fake powerhouses" etc. yet now RT on Switch 2 is important, guaranteed etc.

It's almost as if horsepower is a bad thing on other consoles but good on Nintendo consoles.
It's not a matter of importance. It's just a matter of not being a chip designed a decade ago.
If it was like say…. those that make the GPDwins or the AyaNeos, they would be more keen on making a profit (they already have high PM). But for Nintendo across a span of 7 years? 500-600k isn’t really much for their scale.
Where are you getting this 500-600K from again? You were talking about like 50 cents a unit, and that would be $60+ million for Switch so far.
 
Ray tracing is a lock-in for the successor, not something we’re saying hopefully. It was revealed in the leaks, but truthfully, this was pretty much known four years ago. To believe it’s “hopeful” requires you to believe that Nvidia would speak of its importance, then proceed to omit this feature from future hardware for one of their biggest partners, having tried to secure a deal with them well before the Switch. It’s hard to reconcile those points.

AMD had stolen a march on them with PS/XBox, THEN Samsung/AMD brought the first RT phone GPU to a retail product when the Xclipse 920 was in the Exynos 2200 SoC for the S22 line of phones, and there is also Steam Deck in the portable arena, and more recently, ROG Ally. So, let’s put to bed the idea that RT is this precious thing which only PS/XBox and high-end PCs will do now, and which Nintendo will follow in a generation’s time, playing catch-up with the industry. They might well have the best RT on consoles, considering the consensus view is that Nvidia’s RT is better than AMD’s RT. There is something called the “Common Endeavour”, or the “Great Determination” - Ultimately, one has to ask, do you believe that Nintendo aligned themselves with the leading graphics processing company in the world, only to make backwards-thinking, unambitious hardware? That would be a colossal failure to learn the lessons of the 3DS, and leave them susceptible to a worse-than-Wii U fate. Do you also believe that Nvidia courted them for years, only to enter a long-term partnership which involved creating hardware that left them behind? Surely, as the leading graphics processing company, they would want to act the part and uphold their reputation?

It’s a myth that Nintendo is somehow averse to the idea of higher performance - In fact, from a developer’s standpoint, it’s very much in their interests to have higher performance. It means they can spend less time on nip-and-tuck solutions to mask hardware limitations, AND increase their productivity - Something they’ll need to do because the well of past-gen ports WILL run dry, won’t grow future sales in meaningful ways, and the prolific software momentum has to be sustained as much as possible. For Nvidia, it’s getting DLSS tech into 140m+ pairs of hands, making their RT the first choice for developers, and growth in the low-power consumption, high performance arena. Those points combined form the Common Endeavour/Great Determination.

But please, don’t take my word for it - I mean, I was confident about this as far back as 2019. Nvidia stated that "their leadership had turned ray tracing into a standard", and announced partnerships with Adobe, Autodesk, Unity, Epic, EA, and Microsoft, among others - Seeing as this list includes Nintendo partners, too, a thing called "Reasonable Deduction" told me that it will surely be a part of their next system. Later that year, Morgan McGuire of Nvidia predicted that the first game to require it would ship in 2023, and every gaming platform would offer accelerated ray tracing by then - There is a pandemic, so, the timing could be a little off, BUT we are still talking about the same time window for a prospective successor to the Switch (i.e., from 2024 onwards). This reinforces a past point I made about ambition, and why I continue to push back against the narrative about "curbed expectations" in the Nintendo community. It's not misty-eyed fanboyism or dreamy fanfic hope, it's reasonable deduction and receipts. Finally, one has to reconcile all of this with the "Do Or Die" rhetoric of the leadership, because serious questions would need asking if they didn't deliver on these industry requirements. It would represent collective internal failures for Nintendo, Nvidia, and their partners.
In 2019, Nvidia planned to have made and had Nintendo release a platform with accelerated raytracing.

In 2023 or earlier.

Interesting.
 
the first major game to require RT hardware was back in 2021.

in 2022, one of the most popular multiplayer games has mandated ray tracing on console (though not hardware accelerated)

in 2023, two games have released that has mandatory ray tracing on consoles, one will be mandatory on pc as well (with a no HWRT mode), and another by the same studio will release in 2024

I don't know about yall, but I'd say Nvidia was too conservative
 
Except they don't trim just pennies from a single component, is the thing.
But that literally has nothing to do with my post. You’re discussing something tangentially related, but that’s not really what I’m getting at.


Where are you getting this 500-600K from again? You were talking about like 50 cents a unit, and that would be $60+ million for Switch so far.
You’re right, I was thinking of it from the angle of 5c not 50c. I’m sorry. and I forgot a Zero


I don't know how much silicon or how much price the OFA or a53s for that matter would add. But the fact is that every little thing adds up when you're selling the same very purpose built product over a 7+'year period.
We know they aren’t doing that though.

Consoles aren’t always the same product, they get revised eventually.

and I did the math wrong but at their scale it’s not as significant still as what they normally spend in a year vs the entire lifecycle of the platform from just hardware



Like here’s a funny perspective: they BOM optimize their consoles… but then decide to go for expensive proprietary media to deliver their games.

So they are shifting how the money is spent.


They could just do a small standard console with a generic play style, but the R&D for their “gimmick” (eg unique feature) then raises the cost of said console, see Wii U gamepad.

It’s an interesting perspective to look at it, since people only focus on one element but ignore the other.
How much was saved with the Gamecube Digital AV Out port removal?
Idk, ask Nintendo. I don’t have the financials for that or the known cost of the part.

That said, sometimes when something is removed from a product, it’s because it served no purpose. It was a literal waste of money to keep it as is.

Sometimes it’s because it’s easier to actually make said product and they swap for something else that offers a similar experience, which makes getting their product to shelves more efficient, which makes getting revenue easier. Etc.
 
i kinda expect 8 GB of RAM for the Switch sucessor, the ideal would be 12/16GB of RAM, but to keep the console priced under $400 or lower, Nintendo will opt fot 8 GB of RAM on Switch sucessor.
market conditions would allow for 12GB or even 16GB to fit within that budget depending on when procurement happened

 
i kinda expect 8 GB of RAM for the Switch sucessor, the ideal would be 12/16GB of RAM, but to keep the console priced under $400 or lower, Nintendo will opt fot 8 GB of RAM on Switch sucessor.
Actually, Nintendo will use no RAM. That way they can save money.

Because, you know, Nintendo 🤭😆

please, no one take this obvious comment seriously
 
I mean, it's not possible that they special order it. Nintendo likes special ordering shit for memory. But given how they field inquiries from third parties and the declining price of dram, is it really a stretch that they would exceed 8GB?

They've used unusual memory types in the past, like 1T-SRAM in the GC/Wii or FCRAM in the 3DS, but in both of those cases they chose a niche memory type for specific performance reasons (lower latency), which is quite different than special ordering a lower capacity of a widely available type of memory.

Speaking of which, niche memory types don't really exist any more, which is understandable but a bit of a shame. Nowadays it's pretty obvious that Sony and MS are going to use whatever version of GDDR is available, and Nintendo will use LPDDR, but it used to be fun speculating about weird memory types back in the old days.

I'm particularly disappointed that nobody made a console using 3D XPoint (which Intel sold as Optane) before it went out of production. I don't think it necessarily would have been a good idea, but it definitely would have been an interesting one. Combine a small pool of very fast memory (say 4GB or 8GB of HBM) with a huge pool of 3D Xpoint (say 64GB or 128GB) and you've got a somewhat extreme take on the small pool of embedded RAM plus big pool of cheap DDR approach that the Wii U and Xbox One used. Of course there's a reason both Nintendo and Microsoft moved away from that approach, as managing multiple pools of memory is a pain, but it would have been interesting to see what developers would do if you took that and used 3D XPoint to push it to "fit the entire game in RAM" extremes.
 
i never saw the need of a Switch Pro, it would not be beneficial for Nintendo if they actually released this console, we saw this with New 3DS, very few game would take benefict for the Switch Pro, everytime i saw a rumor of the Switch Pro, i LOL and think, cant this insider and rumor mongers let this thing die and let us enjoy Switch, wait for 6/7 years and them
we have a true sucessor to Switch.

Another factor I think is what would’ve been the addition of more power profiles for developers having to mess with. It would’ve felt more like an Xbox Series S/X type situation.

While it is true some games benefit with an Overclocked Switch, the differences are less compared to say PS4 to PS4 Pro.


I’m going to be real with you.

Those pennies don’t mean anything at the scale they operate at or anything they would be worried about. If they need those pennies, at their size and scale, it’s because they are inefficient at using that money they have and have to save somewhere.


I mean like, assume they cut enough to where they save 50-60¢ per units, that’s “only” 500-600k. Across a whole console generation. They routinely spend >10x that for game dev and hardware R&D 😹

Like by March 2021, spending nearly 800Million for game dev, hardware, market research, etc that happened in the previous fiscal year total. Doesn’t account for the previous years. So, I’d be more sensible about that penny pinching description that’s given to them. It’s not that worth it to them.

But this is off topic I think, or it might be a derail I should say. Just wanted to nitpick at this. :p

I work with multi-million dollar contractors on a daily basis on job sites for a living. Large conglomerates absolutely do care about those pennies on the dollar.

Anything that’ll get them extra money for their bottom line, they’ll absolutely do it.

You drive a car? If so, then you know the same exact scenario occurs for automakers. If it’ll help them save a few cents per car sold, and not really affect the overall car (at least from their calculations), they’ll take it out.

As far as whether or not those cores were actually removed from the Mariko chip, I have no idea. At the minimum, like Erista, they were disabled permanently. That said, since Nintendo and Nvidia bothered so much with a die shrink in the first place, it would be a good time to remove other unnecessary silicon that has no purpose. Otherwise, it’s just wasted silicon at that point. Not a great use of precious space with precious silicon.

Also falls in line with Drake being very custom and bespoke. Vanilla Orin would’ve been too large, and with too much wasted silicon for stuff that had no purpose for game development.
 
i wonder if they’ll do this, switch 2 3rd party ports run natively but to still play them on switch here’s the cloud version
I think that's very possible, even for first party games.

Remember, Nintendo is in bed with Nvidia AND Google for their SOC and network technology respectively, they have more or less unlimited access to cloud infrastructure and choose not to use it. Nvidia would kill to have a customer that big on their GForce Cloud, and Google would absolutely love to have Nintendo pay even just the COST rate of Stadia blades.
 
A Switch Pro that just made the handheld perform like the docked console and had a 1080p OLED screen would have been a pretty interesting product ngl.

This seems like it could have been possible to make after the shift to 16nm. Maybe could have added HDR support for docked mode as well?

Adding an additional 4 GB stick of RAM seems like it would have been a lot more work for the engineers and developers, but probably could have allowed more games to run at 60 FPS. Not sure that would be easily possible compared to the other stuff I brought up though.

I think there was some path where you could make a clearly more powerful Switch that provided some immediate benefits without causing too many developer headaches and where you had a clear path to still offering a next-gen console.

But Nintendo didn't pursue that path and that's fine as well.
 
Last edited:
Another factor I think is what would’ve been the addition of more power profiles for developers having to mess with. It would’ve felt more like an Xbox Series S/X type situation.

While it is true some games benefit with an Overclocked Switch, the differences are less compared to say PS4 to PS4 Pro.




I work with multi-million dollar contractors on a daily basis on job sites for a living. Large conglomerates absolutely do care about those pennies on the dollar.

Anything that’ll get them extra money for their bottom line, they’ll absolutely do it.

You drive a car? If so, then you know the same exact scenario occurs for automakers. If it’ll help them save a few cents per car sold, and not really affect the overall car (at least from their calculations), they’ll take it out.

As far as whether or not those cores were actually removed from the Mariko chip, I have no idea. At the minimum, like Erista, they were disabled permanently. That said, since Nintendo and Nvidia bothered so much with a die shrink in the first place, it would be a good time to remove other unnecessary silicon that has no purpose. Otherwise, it’s just wasted silicon at that point. Not a great use of precious space with precious silicon.

Also falls in line with Drake being very custom and bespoke. Vanilla Orin would’ve been too large, and with too much wasted silicon for stuff that had no purpose for game development.
I suppose, but the comparison I’m getting at is that they’d be caring about dollars more than pennies at their large size and how much they routinely spend. 10-20cents vs 1-2 dollars or 10-20 dollars is a big gulf. They’d want to save more dollars.
 
They've used unusual memory types in the past, like 1T-SRAM in the GC/Wii or FCRAM in the 3DS, but in both of those cases they chose a niche memory type for specific performance reasons (lower latency), which is quite different than special ordering a lower capacity of a widely available type of memory.

Speaking of which, niche memory types don't really exist any more, which is understandable but a bit of a shame. Nowadays it's pretty obvious that Sony and MS are going to use whatever version of GDDR is available, and Nintendo will use LPDDR, but it used to be fun speculating about weird memory types back in the old days.

I'm particularly disappointed that nobody made a console using 3D XPoint (which Intel sold as Optane) before it went out of production. I don't think it necessarily would have been a good idea, but it definitely would have been an interesting one. Combine a small pool of very fast memory (say 4GB or 8GB of HBM) with a huge pool of 3D Xpoint (say 64GB or 128GB) and you've got a somewhat extreme take on the small pool of embedded RAM plus big pool of cheap DDR approach that the Wii U and Xbox One used. Of course there's a reason both Nintendo and Microsoft moved away from that approach, as managing multiple pools of memory is a pain, but it would have been interesting to see what developers would do if you took that and used 3D XPoint to push it to "fit the entire game in RAM" extremes.
Wouldn’t the modern day equivalent be the usage of cache or embedded memory to reduce bandwidth constraints?
 
If the OLED model was truly supposed to be the fabled "Switch Pro" (with an overclocked Mariko chip and the built-in 4K compatibility), then can we assume that Covid throttled its development and that Nintendo had to scrap these features in order to meet their 2021 deadline?

Its possible that they ultimately looked at what the 4K capabilities would be for this Mariko based Pro model and realized that they would be so limited that it wasn't worth implementing. Just from a marketing standpoint, would you really want to bring a new model to market with a main selling point of 4K output when practically no games would support it, and Nintendo has not been interested in promoting Switch as a streaming device, so 4K output for stuff like Netflix isn't really something they would have wanted to promote either. Ultimately it was best to save the 4K selling point for hardware that could actually do it.
 
Hopefully it's at least accurate to say that, in the likely worst case scenario, we're less than 12 months away from knowing about this thing. And the mystery will be over.

Unless you're #Team2025
 
i wonder if they’ll do this, switch 2 3rd party ports run natively but to still play them on switch here’s the cloud version
Very unlikely. Whatever extra sales they could extract that way would quickly be overwhelmed by the massive costs of running such a service. They also really want people upgrading to the new hardware.
 
In terms of new releases, what order do you think Nintendo’s main series’ will appear on Switch 2 next gen based on current release patterns?

Obviously just guesses and just for fun!

3D Mario (day 1)
Metroid Prime 4 (cross gen)
Pokemon (could be cross gen)
Mario Kart
Animal Crossing
Smash Bros
Splatoon 4
3D Zelda
2D Mario
 
In terms of new releases, what order do you think Nintendo’s main series’ will appear on Switch 2 next gen based on current release patterns?

Obviously just guesses and just for fun!

3D Mario (day 1)
Metroid Prime 4 (cross gen)
Pokemon (could be cross gen)
Mario Kart
Animal Crossing
Smash Bros
Splatoon 4
3D Zelda
2D Mario
I would swap Mario Kart and Animal Crossing, but aside from that, this is where I'm at.
 
In terms of new releases, what order do you think Nintendo’s main series’ will appear on Switch 2 next gen based on current release patterns?

Obviously just guesses and just for fun!

3D Mario (day 1)
Metroid Prime 4 (cross gen)
Pokemon (could be cross gen)
Mario Kart
Animal Crossing
Smash Bros
Splatoon 4
3D Zelda
2D Mario

Prime 4 will release before the Switch 2.

Pokemon Gen 10 will be November 2025, but the next Pokemon spinoff RPG will probably be November 2024 and be cross-gen.

The interesting question is about Mario Kart, Animal Crossing, Smash.

It's really hard to determine EPD resource allocation so I have no idea how far along the next Mario Kart or Animal Crossing are.

And I don't even know if Smash 6 is like... In development? Or even close to starting development? That's the toughest one to guess. Who would even be making Smash 6 right now?

When RingFit 2 happens is also a super interesting question.

I could see another 2D Mario by 2026 as well, whereas Open Air Zelda 3 won't be until 2029 at the earliest.
 
Wouldn’t the modern day equivalent be the usage of cache or embedded memory to reduce bandwidth constraints?

Well the specific issue that Wii U and Xbox One were designed to deal with is that there are two different types of things in memory which have very different requirements. Buffer objects don't take up very much space, but require lots of bandwidth, and then everything else doesn't require that much bandwidth but takes up a lot of space. The solution Nintendo and MS used was to have a small pool of dedicated fast memory for the buffers and then a big pool of cheap, slow memory for everything else. This wasn't a particularly new idea, either, as dedicated VRAM in older consoles was often intended as "framebuffer memory".

The issues with it were, as I mentioned, the difficulty of developing games to properly utilise the two memory pools, and also simple economics. Embedded RAM is really expensive, so it's difficult to justify compared to just spending the money on a faster main RAM pool. Chipworks told us that the Wii U MCM would have cost Nintendo around $100 a piece, which is the same as Sony and MS were paying for their PS4/XBO SoCs only a year later. Part of this was the use of an MCM with separate CPU and GPU dies, but a large part of it was the big pool of eDRAM on there.

The Wii U's 32MB eDRAM pool was at least reasonably sized for the machine's performance. The Xbox One also has 32MB (of SRAM), but was a much more powerful machine which was intended to target 1080p, meaning the embedded memory pool wasn't actually big enough to hold a full framebuffer, forcing developers to utilise tiling or other approaches. The fact that deferred rendering had become common didn't help either, as now there were extra intermediate buffers such as G-buffers which also want to be in the fast memory pool.

The biggest issue for the XBO, though, was that the economics of it just didn't work out for MS. The plan was that by using embedded SRAM and cheap DDR they could have a much larger main memory pool than Sony would get going with a single GDDR memory pool, with the expectation that they'd have 8GB vs 4GB on the PS4. Sony, however, increased the PS4's memory to 8GB very late on, and MS lost their capacity advantage. Without a big chunk of on-die SRAM, Sony also had space for a bigger GPU, and with the BoM of the two consoles pretty similar (once you take out Kinect), it was pretty clear that Sony's "just use a single big pool of fast RAM" was the way to go.

Which is pretty much where we are now, with a single big pool of fast RAM being the case for PS5, XBSS/X* and Switch NG. While Switch NG's LPDDR5 might not sound that fast next to the other consoles, it appears to be hitting a similar ratio of bandwidth to Gflops as desktop Ampere chips, which would definitely fit it in comfortably as a continuation of the PS4 approach rather than the Wii U/XBO memory model.

In theory you could sort of argue that having a big on-chip cache is similar, but it's going to cache everything, not just buffer objects, and it's still got a large cost associated with it, and diminishing returns as you increase the size. You also still need a fast memory pool behind it; I don't think Sony or MS could have added so much cache to PS5 or XBSX that they would have gotten away with using regular DDR instead of GDDR.

*Technically, Xbox Series S and X do have split pools, although it's all GDDR6. I think Microsoft must enjoy making memory management a hassle on their consoles.
 
Please read this new, consolidated staff post before posting.

Furthermore, according to this follow-up post, all off-topic chat will be moderated.
Last edited by a moderator:


Back
Top Bottom