• 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.
  • Do you have audio editing experience and want to help out with the Famiboards Discussion Club Podcast? If so, we're looking for help and would love to have you on the team! Just let us know in the Podcast Thread if you are interested!

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

But aren't those Tensor Cores presumably going to be busy doing DLSS, especially in docked mode?
I'm assuming DLSS isn't being used here. We don't know how much either of these tools cost on Drake, so it's best to evaluate them independently for now.

In theory they don't have to be done concurrently. You could run the upscaling during loading an store the textures in memory to used or upscale only some of the textures at any given time (which God of War Ragnarok does)
 
Is anyone expecting redesigned Joycons alongside the next console? Even an ergonomic change (I have to use a grip shell thing playing in handheld mode as it’s not a comfortable console to use for extended periods)
Yes, even if it's at a superficial design level, just to make the distinction between the old and new model clearer. However I doubt they'll change the size much, unless they surprise us all by moving away from it being a hybrid console
 
So they're going to GamesCom, remodelling their store, but have nothing to show for the second half of the year?

Seems suspicious. I can't think of anything that could explain their recent actions other than a new console they're waiting to reveal before showing the holiday games, even if they're cross-gen.

I don't think Pikmin 4 will affect the reveal. If anything, I think it's intentionally placed in July so that they have something to release between the June reveal and the release.
 
Is anyone expecting redesigned Joycons alongside the next console? Even an ergonomic change (I have to use a grip shell thing playing in handheld mode as it’s not a comfortable console to use for extended periods)

I'm expecting a new design for connecting it to the tablet.

I would love a total redesign (front facing buttons, shoulders, sticks, ergonomics (not that flat design)), but I'm not holding out hope for it.
 
I know many glossed over it but the most interesting thing from the ROG Ally spec reveal for me was the inclusion of UHS-II.

A low cost/low heat solution that the hardware file decompression engine could very well bridge the gap in storage speeds. This is right up Nintendo's wheelhouse.

I wouldn't be surprised if [TBA] had a little of
everything the ROG Ally has other than the screen and the SoC. All at a lower cost to consumers.
 
Last edited:
I'm expecting a new design for connecting it to the tablet.

I would love a total redesign (front facing buttons, shoulders, sticks, ergonomics (not that flat design)), but I'm not holding out hope for it.
Bah! Absolutely not! The Joy-Con Rail (both the "receptical" style on the console and the "edge connector" style on the Joy-Con controllers themselves.) OUGHT to stay completely compatible. Even if they improve how things attach, they won't lose compatibility.

No point in saying "you can use your old controllers and accessories!" when your new console can't attach to them or charge them.

I'd expect adjustments to be relatively minor. Maybe button placements and stick technology. Certain things have to give to maintain compatibility and their small size. The edge where the Joy-Con connects to the console can't thicken much, and thickening any part of it would require a redesigned Joy-Con Grip, which I don't think is extremely likely.

There are improvements they can make within these limitations, like offset buttons, pressure sensitive triggers and capacitive shoulder buttons (scroll wheels would take up a LOT of space, but dual capacitive touch surfaces on the shoulder buttons would achieve the same input.).
 
So they're going to GamesCom, remodelling their store, but have nothing to show for the second half of the year?

Seems suspicious. I can't think of anything that could explain their recent actions other than a new console they're waiting to reveal before showing the holiday games, even if they're cross-gen.

I don't think Pikmin 4 will affect the reveal. If anything, I think it's intentionally placed in July so that they have something to release between the June reveal and the release.

It’s not that they have nothing, it’s that it’s just as possible that they have more for Switch (1) this year and nothing more.

Being quiet before Tears of the Kingdom and booking events later in the year is far from suspicious. I’m assuming it’s the biggest game they’ve ever released. It’s a disservice to other games to bring them into focus when this game is right around the corner. We’ll see other H2 games soon enough, new hardware or not.

I know this is a pretty major change of tune from where I’ve stood in the past, but I’m trying my best to acknowledge that not everything is an indicator of new hardware.
 
Bah! Absolutely not! The Joy-Con Rail (both the "receptical" style on the console and the "edge connector" style on the Joy-Con controllers themselves.) OUGHT to stay completely compatible. Even if they improve how things attach, they won't lose compatibility.

No point in saying "you can use your old controllers and accessories!" when your new console can't attach to them or charge them.

I'd expect adjustments to be relatively minor. Maybe button placements and stick technology. Certain things have to give to maintain compatibility and their small size. The edge where the Joy-Con connects to the console can't thicken much, and thickening any part of it would require a redesigned Joy-Con Grip, which I don't think is extremely likely.

There are improvements they can make within these limitations, like offset buttons, pressure sensitive triggers and capacitive shoulder buttons (scroll wheels would take up a LOT of space, but dual capacitive touch surfaces on the shoulder buttons would achieve the same input.).
I agree with the retaining backward compatibility for hardware accessories. I just wish they would improve the materials used though. For example, the joycon has a very weak grip that loosen overtime where it wouldn't lock anymore with the switch body. I have to replace mine with a metallic part to fix it.
 
It’s not that they have nothing, it’s that it’s just as possible that they have more for Switch (1) this year and nothing more.

Being quiet before Tears of the Kingdom and booking events later in the year is far from suspicious. I’m assuming it’s the biggest game they’ve ever released. It’s a disservice to other games to bring them into focus when this game is right around the corner. We’ll see other H2 games soon enough, new hardware or not.

I know this is a pretty major change of tune from where I’ve stood in the past, but I’m trying my best to acknowledge that not everything is an indicator of new hardware.
Not every one thing, but everything all at once, I would say yes.

TOTK does not prevent them from revealing games. It just doesn't. It didn't stop Pikmin 4, or the many games initially planned to release after TOTK that now have released before it.

Delaying marketing for what could be half a dozen titles by months and months over one game doesn't make sense. It's just not good business. And they KNOW that, or they wouldn't have shown off Pikmin 4.
 
I agree with the retaining backward compatibility for hardware accessories. I just wish they would improve the materials used though. For example, the joycon has a very weak grip that loosen overtime where it wouldn't lock anymore with the switch body. I have to replace mine with a metallic part to fix it.
Absolutely, I think improvements to the Joy-Con Rail system are due. They did improve it a LOT with better tolerances and materials with the OLED Model, so I'd expect them to continue to improve it without breaking compatibility.

I'd like to see them rethink the materials a bit, like using the plastic that Pro Controller Grips are made of; they're grippy but far less delicate than Joy-Con shells.

That said one problem Joy-Con can't avoid due to their tiny size and half a dozen PCBs (other than reducing internal complexity, which I think they will do to some extent, and have already been doing.), is their shells are extremely thin, and few materials are suitable for it. I hope they go with something that's strong and wears well.

As for the Joy-Con latch you replaced with metal; the reason they're plastic is so that if you drop the console, the Joy-Con absorbs the fall, and rather than damaging the console, the console slices off the plastic latch and lets the Joy-Con take the fall. If they came pre-installed with metal latches, a drop could permanently damage the console.

There are downsides to this approach, Joy-Con ARE more delicate unfortunately because they are DESIGNED to be delicate because they're meant to be the weakest link so they break before the console. If they improve both durability wise, I think a lot of these problems can be overcome. For instance, they could use a pre-installed metal latch that has a slight slope at the top edge, so if the force pushing it up is strong enough to cause damage, it becomes unlatched, rather than just getting sliced off by the console.
 
Delaying marketing for what could be half a dozen titles by months and months over one game doesn't make sense. It's just not good business. And they KNOW that, or they wouldn't have shown off Pikmin 4.
I mean this is the same company that revealed a Mario collection and released it days later. And it sold very well.

Nothing is a sure thing. They always have their reasons.
 
kopite7kimi mentioned that Nvidia won't use TSMC's 3 nm** process node for fabricating Blackwell GPUs. So TSMC's 3 nm** process node isn't guaranteed.

Of course, there was an article from the Korea Economic Daily mentioning that Nvidia's going to use Samsung's 3 nm** process node. But no one knows which products are going to be fabricated using Samsung's 3 nm** process node.

** → a marketing nomenclature used by all foundry companies
What does the 5XXX series have to do with Switch 2 though? And considering Switch /T239 2 is really customizable versus the base Orion (8nm Samsung vs 5nm Samsung/4 or 5nm TSMC), I say anything could go.

But you know what... a newer node is one thing, but Changing foundries for Orion is a big deal and kinda hard for me to believe to think we would originally go from 8nm Samsung to 4-5nm TSMC. 5LPP or 4LPP Samsung and 3nm Samsung modes might just be more likely for Switch 2 and its revision, than tsmc 4/5nm. Hmm
 
What does the 5XXX series have to do with Switch 2 though? And considering Switch /T239 2 is really customizable versus the base Orion (8nm Samsung vs 5nm Samsung/4 or 5nm TSMC), I say anything could go.

But you know what... a newer node is one thing, but Changing foundries for Orion is a big deal and kinda hard for me to believe to think we would originally go from 8nm Samsung to 4-5nm TSMC. 5LPP or 4LPP Samsung and 3nm Samsung modes might just be more likely for Switch 2 and its revision, than tsmc 4/5nm. Hmm
I don't know, it's really customised (and a whole CUDA version above 30XX GPU cards), I could see it being on Lovelace's 5nm, easily. In fact it was sampled and tested at nearly the same time, with finalised versions coming around September last year. I'd almost say it's hard for me to believe that was a coincidence. Nvidia has experience moving Ampere-like architecture to 5nm, that's more or less what Lovelace IS. It's not unreasonable to think they could have done it for a valuable part where power efficiency is important.

I'm not holding my breath on 4N, but I wouldn't rule it out!
 
0
So they're going to GamesCom, remodelling their store, but have nothing to show for the second half of the year?

Seems suspicious. I can't think of anything that could explain their recent actions other than a new console they're waiting to reveal before showing the holiday games, even if they're cross-gen.

I don't think Pikmin 4 will affect the reveal. If anything, I think it's intentionally placed in July so that they have something to release between the June reveal and the release.
No one will argue that Nintendo will have releases also in the second half of the year. This does not mean at all, that it has to be a new console. Nintendo was ALWAYS a supporter of the GamesCom, they just skipped last year because of the unclear situation (it was a chaos, could have still go easily completely south), so I can understand why Nintendo skipped it last year. This year they can safely plan with it, so no reason not to go, if they are regular patrons of the GamesCom.
And renovation of a store is also not the biggest news if its ahead a major release. IF the renovation will take longer as the TOTK release, we most likely talking about the big thing.
Is anyone expecting redesigned Joycons alongside the next console? Even an ergonomic change (I have to use a grip shell thing playing in handheld mode as it’s not a comfortable console to use for extended periods)
I personally don't expect groundbreaking changes, more an evolution as we had from NES to SNES controller. So better handling, detail improvements and visible different, but similar design. Not sure if they can fix the joycon drift at all, as they now offer a free repair no question asked finally world wide, I would hope so. In general, I expect the old ones to be compatible, but not always fully functional, because they are missing some features. For most fun games, they should be still fine.
 
Sure, but that was a slapdash rerelease collection.
Yeah though the supposed original plan was the show it at E3 months ahead before all the Covid stuff happened.

For all we know there could be some other factor that caused them the market a bit unusually this year. Could be a new console coming, could be their year is lighter first party wise, or a number of different other things.
 
What does the 5XXX series have to do with Switch 2 though? And considering Switch /T239 2 is really customizable versus the base Orion (8nm Samsung vs 5nm Samsung/4 or 5nm TSMC), I say anything could go.
By the time Nintendo releases a "revision refresh" that supports smaller process nodes (e.g. a 3 nm** process node) within 2 years, Blackwell GPUs are probably already out on the market.

Assuming Nintendo and Nvidia choose TSMC to fabricate Drake, if Nvidia doesn't use TSMC's 3 nm** process node to fabricate Blackwell GPUs, which seems likely going by what kopite7kimi said, then I don't think there's a likely chance Nintendo and Nvidia are going to use TSMC's 3 nm** process node to fabricate a die shrunk Drake.

** → a marketing nomenclature used by all foundry companies
 
"shouldn't be possible on the switch at 60 fps"

A bit of an exageration considering Fast Racing Neo did it.. on Wii U.
 
I thought it would be obvious, such as reducing the memory requirement for the system or better streaming that doesn’t result in larger textures at a time for a slower storage medium vs smaller more textures that get scaled up.
Unless the upscaling work is being repeated each frame immediately before use the upscaled form going to need to sit in memory like a texture originally of that size.
 
But you know what... a newer node is one thing, but Changing foundries for Orion is a big deal and kinda hard for me to believe to think we would originally go from 8nm Samsung to 4-5nm TSMC. 5LPP or 4LPP Samsung and 3nm Samsung modes might just be more likely for Switch 2 and its revision, than tsmc 4/5nm. Hmm
The enterprise GPU and DC AI Ampere cards were manufactured on TSMC 7nm node. So there's no issue with Ampere crossing foundries. And I assume you wouldn't have an issue with ARM cores being produced on different foundries? The fact that A100 cards were on TSMC 7nm makes TSMC 6nm (an improved 7nm node) an option for Drake as well. It's what the PS5 and XBsX/S SoCs are manufactured on at the moment.
 
Unless the upscaling work is being repeated each frame immediately before use the upscaled form going to need to sit in memory like a texture originally of that size.
I'm not sure. Mipmapping works something like that. I don't think it's impossible for a texture to say, sit at a small size in memory and have its upscaled, mipmapped counterpart moved to screen memory upon draw. You could also have parts of the upscaling precalculated so not everything has to be done on the fly. There's lots of ways to use it to save both storage and memory.

However, it might be as well off doing something simpler, like just rendering at a low resolution and using textures to fit, then DLSSing the whole thing at once. Even if that isn't recommended, it can work.

I think upscaling is more likely to be used to save storage, rather than memory, due to the reasons you pointed out.
 
0
I know you’ve chosen conservative specs to illustrate a point, but wouldn’t the 8GB RAM be limiting if using 4K textures in a game?
They aren’t 4K textures, which is the point. They are upscaled textures in real time. So you can stream lower quality textures, and take up less memory in theory, to scale them to look like a higher quality.


But if you’re asking at all if 4K textures are limited on 8GB, yes.
 
I know many glossed over it but the most interesting thing from the ROG Ally spec reveal for me was the inclusion of UHS-II.

A low cost/low heat solution that the hardware file decompression engine could very well bridge the gap in storage speeds. This is right up Nintendo's wheelhouse.

I wouldn't be surprised if [TBA] had a little of
everything the ROG Ally has other than the screen and the SoC. All at a lower cost to consumers.
I don't think there are any micro SD UHS-II cards out there, which could be a problem. The solution would be to just support regular SD cards and tell people to use a converter to reuse their old cards

Ninja Edit: just Google, there are some but they are expensive. $100USD for a 256GB. Might be due to volume though, which Drake would help with
 
I don't think there are any micro SD UHS-II cards out there, which could be a problem. The solution would be to just support regular SD cards and tell people to use a converter to reuse their old cards

Ninja Edit: just Google, there are some but they are expensive. $100USD for a 256GB. Might be due to volume though, which Drake would help with
Going back to full sized SD card? That seems extremely unlikely. Remember this device is already size constrained...

If (big if, I'm not convinced) they NEED expandable storage with better speeds, UHS Card is supported by Nvidia's existing Orin kits and can fit into an envelope so similar to MicroSD that it can take both in the same receptical.

That said, Game Cards are likely to be the lowest common denominator, speed wise. And really, they achieve speeds the same as or better than just about any spinning rust out there. Games may complain, but no game will refuse to install on a PC HDD.

You have Direct Storage (or equivalent), a dedicated file decompression block and existing speeds that games must support as low as the Game Card. I think the thinking around faster storage is a little skewed. Nintendo Switch is not limited by its storage speed. The same speeds with dedicated decompression and a much faster CPU would dramatically improve load times and streaming. That appears to be the solution they have gone with.

Because of this, I don't think there will be much change to the storage situation. Other than improved density, I don't expect Game Cards to change, and I don't think they'll spring for anything other than UHS-1 MicroSD as a minimum for compatibility. In line with this, I think it's likely they stick with eMMC for internal storage, which is really, perfectly fine. 128 or 256GB of eMMC with up to 2TB of MicroSD support has it sitting pretty in terms of storage capabilities.

To be extremely clear:

128GB or 256GB is a decision I think NINTENDO will make. I don't think they'll ever offer storage variations that aren't seperate, visually distinct redesigns. Simplicity of production, simplicity of marketing. And avoiding anything akin to Wii U.
 
I think it's 70/30 in favor of no out of the box LPDDR5X support. There are licensable IPs from Synopsys that were available late 2021 that supported LPDDR4/5/5X, and memory to test with, but considering we have hints of verification starting in April of 2022, that might not be time to integrate a new MC with unknown power consumption properties.

I believe Nvidia makes it's own memory controllers for the discrete GPUs, but those support DDR/GDDR, not LPDDR, so I dunno if that's true for the Tegra line. Supporting LPDDR5X makes sense from a future proofing perspective, but the timing is kinda unfortunate.

I'm reasonably confident the LPDDR memory controllers used in the Tegra line are Nvidia's own work as well. They actually already have an LPDDR5X controller being used in Grace, which is currently sampling (and I believe was designed by the Tegra team). They also advertise it as being the world's first LPDDR5X controller with ECC, which makes me think it was designed in-house. However with Grace taping out at a similar time or possibly later than T239, the timelines may have been too tight to use it on T239.

I know many glossed over it but the most interesting thing from the ROG Ally spec reveal for me was the inclusion of UHS-II.

A low cost/low heat solution that the hardware file decompression engine could very well bridge the gap in storage speeds. This is right up Nintendo's wheelhouse.

I wouldn't be surprised if [TBA] had a little of
everything the ROG Ally has other than the screen and the SoC. All at a lower cost to consumers.

UHS-II isn't really that low cost. A quick look at B&H shows this useful comparison:

SanDisk 128GB Extreme PRO UHS-II SDXC Memory Card - 300MB/s - $129.99
SanDisk 128GB Extreme PRO CFexpress Card Type B - 1700MB/s - $119.99

The CFexpress Type B card is almost 6x as fast and costs less. Of course these were just the first cards that came up in each category, so there may be cheaper options for each, and a portable console would be unlikely to use either option, with microSD and CFexpress Type A being more realistic (and currently more expensive in each case), but this is just for illustrations's sake. For the brief period that UFS cards were available they were also cheaper than UHS-II SD cards while offering about twice the speed.

My point is that slower doesn't necessarily mean cheaper. The reason why CFexpress and UFS cards are/were so competitive on cost with UHS-II while offering much higher speeds is that they use parts that are already mass-produced. CFexpress is effectively an NVMe SSD in a plastic casing, while UFS cards are eUFS wrapped in plastic. Conversely, UHS-II SD cards have a custom interface, which means they need memory controller/interface chips which are custom designed and custom built just for UHS-II SD cards. The market for UHS-II cards is very small (and being eroded by CFexpress), so there's almost no economy of scale. By comparison, NVMe and UFS have enormous economies of scale, so even though CFexpress is also niche, the costs are much lower because they can use existing parts.

I don't have data on it, but I'd also suspect that the CFexpress card is much more power-efficient. It likely consumes more power overall than the UHS-II card, but probably consumes much less per GB transferred. The small market for UHS-II cards means UHS-II controllers are likely manufactured on older, less efficient manufacturing processes than would be used for mass-market technologies like NVMe and UFS. This is a limiting factor for SD Express as well. Because it features backwards compatibility with old SD card interfaces (but not UHS-II, as it happens), SD Express card manufacturers can't use off-the-shelf NVMe controllers. Hence you have SD Express cards using controllers manufactured on an old 28nm process consuming lots of power and running extremely hot.
 
I don’t think it’s right to rule out something quite yet, this is all for speculation after all, the tech exists and Sony FP used it which means it is possible, no one is saying Nintendo is going to adopt it right now, it is still a speculation thread where we can speculate. PS5 lacks ML inference capabilities, everything would need to be done in the shaders. PS5 can top out at 20TF of FP16 that can be used for this

Drake, let’s give it a theoretical 2TF docked and assume it can only deliver 1TF in portable mode. It can still achieve 16TF (with sparsity which is a hardware feature on Ampere and later) of FP16 thanks to those tensor cores. Not that far off from the PS5 peak FP16 capability.


With that in mind, let’s assume that Drake is 1TF portable, 2TF docked at best. CPU is clocked to no better than 800MHz and it only has 8GB of RAM. Let’s also assume that it has really slow storage speed, storage speed that is actually not at all different from what the switch has right now, let’s put it at 25MB/s for simplicity sake.

A feature like this that can upscale textures in real time would be pretty useful for such a limited device. Sure it has a decompression engine, but we can assume for this example that it operates at only a factor of 2x. So, the storage speed operates better than what it does right now by 2.


Ok we have the device in mind, right? It’s an upgrade from the switch, but a far cry from the PS4 Pro, XBox One X and Series S.

Of that 8GB of memory, 1-1.5GB let’s reserve it for OS related tasks.

So here is what we have in essence:

-2TF TV mode/1TF HH mode
-8 A7x cores clocked to 800MHz
-8GB of RAM, 6.5-7GB available for game usage
-Bandwidth of 102.4GB/s at the highest peak
-64GB of eMMC storage that operates with delivering 25MB/s of throughput in its read speed but an FDE that would aid this factor to be more like a 50MB/s.
-and for extra sake, all of this only delivers at best 4.75H of battery life in portable mode and 1.75H at the lowest


These stats helps us with quantifying the purpose or use case that this can have if it was implemented into Nintendo’s own first party tools and even more for third party tools. The low memory amount and the slow speed of the storage helps to allow for lower textures to be scaled in real time to a higher quality one. Not only that, but it can help alleviate a cart issue quite a bit I think, rather than having to ship with large textures, ship with a step below and allow the system to scale it up.

So, maybe a game that is, say, 50GB. Would need a 64GB cart to fit, but maybe with smart utilization they can squeeze it to the 32GB and have all included.

On top of all that, you can get faster load times for this device and it can have competitive load times if devs accommodate for the system and it’s limitations.



But this would require a lot of work from NV and Nintendo to make it part of its tools for devs to use and make proper use of.


In a different world, there would be a 4GB, 8GB, 16GB, 20GB, 24GB, 32GB, 40GB, 48GB, 64GB and an elusive but expensive, 128GB, cart for devs to choose from and the first 6 being relatively cheap.
i found 8GB of RAM too restric, 12/16GB of RAM would be ideal for Switch sucessor, given that Nintendo Switch games and possibly it sucessor games, will be far more ambitous then now, 8GB of RAM it would be extremely hard to do the next mainline Legend of Zelda game, the next 3D Mario, Metroid
 
i found 8GB of RAM too restric, 12/16GB of RAM would be ideal for Switch sucessor, given that Nintendo Switch games and possibly it sucessor games, will be far more ambitous then now, 8GB of RAM it would be extremely hard to do the next mainline Legend of Zelda game, the next 3D Mario, Metroid
he knows, he's talking about a hypothetical scenario, not actual Drake specs
 
Yes, even if it's at a superficial design level, just to make the distinction between the old and new model clearer.
My sentiments exactly. I also think they’ll sell several colors of the newer Joycons and maybe call ot something else besides regular Joycons. They need to justify the sale of a new variety of Joycons, lol

So they're going to GamesCom, remodelling their store, but have nothing to show for the second half of the year?
True. Too many actions together. Or maybe just a coincidence

Seems suspicious. I can't think of anything that could explain their recent actions other than a new console they're waiting to reveal before showing the holiday games, even if they're cross-gen.
Rumor has it the remodeling is a new holographic monitor

I don't think Pikmin 4 will affect the reveal. If anything, I think it's intentionally placed in July so that they have something to release between the June reveal and the release.
I concur

and stick technology.
Yes, please. Lol No more drift

I know this is a pretty major change of tune from where I’ve stood in the past, but I’m trying my best to acknowledge that not everything is an indicator of new hardware.
Wooord! Same here, hahahahahah

Maybe Nintendo is being veeery attentive with TOTK because of how late in the console’s life cycle it’s releasing. Like, this game will certainly move Switch units. They’re going all out on Marketing because of this.

And yeah, it’s their biggest release ever. It’s the first time they release a direct Zelda sequel and an expanded open world

Delaying marketing for what could be half a dozen titles by months and months over one game doesn't make sense. It's just not good business. And they KNOW that, or they wouldn't have shown off Pikmin 4.
Good point. Especially if part of those game really need an extended marketing campaign

128GB or 256GB is a decision I think NINTENDO will make. I don't think they'll ever offer storage variations that aren't seperate, visually distinct redesigns. Simplicity of production, simplicity of marketing. And avoiding anything akin to Wii U.
I agree. They want an ecosystem of consoles, not a variety. And it’s waaay easier in the supply chain management side.

Also, I’ll present a random counterargument to 2H being empty because of new hardware:
• They’re actually waiting to finish a big first party software before unveiling the second half of software
 
I'm reasonably confident the LPDDR memory controllers used in the Tegra line are Nvidia's own work as well. They actually already have an LPDDR5X controller being used in Grace, which is currently sampling (and I believe was designed by the Tegra team). They also advertise it as being the world's first LPDDR5X controller with ECC, which makes me think it was designed in-house. However with Grace taping out at a similar time or possibly later than T239, the timelines may have been too tight to use it on T239.



UHS-II isn't really that low cost. A quick look at B&H shows this useful comparison:

SanDisk 128GB Extreme PRO UHS-II SDXC Memory Card - 300MB/s - $129.99
SanDisk 128GB Extreme PRO CFexpress Card Type B - 1700MB/s - $119.99

The CFexpress Type B card is almost 6x as fast and costs less. Of course these were just the first cards that came up in each category, so there may be cheaper options for each, and a portable console would be unlikely to use either option, with microSD and CFexpress Type A being more realistic (and currently more expensive in each case), but this is just for illustrations's sake. For the brief period that UFS cards were available they were also cheaper than UHS-II SD cards while offering about twice the speed.

My point is that slower doesn't necessarily mean cheaper. The reason why CFexpress and UFS cards are/were so competitive on cost with UHS-II while offering much higher speeds is that they use parts that are already mass-produced. CFexpress is effectively an NVMe SSD in a plastic casing, while UFS cards are eUFS wrapped in plastic. Conversely, UHS-II SD cards have a custom interface, which means they need memory controller/interface chips which are custom designed and custom built just for UHS-II SD cards. The market for UHS-II cards is very small (and being eroded by CFexpress), so there's almost no economy of scale. By comparison, NVMe and UFS have enormous economies of scale, so even though CFexpress is also niche, the costs are much lower because they can use existing parts.

I don't have data on it, but I'd also suspect that the CFexpress card is much more power-efficient. It likely consumes more power overall than the UHS-II card, but probably consumes much less per GB transferred. The small market for UHS-II cards means UHS-II controllers are likely manufactured on older, less efficient manufacturing processes than would be used for mass-market technologies like NVMe and UFS. This is a limiting factor for SD Express as well. Because it features backwards compatibility with old SD card interfaces (but not UHS-II, as it happens), SD Express card manufacturers can't use off-the-shelf NVMe controllers. Hence you have SD Express cards using controllers manufactured on an old 28nm process consuming lots of power and running extremely hot.
Tbf UHS-I cards were pretty pricey before the Switch happened. The possibility of that reoccurring is higher than you think.

Fake edit: Can it be July already? I feel like every topic has been recycled to the point of exhaustion.
 
Last edited:
he knows, he's talking about a hypothetical scenario, not actual Drake specs
i see, let hope Nintendo opt for 12/16GB of RAM for Switch sucessor, if Nintendo choose 12 or 16GB of RAM for Switch sucessor, that would be excelent for third party and Nintendo internal studios and it subsidiaries, allowing they to make far more ambitous games such as Bayonetta 3 e Legend of Zelda Tears of the Kingdom kinda games, far more easier
 
i see, let hope Nintendo opt for 12/16GB of RAM for Switch sucessor, if Nintendo choose 12 or 16GB of RAM for Switch sucessor, that would be excelent for third party and Nintendo internal studios and it subsidiaries, allowing they to make far more ambitous games such as Bayonetta 3 e Legend of Zelda Tears of the Kingdom kinda games, far more easier
I don't see 8GB happening. ram is cheap and it seems that 4GB chips wasn't really a thing at the time of Drake starting development due to the direction LPDDR5 was heading (phones were building onto stacked ram while laptops were on dedicated chips and could use higher capacities).

honestly, with the way ram prices have been, I wouldn't be shocked if they did go with 16GB of LPDDR5. I don't expect it, but I just wouldn't be surprised
 
I'm reasonably confident the LPDDR memory controllers used in the Tegra line are Nvidia's own work as well. They actually already have an LPDDR5X controller being used in Grace, which is currently sampling (and I believe was designed by the Tegra team). They also advertise it as being the world's first LPDDR5X controller with ECC, which makes me think it was designed in-house. However with Grace taping out at a similar time or possibly later than T239, the timelines may have been too tight to use it on T239.
I could be wrong, but looking at the Figure 3-1 from p. 39 of the Arm Neoverse V2 Core Technical Reference Manual and Figure 2 from Nvidia's technical blog "NVIDIA Grace CPU Superchip Architecture In Depth", the RAM memory controller doesn't seem to be integrated into Grace. So if that's the case, I imagine Nvidia doesn't have to worry as much about choosing which RAM memory controller to use before taping out Grace since the RAM memory controller seems to be outside of Grace.
 
i found 8GB of RAM too restric, 12/16GB of RAM would be ideal for Switch sucessor, given that Nintendo Switch games and possibly it sucessor games, will be far more ambitous then now, 8GB of RAM it would be extremely hard to do the next mainline Legend of Zelda game, the next 3D Mario, Metroid
It’s intentionally lowest configuration for this system to show how an Upscaling can work real time in a beneficial sense for said system.

Unless the upscaling work is being repeated each frame immediately before use the upscaled form going to need to sit in memory like a texture originally of that size.
I mean, if you are housing lower quality textures, and you are having the trained algorithm work (it’ll still have its own resident space in memory), you can in theory save space without having to house say, 3GB of HQ textures, can be 1GB textures, algo of 1GB, and you save 1GB in this case vs just using the HQ textures.

In case A, 3GB out of the 6.5-7GB are used in textures, but in scenario B, 2GB are used out of the 6.5-7GB that are available for textures and Upscaling.



(This is just for example purposes)
 
for shits and giggles, I looked up the Switch assets I have on hand (BotW Link and Urbosa, some Pokemon SV models) and a lot of them have the same trend of having a lower resolution normal map (1024 diffuse, 512 normal/ao/rough/etc) in addition to many smaller textures. this is pretty much the exact usecase that Sony Santa Monica uses for God of War, upscaling normal maps.

this is definitely in NERD's wheelhouse to look into
 
Just for shits and giggles, and for entertainment purposes only, but if the FDE in Drake is potent, then the eMMC and/or the SD card speeds would be fine.

If it can decompress bits of data into a large amount of data, then they can manage that probably pretty fine and not need eUFS/NVMe/etc speeds.

But that’s if the FDE is potent.
 
0
Tbf UHS-I cards were pretty pricey before the Switch happened. The possibility of that reoccurring is higher than you think.

Fake edit: Can it be July already? I feel like every topic has been recycled to the point of exhaustion.
UHS-I had become the standard for SD cards (including microSD) long before Switch appeared. Here's a review of a typical UHS-I microSD card from four years before the Switch launched. Even UHS-II started to be adopted in the camera sector way back in 2014. UHS-I cards have come down in price since the Switch launched, but that's more to do with the natural drop in prices of NAND over that period than Nintendo driving prices down, and they're still pretty expensive, relatively speaking. The cheapest 1TB UHS-I microSD cards (like this one) are around $100. Meanwhile, for $90, you can get a 1TB PCIe 4.0 NVMe SSD capable of over 45x the read speeds. Or, for $60, a PCIe 3.0 variant that offers a meagre 23x the read speed of the SD card.

That being said, you're not wrong about Nintendo needing to drive adoption of a new format, I just don't think UHS-II SD cards are the right format to push. The digital camera industry has typically been the driving force behind the adoption of new memory card formats, but the industry has dropped off significantly even since Switch launched, let alone 10-15 years ago. Last year a total of just over 8 million digital still cameras were sold, which means even on a year of declining sales for Nintendo, the Switch is outselling the entire camera industry by over 2 to 1. Back in 2010, there were over 120 million digital cameras sold in a year, and since the drop-off from 2013 onwards no new memory format has managed to overtake UHS-I SD cards in the mainstream.

Without a significant outside market for cards faster than UHS-I, if Nintendo has to drive adoption themselves, it makes much more sense to leverage technologies that already have significant economies of scale. In terms of flash memory, this means using either an NVMe or UFS interface. Anything else requires a custom controller which limits both cost and power efficiency compared to off-the-shelf NVMe/UFS hardware. So, the options are CFexpress (type A, presumably), UFS cards, or some kind of proprietary card using either interface. UFS cards are smaller and should consume less power, whereas CFexpress Type A cards are a bit bigger (about the size of a Switch game card), but have the benefit of already being on the market. UFS cards would presumably be the 3.0 version, offering up to 1.2GB/s, whereas CFexpress Type A currently hit around 800MB/s, but future PCIe 4.0 cards should hit around 1.7GB/s and be compatible.

I could be wrong, but looking at the Figure 3-1 from p. 39 of the Arm Neoverse V2 Core Technical Reference Manual and Figure 2 from Nvidia's technical blog "NVIDIA Grace CPU Superchip Architecture In Depth", the RAM memory controller doesn't seem to be integrated into Grace. So if that's the case, I imagine Nvidia doesn't have to worry as much about choosing which RAM memory controller to use before taping out Grace since the RAM memory controller seems to be outside of Grace.

Grace's memory controller is definitely on the die. If it wasn't Nvidia would have to use a MCM/chiplet approach with a separate die (or dies) with the memory controller, which we don't see in any of the photos, where it's pretty clearly a monolithic design. Moving the memory interface off the main compute die is pretty common in chiplet setups, so if Nvidia move in that direction in the next few years we may see the memory controllers being moved off-die on Grace's successors, but for now it's a pretty standard monolithic setup.
 
Last edited:
It’s intentionally lowest configuration for this system to show how an Upscaling can work real time in a beneficial sense for said system.


I mean, if you are housing lower quality textures, and you are having the trained algorithm work (it’ll still have its own resident space in memory), you can in theory save space without having to house say, 3GB of HQ textures, can be 1GB textures, algo of 1GB, and you save 1GB in this case vs just using the HQ textures.

In case A, 3GB out of the 6.5-7GB are used in textures, but in scenario B, 2GB are used out of the 6.5-7GB that are available for textures and Upscaling.



(This is just for example purposes)

Upscaled HD textures still take up significant amounts of RAM, making this is a completely pointless application of pseudo real time upscaling. Chrono Cross HD had Square forget to optimize any of their upscaled visuals and it ate up all of the RAM, causing the original remaster to run like complete shit.

There's a likely a future where machine learning helps to reduce the size of models and textures without reducing the visual quality of these assets very much (and this would help with RAM), but this will be a completely production-side thing.

When trying to think of how to use the probably 64 tensor cores in the Switch, you always need to ask "wait, wouldn't this work a lot better in production or on the Cloud?"

Image reconstruction and image generation are not production-side or cloud-side and real time is needed. Other real time applications will also exist, but saving RAM via machine learning upscaling does not make sense and will not be one of them.
 
Upscaled HD textures still take up significant amounts of RAM, making this is a completely pointless application of pseudo real time upscaling. Chrono Cross HD had Square forget to optimize any of their upscaled visuals and it ate up all of the RAM, causing the original remaster to run like complete shit.
This isn’t at all what I’m talking about. Upscaled textures already loaded in =/= Upscaling textures in real-time.

These are two completely different concepts.
 
I wonder if with this piece of news TOTK might be the last Nintendo AAA title under the $69.99 price tag:

I think that depends on how well Tears of the Kingdom sells, because unlike Sony and Microsoft games, Nintendo games very rarely have price cuts.

Anyway, speaking of the type of RAM, I think the absolute best case scenario is LPDDR5X-7500, which is 120 GB/s of bandwidth. And the worst case scenario is LPDDR5-6400, which is 102.4 GB/s of bandwidth. (I think Nintendo's probably going to choose the worst case scenario.)
 
This is obvious off topic but honestly, PS5 and Xbox Series S/X is the first generation I've been generally priced out of (what I get on disability is nowhere near enough what I need) and I'm just really worried how Nintendo will respond overall, as it's been pointed out numerous times they haven't done any price drops...
 
This is obvious off topic but honestly, PS5 and Xbox Series S/X is the first generation I've been generally priced out of (what I get on disability is nowhere near enough what I need) and I'm just really worried how Nintendo will respond overall, as it's been pointed out numerous times they haven't done any price drops...
In terms of their next console I'd imagine that it's gonna be a little more expensive than the switch oled, as for games their flagships will probably be around 70 for your marios, zeldas, animal crossings and god forbid pokemon
 
Anyway, speaking of the type of RAM, I think the absolute best case scenario is LPDDR5X-7500, which is 120 GB/s of bandwidth. And the worst case scenario is LPDDR5-6400, which is 102.4 GB/s of bandwidth. (I think Nintendo's probably going to choose the worst case scenario.)
there's still 88GB/s as an option. but bandwidth is something they will want to mitigate as much as possible
 
T - 9 Days to go
By the way:



Guess 2H isn’t gonna be warren. We’ll definitely get a June Direct announcing all software. I presume they’re over focusing on TOTK to then unveil the rest of the 2023 lineup later.

But whether or not this next Direct will have any ties to the successor, who knows

Regardless, if people believe it’s this year, it’ll be announced this year. If people believe it’s early next year… it’ll be announced this year.

If it’s not early next year, it’s not going to get announced this year at all. They aren’t going to announce a system right after the holiday shopping.
 
Last edited:
Please read this staff post before posting.

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


Back
Top Bottom