• 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| (Read the staff posts before commenting!)

Reader mode
mjayer said:
aWY.gif
Click to expand...
both big and small?
 
The Switch had multiple power profiles, yall telling me you wouldn't use a higher tier power profile if it meant that you could get better performance in places where you are near an outlet or have access to more power? We have seen moderate overclocks improve switch performance noticeably, MVG has done a bunch of videos on them. Considering all the performance issues we've seen with late gen Nintendo published games, i think its worth looking into.
 
Mr.Gamerson said:
The Switch had multiple power profiles, yall telling me you wouldn't use a higher tier power profile if it meant that you could get better performance in places where you are near an outlet or have access to more power? We have seen moderate overclocks improve switch performance noticeably, MVG has done a bunch of videos on them. Considering all the performance issues we've seen with late gen Nintendo published games, i think its worth looking into.
Click to expand...
I have to imagine making 2 profiles for a game can be tricky enough during development, asking for a 3rd can't be any simpler. Not to mention it would kind of be janky. The Switch already goes black for a few seconds when changing between modes, and it's made to be as easy as possible to do. Would the system do the same thing when plugging it into a wall outlet without a dock? What happens if the plug gets knocked out of the wall or your USB C has a crappy connection? Would the system just start flickering? You know to expect the game to not work for a few seconds while (un)docking because you're (un)docking the system, but it certainly would feel terrible for the profile to change while you're playing because of some unforseen circumstance irl. Now what if you just want to play normally while plugged in? Would you have to navigate a bunch of different menus to enable and disable everything? This seems like a clunky idea that won't make much of a difference anyways on an 8inch 720p screen.
 
Last edited:
Mr Swine said:
I would keep my expectations in check as of what happened last with Switch (people thinking it was Tegra X2 on 16nm) and Wii U.

I firmly believe Drake is on 8nm and I would be really happy if it’s revealed in the future to be on 4N.
Click to expand...
This is a good way to manage expectations in check, yes. Assume 8nm until otherwise and be happy if it's fabbed on anything better.
 
0
Mr Swine said:
I would keep my expectations in check as of what happened last with Switch (people thinking it was Tegra X2 on 16nm) and Wii U.

I firmly believe Drake is on 8nm and I would be really happy if it’s revealed in the future to be on 4N.
Click to expand...
I get the sentiment, but there were very specific reasons for both of those.

An off the shelf TX2 was probably never an option for various reasons. Mariko at launch might have been possible, also might not have. A custom hybrid would probably have been ideal, but Nintendo needed something fast, so that might have been why they just went off the shelf (along with helping Nvidia meat their 20nm quotas and a avoid being fined).

Unless Nvidia has somehow achieved huge efficiency gains with Ampere on the 8nm node, why they would design a custom 12 SM soc on that node is very hard to explain from our vantage point.
 
Mr.Gamerson said:
The Switch had multiple power profiles, yall telling me you wouldn't use a higher tier power profile if it meant that you could get better performance in places where you are near an outlet or have access to more power?
Click to expand...
This is a really terrible idea.

Switch’s different power profiles are selected by developers, not users. Developers pick the profile that works best, and try to deliver a good experience. And unlike PC, you don’t have control over visual settings.

Adding a plugged in power profile would either cause it to be ignored by devs, or cause more games to launch broken in the base mode. Users would complain that it wasn’t really a handheld system, if you have to be plugged in to get a good experience. Charging would be slower due to increased power draw, further limiting the quality of handheld mode.

There is a market for a highly flexible device, supported by an ecosystem of highly scalable and flexible software - the PC, and there are numerous handheld PCs to choose from.

There will always be games that push the hardware limits to the breaking point, and overclocking can help. But you can’t bypass that by offering the “overclock” on day 1. Once devs can target that mode you’ll get the same number of games broken, just at this higher profile.
 
ItWasMeantToBe19 said:
How well do we think the Switch 2 will sell if the worst case scenario happens and it’s near PS4 Pro level but 8nm causes it to be $500 and heavier than the Steam Deck?
Click to expand...
If it's 8nm, it won't reach PS4 Pro performance in Docked because portable and docked clock profiles differences are synced due to cooling reasons. Regardless, it will sell bazillions because most people will see Nintendo games + 3P games will good IQ and visuals and they won't care that the machine won't reach "XSS or PS4 Pro" performance.

Also, if fabbed on 8N, the chip will be fairly decent cheap to manufacture and sell, assuming reasonable small size and good parametric yields, of course.
 
Ghostsonplanets said:
If it's 8nm, it won't reach PS4 Pro performance in Docked because portable and docked clock profiles differences are synced due to cooling reasons. Regardless, it will sell bazillions because most people will see Nintendo games + 3P games will good IQ and visuals and they won't care that the machine won't reach "XSS or PS4 Pro" performance.

Also, if fabbed on 8N, the chip will be fairly decent cheap to manufacture and sell, assuming reasonable small size and good parametric yields, of course.
Click to expand...

People calculated that the chip will have to be clocked fairly high to work at all and the chip would be bigger than the Series S chip at 8nm. So an 8nm chip could still have very good performance, just while eating maybe 3x as much power as the original Switch (5x as much as the revised Switch) and being Steam Deck or bigger in physical size.
 
TX2 was never a good option. It was not for portable devices and the Denver cores were large, power hungry and not that fast. TX1 in 16nm would have been fine, but we got that anyway.
 
PedroNavajas said:
TX2 was never a good option. It was not for portable devices and the Denver cores were large, power hungry and not that fast. TX1 in 16nm would have been fine, but we got that anyway.
Click to expand...
We got that anyway, but had we gotten it at launch, the clocks would likely have been much better (especially cpu).
 
0
ns156 said:
If it's 8nm, then our 12SM figure was wrong from the get go, and it will perform like ass relative to our expectations, assuming it's not a huge device and keeps a similar form factor to the original Switch.
Click to expand...

Marking these down for the future “IWMTB19 was right about the Switch 2 being a weight lifting device” thread (not really, but…)
 
Christer said:
Regarding the production node discussion, may I ask what the original source is for the chip having 12 SM?

just want to get an idea of the confidence level we can have about that detail since the feasibility of 8nm seems to hinge quite a lot on that.
Click to expand...
the nvidia theft
 
0
Christer said:
Regarding the production node discussion, may I ask what the original source is for the chip having 12 SM?

just want to get an idea of the confidence level we can have about that detail since the feasibility of 8nm seems to hinge quite a lot on that.
Click to expand...
Kopite7Kimi. Used to be very reliable Nvida leaker, but has gotten some things wrong about Lovelace. Also has gotten some things wrong about Drake.

Edit: thought you asked about the source for 8nm. The source for 12SM is the nvn 2 leak.
 
0
Christer said:
Regarding the production node discussion, may I ask what the original source is for the chip having 12 SM?

just want to get an idea of the confidence level we can have about that detail since the feasibility of 8nm seems to hinge quite a lot on that.
Click to expand...
The source of 12 SMs for the chip was from the Nvidia leak. It's one of the big points on why 8nm is unlikely, as least as far as I understand. Before that, folks were speculating no more than 8 SMs, which is why seeing 12 SMs took people by surprise.
 
Machrider said:
I mean well Mortal Kombat 1 on Switch is a game that probably the Switch can ran a lot better than that. People can call the developers lazy bums all they want but the fact is when you really have to make a baseline version of a game that only utilizes about 40% of any chip ... it can be a massive hinderance in trying to port a game like that.

If they were allowed to use the Switch chip to its full potential, they could probably deliver a version that looks better than that.

Now some people will say "well if that's the case they shouldn't release the game period and no one should be able to play it!".

I think it's just better to allow in situations like that an option where a developer can utilize the full potential of the chip even when portably.

If it was a case of some exotic or extreme solution was the only way to do it, I'd be against it, but when it really just comes down to battery size basically and extending a battery in this day and age is so cheap and easy to do, I say allow/make it easier for that to be possible.

It can help a system like a Switch an awful lot and I think lots of people are perfectly willing to spend a few bucks extra on an accessory if they have to. No one's asking for like a $100 upgrade here, a 5000 MaH ion-lithium battery can be sold at a profit even at $20 these days. I paid more for my GameCube and PS2 memory cards and that used to be just a standard thing you'd pay for without even thinking twice about.

You can dramatically alter the performance of a hardware like the Switch for something that costs about the same or less than a PS2/GC memory card back in the day, if there was something that could do that for the GameCube for that price, I'd have done backflips of joy.
Click to expand...
unless your console have a good arquicture for developers to work, it will be pointless to work on it, developers want a console that is easy to program, not a chimera that make even
the simplest game a hurdle/pain to make(PS3)
 
0
Christer said:
Regarding the production node discussion, may I ask what the original source is for the chip having 12 SM?

just want to get an idea of the confidence level we can have about that detail since the feasibility of 8nm seems to hinge quite a lot on that.
Click to expand...
That comes from the Nvidia hack. The authenticity isn't really in question, though there is a very small, but extant risk of interpretation error.
 
Christer said:
Regarding the production node discussion, may I ask what the original source is for the chip having 12 SM?

just want to get an idea of the confidence level we can have about that detail since the feasibility of 8nm seems to hinge quite a lot on that.
Click to expand...
Others already pointed out, but it's from those lines in the leaked NVN2 doc

SM = streaming multiprocessors

// Number of warps per SM on ga10f
#define __NVN_NUM_WARPS_PER_SM_GA10F 48

// Number of SMs on on ga10f
#define __NVN_NUM_SMS_GA10F 12
Click to expand...
Referenced to for first time here I believe
 
Giancarlo said:
which of this options is more likely for Nintendo to choose on it next hardware, given the Eurogamer/VGC report on the console?
Click to expand...
The middle one seems most plausible IMHO.

The first one, is maybe plausible, but I hope it's not the direction Nintendo/nvidia went into.

Last one, nah.
 
Giancarlo said:
which of this options is more likely for Nintendo to choose on it next hardware, given the Eurogamer/VGC report on the console?
Click to expand...
The woke option.

We all want this thing to be as impressive as possible, so threads like this have an inherent bias. there's also the possibility that it will have some sort of variable clock setup, or some factor we didn't think of that makes all our math invalid.
 
ItWasMeantToBe19 said:
People calculated that the chip will have to be clocked fairly high to work at all and the chip would be bigger than the Series S chip at 8nm. So an 8nm chip could still have very good performance, just while eating maybe 3x as much power as the original Switch (5x as much as the revised Switch) and being Steam Deck or bigger in physical size.
Click to expand...
ns156 said:
If it's 8nm, then our 12SM figure was wrong from the get go, and it will perform like ass relative to our expectations, assuming it's not a huge device and keeps a similar form factor to the original Switch.
Click to expand...
I'm aware of the fantastic job Thraktor and others did regarding the node choice. However, we can't affirm x or y with certainty as Samsung, Nvidia and Nintendo could optimize/co-technology the node for further efficiency/performance. The bullshit rumor floating around of "7LPH" being an optimized node for T239 and Nintendo could very well be true in the sense that the parties enhance/tweak the node further to serve their objectives. We also don't know much about Samsung/Nvidia 8N node. It's very probable that Nvidia did their DTCO on the 8LPP node, which was the first iteration. Samsung kept investing and refininf further the 8nm node with 8LPU (adds extreme low voltage cells) and 8LPA (iirc 15% better efficiency). We also don't know how big or how guzzling the different automotive structures of Orin are, which Nintendo and Nvidia have cut on T239.

Another fact is that we aren't aware of Nintendo battery life goals, clock targets, battery size, device size and weight, etc. The only certainty is that Nintendo will deliver a device with good battery life and that is around the same weight and size class of OG Switch rather than PC Handhelds.

There's also the fact that 8nm is the cheapest modern non-EUV node. And Samsung Foundry is desperate for customers. Enough that a deal between Samsung Group as a whole and Nintendo, where Samsung get exclusive suppliers rights for RAM, Storage, SoC Manufacturing, Future OLED Screens, Battery, etc ,while Nintendo gets a sweetned price on these parts, isn't out of possibility.

TLDR: Public data and fantastic work by the amazing people here based on this data show us that 8N is very unfeasible for T239. But public data is the extent of what we can infer. We don't know what Nintendo and Nvidia will do, we lack insight into the size of Computer Vision/Automotive and redundant structures to properly estimate T239 size when compared to GA107 (Closest Consumer cousin), etc. While our data shows us that 4N is the node of choice for a SoC like T239, we don't have knowledge on what can be worked out behind the scenes.
 
redmutineer75 said:
Broke: Switch 2 will be Steam Deck size with T239 on SEC 8N
Woke: Switch 2 will be original Switch size with T239 on TSMC 4N
Bespoke: Switch 2 will be Steam Deck size with T239 on TSMC 4N because it's secretly 16 SMs
Click to expand...

Others here longer than I have been can correct me if I'm wrong, but the belief, before NVN2 leak, used to be it's either 4 SM or 8 SM? Optimists say 8 SM, pessmists say 4 SM.

Then NVN2 leak came, even the optimists were wrong. That would be the "bespoke" scenario back then in your example, lol (secretly it has been 12 SMs all all along, not 4 or 8 SMs)
 
Re this whole battery life versus performance model debate.

Battery life does actually matter. Maybe people don't play in 6 hour mega marathons, but they take their Switch out and about: on their commute, to school, on the plane, in the back seat on a long road trip, while visiting Grandma. Because people are imperfect, they sometimes forget to charge or forget their charger and don't want to miss out on playing. A battery pack is a second thing to forget to pack or forget to charge, and is an extra powerpoint if you're travelling.

And bulk/weight matter. Some people want to throw their Switch in a bag or hand bag, or want their kids to be able to carry it around. They want to be able to travel light, without having to bring a kitchen sink of extra accessories.

It's easy to talk about an extra profile as a value add, but, realistically, there would be some developers who could get a game running on a standard-handheld profile who may not bother with that optimisation if they can use a boost-profile instead (because capitalism and profit). That means, in practice, the option of an extra profile would mean less handheld functionality for handheld gamers in some games. At which point you have to weigh up - are there more gamers who'd want a boost profile, compared to the ones who just want an easy standard handheld experience?

The Switch isn't a PC. A key advantage of consoles over PC is that stuff just works without need for a bunch of fussing around and optimising. I'm saying that as someone who has a handheld gaming PC - I like that it lets me play games that aren't on the Switch, but the added PC frustrations do sometimes get in the way of just playing the damn game.

I can see why some gamers want more power in handheld, but there are reasons why it would make the Switch a worse experience for others. I think it's important not to lose sight of that, especially when we all tend to consider things through the prism of our own needs and preferences - when those of us here aren't going to be a representative sample of Switch owners.
 
Ghostsonplanets said:
I'm aware of the fantastic job Thraktor and others did regarding the node choice. However, we can't affirm x or y with certainty as Samsung, Nvidia and Nintendo could optimize/co-technology the node for further efficiency/performance. The bullshit rumor floating around of "7LPH" being an optimized node for T239 and Nintendo could very well be true in the sense that the parties enhance/tweak the node further to serve their objectives. We also don't know much about Samsung/Nvidia 8N node. It's very probable that Nvidia did their DTCO on the 8LPP node, which was the first iteration. Samsung kept investing and refininf further the 8nm node with 8LPU (adds extreme low voltage cells) and 8LPA (iirc 15% better efficiency). We also don't know how big or how guzzling the different automotive structures of Orin are, which Nintendo and Nvidia have cut on T239.

Another fact is that we aren't aware of Nintendo battery life goals, clock targets, battery size, device size and weight, etc. The only certainty is that Nintendo will deliver a device with good battery life and that is around the same weight and size class of OG Switch rather than PC Handhelds.

There's also the fact that 8nm is the cheapest modern non-EUV node. And Samsung Foundry is desperate for customers. Enough that a deal between Samsung Group as a whole and Nintendo, where Samsung get exclusive suppliers rights for RAM, Storage, SoC Manufacturing, Future OLED Screens, Battery, etc ,while Nintendo gets a sweetned price on these parts, isn't out of possibility.

TLDR: Public data and fantastic work by the amazing people here based on this data show us that 8N is very unfeasible for T239. But public data is the extent of what we can infer. We don't know what Nintendo and Nvidia will do, we lack insight into the size of Computer Vision/Automotive and redundant structures to properly estimate T239 size when compared to GA107 (Closest Consumer cousin), etc. While our data shows us that 4N is the node of choice for a SoC like T239, we don't have knowledge on what can be worked out behind the scenes.
Click to expand...

“Nintendo won’t make a system massively larger than the Switch 1” just feels like one of these assumptions that is held onto way too strongly. The Steam Deck isn’t popular, but I don’t know if the weight is the reason.
 
darthdiablo said:
Others here longer than I have been can correct me if I'm wrong, but the belief, before NVN2 leak, used to be it's either 4 SM or 8 SM? Optimists say 8 SM, pessmists say 4 SM.

Then NVN2 leak came, even the optimists were wrong. That would be the "bespoke" scenario back then in your example, lol (secretly it has been 12 SMs all all along, not 4 or 8 SMs)
Click to expand...
Or 6 SM.
 
darthdiablo said:
Others here longer than I have been can correct me if I'm wrong, but the belief, before NVN2 leak, used to be it's either 4 SM or 8 SM? Optimists say 8 SM, pessmists say 4 SM.

Then NVN2 leak came, even the optimists were wrong. That would be the "bespoke" scenario back then in your example, lol (secretly it has been 12 SMs all all along, not 4 or 8 SMs)
Click to expand...
I believe the expectations where ranging from 4 to 8 SM on 8nm
 
Mr Swine said:
I would keep my expectations in check as of what happened last with Switch (people thinking it was Tegra X2 on 16nm) and Wii U.

I firmly believe Drake is on 8nm and I would be really happy if it’s revealed in the future to be on 4N.
Click to expand...
I’ll join you and believe it doesn’t exist, we can both be in the pessimistic boat together!
 
ItWasMeantToBe19 said:
“Nintendo won’t make a system massively larger than the Switch 1” just feels like one of these assumptions that is held onto way too strongly. The Steam Deck isn’t popular, but I don’t know if the weight is the reason.
Click to expand...
It just makes intuitive sense. Nintendo consoles are designed to sell to kids and their parents. A bulky console that a child can't easily hold wouldn't be as marketable. Not to mention that a larger device means less can be shipped in a given space, and when you're going to be selling a hundred million of them, how many consoles can fit in a box or on a truck is a real concern.
 
I have just think of something…

If next Switch HW has a camara, focused on AR, somewhere as its new gimmick. Maybe they can launch a new Nintendogs title? With Switch NG specs, they can make a very realistic game with a heavy component in simulations/physics between the pet (either dog or cat) and the environment + photorealistic models of dogs + AR gimmick.
 
oldpuck said:
This is a really terrible idea.

Switch’s different power profiles are selected by developers, not users. Developers pick the profile that works best, and try to deliver a good experience. And unlike PC, you don’t have control over visual settings.

Adding a plugged in power profile would either cause it to be ignored by devs, or cause more games to launch broken in the base mode. Users would complain that it wasn’t really a handheld system, if you have to be plugged in to get a good experience. Charging would be slower due to increased power draw, further limiting the quality of handheld mode.

There is a market for a highly flexible device, supported by an ecosystem of highly scalable and flexible software - the PC, and there are numerous handheld PCs to choose from.

There will always be games that push the hardware limits to the breaking point, and overclocking can help. But you can’t bypass that by offering the “overclock” on day 1. Once devs can target that mode you’ll get the same number of games broken, just at this higher profile.
Click to expand...

Well, I have seen the homebrew community do it, I figured Nintendo could do it even better and make it safer. It's clearly useful because people are going out there way to hack and mod their switch's to get more juice out of it. You also wouldn't need to plug in a power plug for it to run, you just plug in an external source if you want to play for longer than 1 to 2 hours you would get at the higher settings on battery, like when I plug in my laptop. In my head a game like Hyrule warriors would of had a toggle in the menu settings to allow for the higher profile that we would toggle to then allow for the better performance in handheld. It's also not just about getting to acceptable performance, it's also useful for making a game that runs at locked 30 fps now run at an unlocked 40-60fps. Ultimately, if it's hard to implement properly I definitely don't expect them to do it all, it's just a request if it's something that doesn't take much to implement. The Switch is literally designed to be able to switch profile's quickly and then run, and if they have to design the higher power profile for handheld in a way that disables docked mode once you toggle it then so be it, let me get me get more out of my handheld. Devs and consumers don't have to use this function, it's just an option.
 
Since the Switch 2 runs The Matrix demo reasonably well, it has to be pretty powerful.

The 4N vs Samsung 8nm debate is therefore entirely about how fucking huge this thing will be.
 
LoneRanger said:
I have just think of something…

If next Switch HW has a camara, focused on AR, somewhere as its new gimmick. Maybe they can launch a new Nintendogs title? With Switch NG specs, they can make a very realistic game with a heavy component in simulations/physics between the pet (either dog or cat) and the environment + photorealistic models of dogs + AR gimmick.
Click to expand...
+ it has dedicated AI hardware. Nintendogs would definitely prioritize a dog with machine learning over having dlss.
 
Ghostsonplanets said:
I'm aware of the fantastic job Thraktor and others did regarding the node choice. However, we can't affirm x or y with certainty as Samsung, Nvidia and Nintendo could optimize/co-technology the node for further efficiency/performance. The bullshit rumor floating around of "7LPH" being an optimized node for T239 and Nintendo could very well be true in the sense that the parties enhance/tweak the node further to serve their objectives. We also don't know much about Samsung/Nvidia 8N node. It's very probable that Nvidia did their DTCO on the 8LPP node, which was the first iteration. Samsung kept investing and refininf further the 8nm node with 8LPU (adds extreme low voltage cells) and 8LPA (iirc 15% better efficiency). We also don't know how big or how guzzling the different automotive structures of Orin are, which Nintendo and Nvidia have cut on T239.

Another fact is that we aren't aware of Nintendo battery life goals, clock targets, battery size, device size and weight, etc. The only certainty is that Nintendo will deliver a device with good battery life and that is around the same weight and size class of OG Switch rather than PC Handhelds.

There's also the fact that 8nm is the cheapest modern non-EUV node. And Samsung Foundry is desperate for customers. Enough that a deal between Samsung Group as a whole and Nintendo, where Samsung get exclusive suppliers rights for RAM, Storage, SoC Manufacturing, Future OLED Screens, Battery, etc ,while Nintendo gets a sweetned price on these parts, isn't out of possibility.

TLDR: Public data and fantastic work by the amazing people here based on this data show us that 8N is very unfeasible for T239. But public data is the extent of what we can infer. We don't know what Nintendo and Nvidia will do, we lack insight into the size of Computer Vision/Automotive and redundant structures to properly estimate T239 size when compared to GA107 (Closest Consumer cousin), etc. While our data shows us that 4N is the node of choice for a SoC like T239, we don't have knowledge on what can be worked out behind the scenes.
Click to expand...

Quite frankly, Samsung's nodes are such stinkers that I have close to 0 faith in them improving it enough to be usable over 4N.
 
ItWasMeantToBe19 said:
Okay so how is this even this pessimistic of a take.

“The Switch 2 could be heavier than a Steam Deck if all the reasonable rumors are true” is just

1. Accurate
2. Not that big of a deal

I’m just poking fun at the potential size.
Click to expand...
That’s not what I asked about lol, nothing to do with your “pessimistic” take.
 
Reader mode
You must log in or register to reply here.
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