• 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!)

Because it would be exspensive for a half step console. The same reason Sony did not do it for the PS4 Pro and Microsoft too
It wouldn't be hardware compatibility, it'd be an OS function. Which would also carry over to future hardware. This is how they will most likely solve BC on Dane
 
0
I'm really skeptical about Dane being made on 8 nm. We haven't had any official data concerning Orin S and a lot of mobile chips will have transitioned to nodes with extensive use of EUV with only entry level proposition still being made on 8 nm.

Nintendo having postponed the use of Dane separately from the OLED components and form factor could be an indication of a change of SoC and thus using a better node. Especially when fall 2022 will see the emergence of firsts 3 nm chips and 4/5/6 nm being cheaper.
 
I'm really skeptical about Dane being made on 8 nm. We haven't had any official data concerning Orin S and a lot of mobile chips will have transitioned to nodes with extensive use of EUV with only entry level proposition still being made on 8 nm.

Nintendo having postponed the use of Dane separately from the OLED components and form factor could be an indication of a change of SoC and thus using a better node. Especially when fall 2022 will see the emergence of firsts 3 nm chips and 4/5/6 nm being cheaper.
There isn't any indication that Switch OLED was ever supposed to use Dane. It's always been Mariko since it first showed up in firmware.
 
We may never see Dane or a newer chip in the TX1 version of the switch OS. We don't know if Nintendo are planning to use switch OS and letting informations concerning their new SoC at the mercy of dataminers.

This was brought up on the old boards when it got noticed (if anyone can source the thread, please do so, since not sure how to search for it), but earlier in the year, late last year, Nintendo had a job listing was looking for people to help port the OS to different hardware. So the intent seems "currently" as of whenever that job listing got posted, is to likely keep the OS the same. But with anything, plans can change.
 
This was brought up on the old boards when it got noticed (if anyone can source the thread, please do so, since not sure how to search for it), but earlier in the year, late last year, Nintendo had a job listing was looking for people to help port the OS to different hardware. So the intent seems "currently" as of whenever that job listing got posted, is to likely keep the OS the same. But with anything, plans can change.
Also, they can just as easily be writing updates to the Horizon OS behind a whole new code base/major version update, not unlike what other OS creators do when they have new major features/UI changes planned. The fact we knew about OLED through data mining tells me that they weren't as concerned about knowing a new model of the same hardware was planned and used the active code base for the current version of Horizon OS.
Since Orin chip prototypes are likely in play and they have actual hardware they can start to program it for, we'll know one way or another, but my thinking is they'll absolutely port Horizon to new hardware as-is using a private version of the code base, but I doubt it will be totally unchanged, they just want to keep what works now and work from that knowing their basic OS functionality will run at the same incredibly-efficient level it does now.
 
0
I would have thought 8 nm relevant in 2020. A little less but still one of the most plausible one in 2021 due to the rumored delay and chip shortages but harder to consider in 2022. Consoles/Apple/high end android chipsets have all seen record numbers in 2020 and 2021. Some middle end chipsets have been delivered in low supply (S780/E1080) or switched to N6 (S778) due to their high end counterparts being more profitable and seeing more demande.
 
0
Mobile SoCs move to new nodes immediately because that's just how they do things. The cheaper phones stick to the most cost effective nodes.

This whole 8nm thing feels like a forced problem. It's fine on its own, but if you start comparing it to anything, it's a terrible decision or something. Nintendo operates in their own little world and will continue to do so.
 
0
These phones are heavily subsidized and are being produced at a loss.
I would not say heavily subsidized. At one point, those brands have to recoup on these supposed losses (they don't benefit from software revenues). As for being produced at a loss, I would say that PS5, XSS and XSX are produced at a loss too. One of the last successful Nintendo product has been sold at a loss (3DS).
 
Quoted by: SiG
1
I would not say heavily subsidized. At one point, those brands have to recoup on these supposed losses. As for being produced at a loss, I would say that PS5, XSS and XSX are produced at a loss. One of the last successful Nintendo product has been sold at a loss (3DS).
I'd say the 3DS was sold more with "zero profit" but the Wii U was definitely sold at a loss (as in it cost more to produce than sell).
 
I'd say the 3DS was sold more with "zero profit" but the Wii U was definitely sold at a loss (as in it cost more to produce than sell).
Zero profits is literally saying at a loss.

Even if some Chinese brands are heavily subsidized, they constitute one of their several opponents in the hybrid mobile gaming market. Steam Deck, xCloud/xbox console streaming, the rest of the Cloud gaming market, powerful middle end smartphones (>1tflop GPU) weren't there in 2017 while some of these markets are reaching maturity in 2021. Moreover, they may not have another BOTW at launch date.
 
Quoted by: SiG
1
Zero profits is literally saying at a loss.

Even if some Chinese brands are heavily subsidized, they constitute one of their several opponents in the hybrid mobile gaming market. Steam Deck, xCloud/xbox console streaming, the rest of the Cloud gaming market, powerful middle end smartphones (>1tflop GPU) weren't there in 2017 while some of these markets are reaching maturity in 2021. Moreover, they may not have another BOTW at launch date.
...Lest BOTW sequel gets delayed/date given for 2023.
 
Rumors point that Nvidia is going to TSMC for their Lovelace based GPUs because Samsung’s fabs are prone to low yields.

That doesn’t sound good even if Dane is roughly 1/5 or 1/4 of the size of a normal GPU die.

I hope Nintendo ponies up and goes with TSMC for better yields or doesn’t it matter since Samsung will be cheap as “dirt” compared to TSMC?

Edit: nvm I forgot that the Orin is produced on Samsung 8nm and that would require Nvidia and Nintendo to spend money on producing a new chip
 
At this rate, I would think that the next switch model will be scheduled for 3 months before the Mario movie with a 3D Mario game with voice acting.
Zack Snyder's Nintendo Switch DLSS model...featuring Chris Pratt as Mario and Idris Elba as Knuckles...and Dante from the Devil May CryTM serires. New Funky Modes, etc.

Considering how COVID really put a wrench in Mario's 35th Anniversary, not to mention the whole Tokyo Olympics promotion fiasco, I'd take it they're just waiting for the stars to allign again regarding the new successor model.

And as most people hate to admit, they're going to have to sell it on a gimmick. 4K, DLSS, and ray tracing aren't going to be the primary draws, and reusing the same hybrid concept isn't going to be enough to differenciate them this "next-gen": They pretty much learned from the GameCube era that power isn't everything (yet they also learned from the Wii U era that power and ease of development is still important).
Rumors point that Nvidia is going to TSMC for their Lovelace based GPUs because Samsung’s fabs are prone to low yields.

That doesn’t sound good even if Dane is roughly 1/5 or 1/4 of the size of a normal GPU die.

I hope Nintendo ponies up and goes with TSMC for better yields or doesn’t it matter since Samsung will be cheap as “dirt” compared to TSMC?
I'm curious about this. Wouldn't this mean they'd have to start their SoC from scratch should they change fab nodes?
 
0
Rumors point that Nvidia is going to TSMC for their Lovelace based GPUs because Samsung’s fabs are prone to low yields.

That doesn’t sound good even if Dane is roughly 1/5 or 1/4 of the size of a normal GPU die.

I hope Nintendo ponies up and goes with TSMC for better yields or doesn’t it matter since Samsung will be cheap as “dirt” compared to TSMC?
At the same time, other rumors are pointing out that AMD and Qualcomm have grown tired of TSMC's politics that are heavily favouring Apple that have the priority and most of all their cutting edge capacities and thus would have switched for Samsung 3GAE for 3 nm.

Chip shortages and Samsung supposed lower yields didn't prevent Qualcomm making a profit with their 5 nm generation.
 
0
I do have a question: Lower yeilds with compared to what exactly? Is it TSMCs at a comparable node? And since it's likely to be one of the more mature nodes, doesn't that mean that it wouldn't be a s supply constrained as the more bleeding-edge ones like 3nm, etc.
 
I do have a question: Lower yeilds with compared to what exactly? Is it TSMCs at a comparable node? And since it's likely to be one of the more mature nodes, doesn't that mean that it wouldn't be a s supply constrained as the more bleeding-edge ones like 3nm, etc.
Yes. 5LPE was seeing lower yields compared to N5. But it was still better and more profitable than N7+/N6. Samsung nodes are good as long as we talk about small mobile chipsets. A bad yield tends to have a bigger impact when it comes to bigger GPU chipsets. That's why GPU manufacturers are going MCM the same AMD have gone MCM for their CPU products. To benefit from mobile yields and denser libraries.
 
0
Rumors point that Nvidia is going to TSMC for their Lovelace based GPUs because Samsung’s fabs are prone to low yields.
I think the rumours so far are only talking about high-end consumer Lovelace GPUs when talking about TSMC's N5 process node being used. No one knows if entry-level and/or mid-range consumer Lovelace GPUs will also be fabricated using TSMC's N5 process node, or will be fabricated using Samsung's 5LPP process node.

Especially when fall 2022 will see the emergence of firsts 3 nm chips and 4/5/6 nm being cheaper.
I highly doubt the first 3 nm** chips are coming in autumn 2022. TSMC won't be shipping 3 nm** wafers until Q1 2023. Samsung's 3GAE process node, which is planned to start high volume manufacturing by the end of 2022, is only available to Samsung, not other consumers. And Samsung's 3GAP process node is planned to start high volume manufacturing by the end of 2023.

And I highly doubt that a 4 nm** process node, at least from TSMC, will become cheaper in autumn 2022. Apple's rumoured to use one of TSMC's 4 nm** process nodes for fabricating the Apple A16 Bionic.
 
Yes. Volume production (risk production) beginning in early 2022 and mass production in fall 2022 with first chips delivery in 2023.

I would expect 5LPP to be more profitable than 8 nm at the end of next year. Especially when you could theorically make 4*5nm chips at the cost of one 8 nm one in the same wafer surface.
 
0
There still cheaper even with better OLED screens, 2-3 camera modules, 4g-5g capabilities, upgraded RAM/MEM and a similar battery capacity.
a better oled screen that is irrelevant in use for gaming, they barely match the performance that their specs dictate. The cameras in those phones are pathetic compared to the higher end phones and upgraded memory only just now, the latest switch is on the LPDDR4X, phones are briefly getting to LPDDR5 and next year 5X.

Battery is due to taking advantage of time.
 
0
We may never see Dane or a newer chip in the TX1 version of the switch OS. We don't know if Nintendo are planning to use switch OS and letting informations concerning their new SoC at the mercy of dataminers.
While it is certainly true that direct references to Dane are unlikely to show up in the public firmware builds for Switch 1, one would expect the Mariko version to show up fairly late if it was a last minute decision, which isn't really what happened. Switch OLED has been in the public firmware since right around when the pandemic started, and probably earlier than that for developers.

Also there's really no reason to think that the OS for Dane won't at least be descended from their current one, if it isn't just the exact same one.
 
While it is certainly true that direct references to Dane are unlikely to show up in the public firmware builds for Switch 1, one would expect the Mariko version to show up fairly late if it was a last minute decision, which isn't really what happened. Switch OLED has been in the public firmware since right around when the pandemic started, and probably earlier than that for developers.

Also there's really no reason to think that the OS for Dane won't at least be descended from their current one, if it isn't just the exact same one.
We have started to get concrete Orin/Ampere informations in the second half of 2019 that led to the taping out process in early 2020. I suppose that Nintendo may have had T239 data on 8 nm at the same time during Ampere evaluation on Samsung's node. They may have made their decision to stay on 12 nm for 2020-2021 at the moment they known the power consumption of Dane on 8N. I believe that an 8 nm Switch should have been released in the 2020-2021 time frame instead of 2022-2023. IMHO the fact that we didn't get a new model in 2020-2021 would be a indication of the use of 7 nm (or better) instead of 8 nm.
 
0
I want to mention that I think the advantages of quickly adopting the most advanced process nodes are becoming not as obvious as in the past, especially when more technological and economical problems arise from process nodes becoming more and more advanced, such as slow SRAM scaling, poor power scaling, logic process steps increasing as process nodes become more and more advanced, advanced packaging challenges, rising costs of semiconductor tools, cost of transistors rising slowly and steadily after 28 nm**, etc.
 
I want to mention that I think the advantages of quickly adopting the most advanced process nodes are becoming not as obvious as in the past, especially when more technological and economical problems arise from process nodes becoming more and more advanced, such as slow SRAM scaling, poor power scaling, logic process steps increasing as process nodes become more and more advanced, advanced packaging challenges, rising costs of semiconductor tools, cost of transistors rising slowly and steadily after 28 nm**, etc.
It also coincide with TSMC's taking the lead which will probably be less obvious with Intel 7 and 3GAE.
 
Last edited:
If I understand things, the newest and greatest process nodes also tends to have a temporary increase in binning or disqualified chips that do not meet any retail product requirements, thereby reducing production capacity. A matured process node with far fewer fabrication hiccups makes better sense for a mass-market product selling for less than $500. TX1 was using a matured 20nm process in 2017 when Switch launched, lest we forget.

And, to be clear, waiting for a new process node to mature down the line gives room for a Dane Lite after the initial release.
 
If I understand things, the newest and greatest process nodes also leads to a temporary increase in binning, thereby reducing production capacity.
Generally speaking, that's correct.

So according to NotebookCheck, the CPU frequencies are as follows:
Snapdragon 8cx Gen 3:
4 Cortex-X1 cores → 2.995 GHz
4 Cortex-A78 cores → 2.4 GHz

Snapdragon 7c+ Gen 3:
4 Cortex-A78 cores → 2.4 GHz
4 Cortex-A55 cores → 1.5 GHz

As for which process node Qualcomm's using for the Snapdragon 8cx Gen 3, considering the Snapdragon 8cx Gen 3 Geekbench 5 single-core scores are very similar to the Snapdragon 888 Geekbench 5 single-core scores, I feel confident enough to say that Qualcomm's probably using Samsung's 5LPE process node to fabricate the Snapdragon 8cx Gen 3. And I'm confident Qualcomm's using TSMC's N6 process node for fabricating the Snapdragon 7c+ Gen 3.

Edit: Hardwareluxx did confirm the Snapdragon 8cx Gen 3's fabricated using Samsung's 5LPE process node.
 
Last edited:
0
If I understand things, the newest and greatest process nodes also tends to have a temporary increase in binning or disqualified chips that do not meet any retail product requirements, thereby reducing production capacity. A matured process node with far fewer fabrication hiccups makes better sense for a mass-market product selling for less than $500. TX1 was using a matured 20nm process in 2017 when Switch launched, lest we forget.

And, to be clear, waiting for a new process node to mature down the line gives room for a Dane Lite after the initial release.
I still think Samsung 7nm will be an option for Dane Lite
 
I still think Samsung 7nm will be an option for Dane Lite
I think Samsung's 5 nm** process node is also an option, assuming Nintendo plans on releasing a new Nintendo Switch Lite model equipped with Dane at around 2024 (assuming Nintendo still plans on releasing the DLSS model* in holiday 2022) and Nvidia plans on using Samsung's 5 nm** process node for the entry-level and/or mid-range consumer Lovelace GPUs.
 
I think Samsung's 5 nm** process node is also an option, assuming Nintendo plans on releasing a new Nintendo Switch Lite model equipped with Dane at around 2024 (assuming Nintendo still plans on releasing the DLSS model* in holiday 2022) and Nvidia plans on using Samsung's 5 nm** process node for the entry-level and/or mid-range consumer Lovelace GPUs.
Yeah, they’ll want a mature process that gives them ample performance-per-watt savings, so it‘ll mostly depend on when Dane launches and how Samsung’s process nodes have matured by then.
 
0
I think Samsung's 5 nm** process node is also an option, assuming Nintendo plans on releasing a new Nintendo Switch Lite model equipped with Dane at around 2024 (assuming Nintendo still plans on releasing the DLSS model* in holiday 2022) and Nvidia plans on using Samsung's 5 nm** process node for the entry-level and/or mid-range consumer Lovelace GPUs.
if price scaling kicks back in for 5nm, I can see Nintendo jumping to it. I guess that's what 4nm is for
 
0
I still think Samsung 7nm will be an option for Dane Lite
I think Samsung's 5 nm** process node is also an option, assuming Nintendo plans on releasing a new Nintendo Switch Lite model equipped with Dane at around 2024 (assuming Nintendo still plans on releasing the DLSS model* in holiday 2022) and Nvidia plans on using Samsung's 5 nm** process node for the entry-level and/or mid-range consumer Lovelace GPUs.
If I'm not wrong, Samsung's 5 nm is using the same toolings used for their 7 nm node. That would explain why we are seeing a lot of 5LPE chipsets instead of 7LPP/6LPP. S780 were eating S888 capacities with a lower margin than S888 on 5LPE so Qualcomm decided to make stop the S780 production for the more profitable 5LPE chip.

As for 8cx3, the media coverage is stating that it uses 4*X1 but there is no proof that those big cores aren't just A78s with more $ memory.
 
What would be considered a "mature" node for a 2023/2024 launch window?

I wish there was a chart which shows where 8nm is in terms of node maturity. If Nintendo jumps to a later date, it would possibly be due to newer nodes having more production capacity moreso than the performance benefits. Again, if all depends on a bunch of stars alligning.

People always say "lightning in a bottle" but it's more about figuring out the perfect storm.
 
What would be considered a "mature" node for a 2023/2024 launch window?
N4 and 4LPX (probably 4LPP too). The former is a derivative of N5 which is going to support massive Lovelace/M1max chipsets. The latter will probably follow soon after with smaller middle end Lovelace chipsets.

Intel 7 can probably be considered as a mature node too.
 
Quoted by: SiG
1
N4 and 4LPX (probably 4LPP too). The former is a derivative of N5 which is going to support massive Lovelace/M1max chipsets. The latter will probably follow soon after with smaller middle end Lovelace chipsets.

Intel 7 can probably be considered as a mature node too.
How likely would it be for them to jump the project to a new node?

I'm assuming they have the money to "restart" development on a new SoC or configuration. The other factors being the global chip shortage still impacting any potential 2022/2023 launch.

As much as one could remain skeptical about Nintendo's next flagship console (i.e. the "Nintendo gonna Nintendo" crowd), one can't deny the possibility that sitations outside their control are very likely to force them to more recent (ARM) architectures and/or nodes if they want to avoid a paper launch of a new device...
 
How likely would it be for them to jump the project to a new node?

I'm assuming they have the money to "restart" development on a new SoC or configuration. The other factors being the global chip shortage still impacting any potential 2022/2023 launch.

As much as one could remain skeptical about Nintendo's next flagship console (i.e. the "Nintendo gonna Nintendo" crowd), one can't deny the possibility that sitations outside their control are very likely to force them to more recent (ARM) architectures and/or nodes if they want to avoid a paper launch of a new device...
Global chip shortages has a bigger impact on legacy nodes (28 nm and more) than cutting edge nodes. Founders don't want to increase capacity on these older nodes.

If switch 2 is using 8LPP/8LPU/8LPA, nothing could stop Nintendo to port the design on a newer node. It will be unoptimized but the deficit would be counterbalanced by the better density/efficiency of the newer node.

Sony rumoured to switch PS5 chipsests to N6 could be an indication that DUV nodes for cutting edge devices are coming to an end.
 
0
If I'm not wrong, Samsung's 5 nm is using the same toolings used for their 7 nm node. That would explain why we are seeing a lot of 5LPE chipsets instead of 7LPP/6LPP. S780 were eating S888 capacities with a lower margin than S888 on 5LPE so Qualcomm decided to make stop the S780 production for the more profitable 5LPE chip.

As for 8cx3, the media coverage is stating that it uses 4*X1 but there is no proof that those big cores aren't just A78s with more $ memory.
Samsung's 6LPP process node is not available to general consumers. And hence, there's no chip that's publicly announced to be fabricated using Samsung's 6LPP process node.

And Arm explicitly said that the Cortex-X1 uses a 5-wide decoder in comparison to the 4-wide decoder used for the Cortex-A78; and the Cortex-X1 has an out-of-window size of 224 entries in comparison to the Cortex-A78, which has an out-of-window size of 160 entries. So the increased amount of cache is definitely not the only reason why the Cortex-X1 is physically larger than the Cortex-A78.

As much as one could remain skeptical about Nintendo's next flagship console (i.e. the "Nintendo gonna Nintendo" crowd), one can't deny the possibility that sitations outside their control are very likely to force them to more recent (ARM) architectures and/or nodes if they want to avoid a paper launch of a new device...
I don't think adopting the Armv9 architecture will help with avoiding a paper launch, especially if demand for the DLSS model* will be as strong as demand for the Nintendo Switch. Product shortages are unfortunately inevitable, especially with the global chip shortage predicted to not end until 2023 at the earliest.
 
Last edited:
Samsung's 6LPP process node is not available to general consumers. And hence, there's no chip that's publicly announced to be fabricated using Samsung's 6LPP process node.

And Arm explicitly said that the Cortex-X1 uses a 5-wide decoder in comparison to the 4-wide decoder used for the Cortex-A78, and the Cortex-X1 has an out-of-window size of 224 entries in comparison to the Cortex-A78, which has an out-of-window size of 160 entries. So the increased amount of cache is definitely not the only reason why the Cortex-X1 is physically larger than the Cortex-A78.
Actually, I was saying that there is no evidence that what Qualcomm called 'big' cores in their presentation were X1s instead of A78(C)s in 8cx3.
 
Actually, I was saying that there is no evidence that what Qualcomm called 'big' cores in their presentation were X1s instead of A78(C)s in 8cx3.
So you believe that Qualcomm's outright lying to NotebookCheck, XDA Developers, Hardwareluxx, etc.?

Also, the CPU diagram of the Snapdragon 8cx Gen 3 shown by Qualcomm implies that a heterogenous CPU configuration rather than a homogeneous CPU configuration (which the Cortex-A78C is) is being used.

An argument can be made that the 4 "Efficiency" cores used in the Snapdragon 8cx Gen 3 are Cortex-A55 cores rather than the Cortex-A78 cores, considering how Qualcomm designates the "Efficiency" nomenclature to the Cortex-A55 cores.

I wish Qualcomm would actually be more transparent when hardware details are concerned.

Yes but the A78s in 8cx3 would be configured as big cores than 1080/780 that are middle end chips (Kirin 9000 chips could be more comparable to big A78Cs than E1080 big cores).
Qualcomm has never used the "Prime" nomenclature to describe the Cortex-A78 though. Qualcomm so far has reserved the "Prime" nomenclature to the Cortex-X1. And I'm saying this, because Qualcomm described that the Snapdragon 8cx Gen 3 has 4 "Prime" cores and 4 "Efficiency" cores.

Anyway, the Kirin 9000 Geekbench 5 single-core scores on average seems similar to the Exynos 1080 Geekbench 5 single-core scores.

~
Edit: The FTC has sued to block Nvidia's attempt to acquire Arm. I think Nvidia's attempt to acquire Arm is dead.
 
Last edited:
Yes but the A78s in 8cx3 would be configured as big cores than 1080/780 that are middle end chips (Kirin 9000 chips could be more comparable to big A78Cs than E1080 big cores).
I’m not really sure what you mean here, can you clarify?

They list 4 perf cores and 4 efficiency cores in the slides from QC. Even have the perf cores as big cores and the efficiency cores as small in the diagram.

They are heavily implying it is heterogeneous, not homogenous, in configuration.
 
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