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

[minimalism]

Such a thing will never happen because it shows integrity.

When OLED was announced, I immediately said it was not the hardware I had been discussing and still stood by the Switch hardware with DLSS information as accurate and that devkits had been distributed in late 2020 and throughout 2021. People called it a lie.

Enter the leak...

Silence from those individuals or they pivot to, "You called a Pro & it's a 2... so you were wrong".

I've always felt that OLED Switch was borne out of lack of CPU/GPU chip availability for the 'Switch Pro', am I too far off the mark? Using available components for a future buy make sense in a long term supply chain thought process.

 

[Eclissi]

Makes a ton of sense. If Nintendo releases this as a simple revision that plays current games better and has 4K capability then there is a risk that sales will eventually fall off like it did with previous handheld revisions.

Calling it Switch 2 makes more sense and they can have ads showing it playing current Switch games and the new Switch 2 exclusive ones. Future exclusive games will have a “Only on Switch 2” on the box while cross gen will have ”Works on Switch and Switch 2”.

Otherwise how will you market exclusive Switch games on a revision that is supposedly to be on the same family and that is way more powerful?

We are used to reasoning with the classic concept of "generations".
It is possible that Nintendo wants to move in a different way now, taking its cue from the mobile world.
This would also allow new games to be released on a huge installed base, while exclusive titles would not have that possibility. We must remember that Nintendo has the biggest gains from the sale of the software.

We have already seen cross gen (BotW) in the past... it will be enough to make it the rule rather than the exception, at least for one or two years.

 

[Zedark]

It should be noted that BC isn't completely automatic. For the Series X, for example, they developed a (type-2 I assume) hypervisor to manage the compatibility with the older hardware. I presume Nintendo could do the same, but hypervisors are probably a lot more difficult to produce than simple clock matching or completely native compatibility (or alternatively emulation) would allow for. But the latter are more likely to produce some incompatibilities if the hardware doesn't allow for it I think (but correct me if I'm wrong).

Then again, I imagine NVIDIA will have made it a priority to have similar enough architectures between GPUs to allow for easy compatibility for all games even with signfiicant optimisation for different PC games produced over the years (i.e. something like Crysis should not crash on the RTX30 series GPUs, and should run better than on older hardware). As such, native compatibility might work better than some give it credit for (still watching the MVG/Spawn Wave discussion, so I'm not commenting on his words here - other people have worded similar concerns before).

 

[ReddDreadtheLead]

I've always felt that OLED Switch was borne out of lack of CPU/GPU chip availability for the 'Switch Pro', am I too far off the mark? Using available components for a future buy make sense in a long term supply chain thought process.

I’m absolutely certain that the OLED model was not born out of that, but with the goal to introduce parts that will be utilized for the 4K model and thus making it cheaper while also more difficult to actually leak from a housing or display perspective. Don’t want a switch lite scenario again.

But anyway, my main point is that it was more to introduce parts that they already secured for the other model. They have reused some parts before in their manufacturing, or some elements from previous consoles in the successor console like the DSi having the same charger as the 3DS and it released years ahead of it.


Reduce, reuse, recycle!

 

[Eclissi]

To clarify what I wrote before... it is clear that the new console will be a real generational leap, because the Nvidia leak. This is not contedtable. Instead, what needs to be understood is how Nintendo will "offer" the console to the public. Marketing aspect.
Announcing a successor when current sales are so high, It would be a surprising move.

 

[Pokemaniac]

It should be noted that BC isn't completely automatic. For the Series X, for example, they developed a (type-2 I assume) hypervisor to manage the compatibility with the older hardware. I presume Nintendo could do the same, but hypervisors are probably a lot more difficult to produce than simple clock matching or completely native compatibility (or alternatively emulation) would allow for. But the latter are more likely to produce some incompatibilities if the hardware doesn't allow for it I think (but correct me if I'm wrong).

Then again, I imagine NVIDIA will have made it a priority to have similar enough architectures between GPUs to allow for easy compatibility for all games even with signfiicant optimisation for different PC games produced over the years (i.e. something like Crysis should not crash on the RTX30 series GPUs, and should run better than on older hardware). As such, native compatibility might work better than some give it credit for (still watching the MVG/Spawn Wave discussion, so I'm not commenting on his words here - other people have worded similar concerns before).

Actually, all Xbox games run in a Hypervisor since Xbox One, even native ones. It's not really a backwards compatibility thing specifically, it's just how Xbox games get run. This might simplify certain aspects of BC, but it's not really necessary if the OS is compatible and the graphics driver can take care of any necessary translation.

 

[SiG]

Do you guys see Nintendo re-releasing "Definitive" like physical editions for this more powerful Switch? Or would Nintendo patch released games to take advantage of the stronger hardware? Or both?

Likely patch. It's not like DLSS needs an overhaul in an engine to get it working correctly, though I do wish more developers/publishers tune for a proper balance of IQ/performance.

 

[NineTailSage]

So have we heard or discussed this feature yet possibly being included?
It was definitely present in A100 GPU but I wonder if this will make its way over to Drake also...




"The NVIDIA Ampere architecture adds Compute Data Compression to accelerate unstructured sparsity and other compressible data patterns. Compression in L2 provides up to 4x improvement to DRAM read/write bandwidth, up to 4x improvement in L2 read bandwidth, and up to 2x improvement in L2 capacity."

NVIDIA Ampere Architecture In-Depth | NVIDIA Technical Blog

Today, during the 2020 NVIDIA GTC keynote address, NVIDIA founder and CEO Jensen Huang introduced the new NVIDIA A100 GPU based on the new NVIDIA Ampere GPU architecture. This post gives you a look…

developer.nvidia.com

 

[ITSMILNER]

What if everyone is wrong and this isn’t a new Switch but rather a portable Wii U :shock:

 

[Branduil]

It's hard for me to square the circle of a console "revision" that requires translation layers and/or patches just to properly run the games for the system it is supposedly a revision of.

In terms of marketing, yeah I expect they won't call it something as straightforward as "Switch 2," but they're absolutely going to emphasize exclusive games and reasons to buy it beyond just prettier Switch games. Especially since the leaked hardware is several times more powerful than necessary for 4K Switch games.

 

[Z0m3le]

Hey, so this is just a theory, but I have a mystery that I think I just solved.

Nvidia has a Samsung 5nm contract, they secured years ago, they also secured 5nm contract for TSMC... Kopite7 has stated that Ada lovelace is built on only TSMC's 5nm node... With Ada Lovelace coming this year, there is no more time for an Ampere Super series, so what is Nvidia using Samsung's 5nm process node they signed a contract for?

Drake is the answer, Nvidia secured TSMC 5nm not that long ago, and in the leak we can see that the codename changed from Dane to Drake, the confirmation of this is that Dane is a banned search word (at least this is what I remember from last week). There is nothing else Nvidia has Samsung's 5nm node as a use for, and offering the node to Nintendo, also solves die size mystery that we have been working with, as this could literally be half the size of 8nm die, and would also solve how this chip is so big and still able to be portable, as we have yet to see any evidence of Thraktor's working theory on disabling SMs, (there is nothing listed in the document for the GPU other than 12SM, which is odd if the portable mode does disable them, that is a huge issue for developers trying to make games for the platform).

Anyways, again this is a theory, but I do think that it is correct, I was 50:50 on a die shrink, but now I favor it, I think Nvidia had the same issue here as when they moved from 20nm Maxwell desktop line (that never happened) to 16nm finfet, and offered the node to Nintendo (who they had worked with as early as 2014), this avoids a contract breach worth potential 100s of millions of dollars, just like it did then.

 

[Hermii]

Drake is the answer, Nvidia secured TSMC 5nm not that long ago, and in the leak we can see that the codename changed from Dane to Drake, the confirmation of this is that Dane is a banned search word (at least this is what I remember from last week). T

can you elaborate on this`?

 

[Simba1]

That's what he flat out says around the 9 min 30 sec mark. He thinks it's going to be called a Switch 4K that's a revision and not a new gen, but it won't be BC with everything, but that they'll try to patch as many games as possible.

Yeah that dont makes sense...but frankly not big surprise coming from him, I very often disagree with things he is saying in Nates podcasts.

 

[Zedark]

Actually, all Xbox games run in a Hypervisor since Xbox One, even native ones. It's not really a backwards compatibility thing specifically, it's just how Xbox games get run. This might simplify certain aspects of BC, but it's not really necessary if the OS is compatible and the graphics driver can take care of any necessary translation.

True. I'm not sure if we know whether that's the modus operandi for Nintendo as well, or if they run natively without virtualisation. If not, then it would probably be a bit of work to change things up, and it could be questioned whether it's even necessary.

 

[ShaunSwitch]

Hey, so this is just a theory, but I have a mystery that I think I just solved.

Nvidia has a Samsung 5nm contract, they secured years ago, they also secured 5nm contract for TSMC... Kopite7 has stated that Ada lovelace is built on only TSMC's 5nm node... With Ada Lovelace coming this year, there is no more time for an Ampere Super series, so what is Nvidia using Samsung's 5nm process node they signed a contract for?

Drake is the answer, Nvidia secured TSMC 5nm not that long ago, and in the leak we can see that the codename changed from Dane to Drake, the confirmation of this is that Dane is a banned search word (at least this is what I remember from last week). There is nothing else Nvidia has Samsung's 5nm node as a use for, and offering the node to Nintendo, also solves die size mystery that we have been working with, as this could literally be half the size of 8nm die, and would also solve how this chip is so big and still able to be portable, as we have yet to see any evidence of Thraktor's working theory on disabling SMs, (there is nothing listed in the document for the GPU other than 12SM, which is odd if the portable mode does disable them, that is a huge issue for developers trying to make games for the platform).

Anyways, again this is a theory, but I do think that it is correct, I was 50:50 on a die shrink, but now I favor it, I think Nvidia had the same issue here as when they moved from 20nm Maxwell desktop line (that never happened) to 16nm finfet, and offered the node to Nintendo (who they had worked with as early as 2014), this avoids a contract breach worth potential 100s of millions of dollars, just like it did then.

This sounds plausible to me. Wasn't there something said regarding Nintendo not being happy with the power draw of the supposed chip sometime late last year? I don't remember where I read it, but I'm sure I read it on here or Era. This could also be the solution to that problem.

Would be really interesting if Nintendo releases a product this year on Samsung 5nm, that's pretty cutting edge stuff. What do you think the implications are on price if this is the case? I'm more inclined to believe a $450 price tag for a 5nm device launching this year.

Also regarding CPU cores, do you think 8 cores is now far more likely? We have plenty of examples of Samsung 5nm smart phones with octocore setups (writing this on a galaxy s21 with such a setup) and whilst I appreciate phones run in burst, 8 A78s at 1ghz should be doable on such an advanced node.

It falls in line with phone manufacturing as well actually as Samsung released the s22 recently so their flagship is no longer taking up 5nm capacity.

 

[RainTech]

regarding BC, what if the answer to that is already right infront of our eyes? Look how Galaxy from the 3D collection is run on Switch. CPU runs natively GPU is emulated. Since ARM cores are compaible throughout generarions, "old" Switch games could run natively on the CPU side while the GPU uses emulation to work.

Though no idea how feasable that is...

 

[Mirev]

I really hope to see it this year. I want a more powerful Switch.

 

[Hermii]

If the 5nm theory is correct, and they basically decided to redesign T239 on 5nm, depending on when the shift happened I do see a significant delay. Even 2023 would be optimistic l.

 

[IamPeacock]

I really hope to see it this year. I want a more powerful Switch.

This!

The new hardware is powerful, which is great.
But I really hope they release it this year or early next year.

If they wait until 2024/2025, then it will fall behind again tech-wise and will be dated the time it releases.

 

[Pokemaniac]

Hey, so this is just a theory, but I have a mystery that I think I just solved.

Nvidia has a Samsung 5nm contract, they secured years ago, they also secured 5nm contract for TSMC... Kopite7 has stated that Ada lovelace is built on only TSMC's 5nm node... With Ada Lovelace coming this year, there is no more time for an Ampere Super series, so what is Nvidia using Samsung's 5nm process node they signed a contract for?

Drake is the answer, Nvidia secured TSMC 5nm not that long ago, and in the leak we can see that the codename changed from Dane to Drake, the confirmation of this is that Dane is a banned search word (at least this is what I remember from last week). There is nothing else Nvidia has Samsung's 5nm node as a use for, and offering the node to Nintendo, also solves die size mystery that we have been working with, as this could literally be half the size of 8nm die, and would also solve how this chip is so big and still able to be portable, as we have yet to see any evidence of Thraktor's working theory on disabling SMs, (there is nothing listed in the document for the GPU other than 12SM, which is odd if the portable mode does disable them, that is a huge issue for developers trying to make games for the platform).

Anyways, again this is a theory, but I do think that it is correct, I was 50:50 on a die shrink, but now I favor it, I think Nvidia had the same issue here as when they moved from 20nm Maxwell desktop line (that never happened) to 16nm finfet, and offered the node to Nintendo (who they had worked with as early as 2014), this avoids a contract breach worth potential 100s of millions of dollars, just like it did then.

That would certainly explain a few of the weirder aspects of this situation. It's definitely a theory worth considering, at least.

regarding BC, what if the answer to that is already right infront of our eyes? Look how Galaxy from the 3D collection is run on Switch. CPU runs natively GPU is emulated. Since ARM cores are compaible throughout generarions, "old" Switch games could run natively on the CPU side while the GPU uses emulation to work.

Though no idea how feasable that is...

That's basically what a lot of us have been suggesting with translating the shaders at runtime.

 

[ShaunSwitch]

If the 5nm theory is correct, and they basically decided to redesign T239 on 5nm, depending on when the shift happened I do see a significant delay. Even 2023 would be optimistic l.

Yeah this could mean a delay if the chip was redesigned for 5nm, for all we know though it could have been designed with 5nm in mind in the first place. Nvidia would have had an idea in mind of what performance Orin would have provided for a given power envelope on 8nm and Nintendo may have wanted more, so Orin and Drake could have been designed side by side with different target nodes. From a business perspective, I can't see them spending R and D designing the chip on one node, only to scrap it and redesign it.

Edit: The more I think of it, the less I see a delay as likely. The leak references GA10F, implying its the sixth version of this design, also wasn't there evidence of actual hardware tests being done in February? If that test was on an 8nm chip a redesign would take a while from now but that doesn't fall in line with Samsung 5nm being secured for now. I think Drake was always 5nm if this is indeed the case.

 

[japtor]

In terms of marketing, yeah I expect they won't call it something as straightforward as "Switch 2," but they're absolutely going to emphasize exclusive games and reasons to buy it beyond just prettier Switch games. Especially since the leaked hardware is several times more powerful than necessary for 4K Switch games.

Course the biggest Switch launch title wasn't even exclusive if you think about it. While I expect exclusives, even just a lineup of prettier new Switch games for launch window can push things too. Kinda like how they used Metroid to help push the OLED model, BotW 2 and such can be used to help push the new Switch.

 

[oracion]

Drake launch title will be Xenoblade 3.
Mark my words.

 

[PedroNavajas]

They could do it like the Switch emulators and precompile the shaders for new GPUs. A patch could consist on recompiled shaders, but I believe Nvidia can automate the process to a degree.

 

[Z0m3le]

So timeline wise, Nvidia trademarked hopper in 2019 and has been in a legal battle with dish network ever since. We know Hopper is a 5nm chip built for their data centers (like GA100, this is GH100). We also know that Nvidia didn't secure 5nm TSMC until 2021, so the 5nm they had secured in 2019 was Samsung's. Now Ada is a 2022 desktop release using TSMC's 5nm and Hopper has been delayed outside of 2022, the contract with Samsung should still exist and these contracts are expensive, can be 100s of millions of dollars to break, Dane also had some power consumption issues rumored, and the size of what Drake is revealed to be is far too big for 8nm without causing some questions about its ability to fit in the form factor, and run a low enough power draw for portable mode. 5nm Samsung solves this issue, as it could have about twice the transistors in the same mm^2 as Samsung's 8nm and offer a much lower voltage, bringing in portable power consumption, which if this theory pans out, would solve Thraktor's questions without disabling SMs (something not seen in the leaked documents and thus unlikely IMO), as developers would have to be made aware of such a thing.

I'm suggesting 5nm solves the mysteries we've had this week, and while this is just a theory, we should start hearing rumors about a die shrink soon if it did happen, the leak has seemingly compelled NDA holders to leak info I believe, and stuff from Nate will likely reflect it this week in his video (I assume it's this week).

 

[Z0m3le]

Yeah this could mean a delay if the chip was redesigned for 5nm, for all we know though it could have been designed with 5nm in mind in the first place. Nvidia would have had an idea in mind of what performance Orin would have provided for a given power envelope on 8nm and Nintendo may have wanted more, so Orin and Drake could have been designed side by side with different target nodes. From a business perspective, I can't see them spending R and D designing the chip on one node, only to scrap it and redesign it.

Edit: The more I think of it, the less I see a delay as likely. The leak references GA10F, implying its the sixth version of this design, also wasn't there evidence of actual hardware tests being done in February? If that test was on an 8nm chip a redesign would take a while from now but that doesn't fall in line with Samsung 5nm being secured for now. I think Drake was always 5nm if this is indeed the case.

The move to 5nm would have happened around a year ago, when Nvidia secured TSMC's 5nm (rumored in feb 2021), this leaves about 20 months for an october launch this year, that is plenty of time to shrink a chip from the same fab company. I don't think a delay is needed to exist, especially when rumors of this FY has not changed, though Nate's video could correct me this week, I do believe it will instead restate this holiday/early next year for release window.

 

[PedroNavajas]

So, we are assuming that part of the $9B that Nvidia spent on 5nm allocation is being used for the Switch 4k? Sorry, but I have hard time believing that. How much would have cost for Nintendo to get some of that? $1B? The cost is too high IMHO.

I think Nintendo would have done their own deal for their chips.

 

[Z0m3le]

So, we are assuming that part of the $9B that Nvidia spent on 5nm allocation is being used for the Switch 4k? Sorry, but I have hard time believing that. How much would have cost for Nintendo to get some of that? $1B? The cost is too high IMHO.

I think Nintendo would have done their own deal for their chips.

No completely different timelines, we actually know Nvidia was designing a 5nm chip in 2019, they didn't secure TSMC 5nm until 2021 with that money, these are separate things.

 

[Simba1]

Just curious though, why didn’t Nintendo release the OLED at $299 and discount the regular hybrid to $249? In your opinion.



Best case scenario at launch:

4K Switch - $449
OLED - $299
Switch - $249
Lite - $149

This would be until Mariko chips and LCD screens last of course.



It’s fine to go high since their support and sales of the current devices isn’t going away any time soon. The whole point of positioning as a mid gen upgrade option (not gen break)

I don’t see how you can argue a $399 price. All that new, custom hardware inside…is costlier than $50. They aren’t going to lose money on each 4K Switch sold, that’s for sure. They don’t need a quick uptake on this model.



If the 4K Switch is going to act as a gen breaking successor, why bother selling the OLED another 2-3 years? The 4K Switch is iterative and BC. You don’t keep last gen consoles around that long anymore.



This is the template for a conventional console lifecycle, sure.



Why? Why not just say it’s a model that enhances Switch games? Why make people think it’s pointless to buy an OLED Switch next year if it’s going to lose support by Nintendo?

My argument is that it makes no business sense to market this expensive model as a gen breaking successor.



Again, Nintendo doesn’t care how fast the 4K Switch is adopted. They don’t have to care. It doesn’t matter what model you buy in the next ~4 years. As long as it keeps you engaged in the services and software they release.



As far as I know, this huge power/feature difference is mostly going to be used to make Switch games run at 4K/60fps instead of 900p/30fps. And a few other graphical effects.

The on-paper specs doesn’t really matter when it comes to the resultant Nintendo product. My wife will notice that Animal Crossing will look different on the 4K Switch vs her Switch…but she won’t think it’s THAT big a difference.

Strange question because its very obvious, there is chip shortage and Switch shortages and Switch selling great in any case, so didnt make any sense to make price cut around OLED launch.


Disagree, that would be best case scenario, $399 is obviously sweet spot for new stronger hardware, and having on mind that there is no big difference between regular Switch and OLED Switch, there is no need to keep selling them in same time for longer time (only if chip shortages makes problems), also I dont see Switch Lite going from $199 to $149 so fast (in around next year).


Thing is that there will not be gen break (regardless we talk about positioning like real next gen or like much stronger current gen upgrade), difference mostly comes to how Nintendo will market it, and at this point and having on mind expected huge differences in any case compared to current Switch models and timing, makes more sense to market it like next gen Switch console.
But point is that, current Switch models will keep selling alongside this next gen Switch and games will keep being cross gen for quite time (at least around 2 years).


Simple, Switch is still selling at launch price point ($299), only reason why Nintendo didnt make price is because it selling great in any case at that price point even they could selling at $249 few years ago and probably again selling at profit. OLED was launched and its basically little improved Switch, so because Switch is selling great in any case and OLED is a little improved Switch, they didnt want make price cut so they put higher price point for OLED for time being.
But I am sure that plan was that OLED actually should be replacement for current Switch model, because doesnt make too much sense to sell Switch and OLED Switch in same time, because basically only real difference is just a OLED screen, but because chip shortages they need to sell them at same time for a while.


So I wrote you that OLED actually should be replacement for main Switch model, and point of keep selling Switch OLED and Switch Lite after "Switch 2" launch is different price points of hardware at least until more Switch 2 revisions and price points are out (Nintendo always had very different price point hardware on market and no way they will having only $399 option all the sadden) and in same time having software sales for at least 110-120m current users that will not upgrade next gen Switch right a way.


Actually its not template for a conventional console lifecycle if we talking about Nintendo, moment Nintendo release new hardware Nintendo was stop releasing all its games for previous platform (going from Wii to Wii U, DS to 3DS..), moment they were announced new platform their support was going significtly down.


Why not just say it’s a model that enhances Switch games?
Simple because it much more than that, its full next gen Switch (based on current rumors) in any way, not to mention that at launch will you have some exclusive games (Nate said some 3rd party games), and by time more and more games will be exclusive until games will stop be developed for current models.
And that's very different to PS4 Pro for instance, even to New 3DS.

Thing is that you are only person that saying gen brake successor, this is not gen break successor like Wii to GC or Wii U to GC was, today gen borders are blurrier/thinner than ever before, and thats ahived by conosle manificaturere stick to same tech (in case of Sony/MS, AMD), that alow full BC and continues support for last gen consoles, and this Switch successor would be similar like PS5 to PS4 (with big difference that Nintendo will keep selling some of current models) or similar like GBC to GB.
Talking about games, latest huge and great AAA game, Elden Ring, it was launched on XB1/PS4/PS5/XsX-S despite PS5/XsX-S were launched 1.5 years ago.


I am pretty sure they will not be happy if this next gen Switch is not selling good, we talking about real next gen console with all R&D and manufacturer costs, this thing cant be comparible in any case with OLED type revision where differences are cosmetics, we talking about new CPU, GPU, RAM, internal memory, tools, APIs, dev kits. DLSS...and who know what other new features/differences.


At 1st, but how time goes, more and more games will specify be made for this new Switch hardware and than will be much more noticeable difference because there will not be cross gen games.
Again, PS5 to PS4 is good example.

 

[Simba1]

This will be covered in my upcoming episode.

When be released on YouTube?

 

[Simba1]

If the 5nm theory is correct, and they basically decided to redesign T239 on 5nm, depending on when the shift happened I do see a significant delay. Even 2023 would be optimistic l.

Yeah, I was saying before there is chance that chip redesign and chip shortages could delay new Switch hardware, but few people reply to me that's not realistic and that is almost certain 8nm and coming this year.


I personally dont expect this new hardware launch this year, I expecting time frame from March 2023. to March 2024.

 

[Hermii]

Yeah, I was saying before there is chance that chip redesign and chip shortages could delay new Switch hardware, but few people reply to me that's not realistic, that is almost certain 8nm and coming this year.

I know, I was one of those people Happy to eat crow.

 

[Simba1]

I know, I was one of those people Happy to eat crow.

There were few people, but when one person reply me that there is also chanche that will be hit in meteorit, I was lost.

People should constantly have on mind not only that this is talk based mostly on rumors but point that plans constantly changing,
especially with Covid, chip shortages, inflation, now war in Ukraine...so nothing is certain or set in stone.

 

[Gianko]

....about BC...if this new model will share the same library with the actual one (at least for 2 years), how are the y gona publish their games on retail? i mean same case of actual Switch games with just a little banner for the new console compatibility? and after 2 years, it will be different?

 

[mariodk18]

....about BC...if this new model will share the same library with the actual one (at least for 2 years), how are the y gona publish their games on retail? i mean same case of actual Switch games with just a little banner for the new console compatibility? and after 2 years, it will be different?

same way GBC and n3DS were handled

 

[Z0m3le]

There were few people, but when one person reply me that there is also chanche that will be hit in meteorit, I was lost.

People should constantly have on mind not only that this is talk based mostly on rumors but point that plans constantly changing,
especially with Covid, chip shortages, inflation, now war in Ukraine...so nothing is certain or set in stone.

these fab contracts wouldn't be changed by stuff like chip shortages, which are just the inability to add more orders, not fill existing orders... Ukraine is a very real conflict that could absolutely delay anything, including slowing or even stopping production of current devices, but that uncertainty is completely out of any company's control.

 

[NineTailSage]

Yeah, I was saying before there is chance that chip redesign and chip shortages could delay new Switch hardware, but few people reply to me that's not realistic and that is almost certain 8nm and coming this year.


I personally dont expect this new hardware launch this year, I expecting time frame from March 2023. to March 2024.

I think Z0m3le is pretty much asserting that this might have been something that happened behind the scene and stuff like the OLED not being a widely known item could also be a sign of this as well. Insiders were all discussing more powerful Switch hardware in development for some time, so moving to a different node could definitely justify where we are today...

 

[Gianko]

same way GBC and n3DS were handled

i think it's different this time because we have one cart running on two different console. N3ds and 3ds were the same console and shared the same games for 99%

 

[mariodk18]

How would 5nm change our expectations for the CPU?

 

[Z0m3le]

How would 5nm change our expectations for the CPU?

it would just be more likely that it is 8 cores, with a higher clock, maybe 2GHz if 5nm. A78 was designed for 5nm, and power consumption would be under 3 watts for 2GHz across 8 cores.

 

[ShaunSwitch]

it would just be more likely that it is 8 cores, with a higher clock, maybe 2GHz if 5nm. A78 was designed for 5nm, and power consumption would be under 3 watts for 2GHz across 8 cores.

This would be an enormous leap over the CPU in the current switch, exciting if true. I wonder how close an 8 Core A78C setup at 2GHZ would get to the 8 Core 3.5GHZ setup in the ps5? Are we now talking somewhere around 50% of the power?

I can see this being great for facilitating simultaneous software releases across all platforms and might help mitigate the "late port" syndrome that Nintendo suffers.

Edit: Actually, this also could change what the max possible dock clock of the GPU is now too. Could we be getting a 3-4 TFLOP docked performance profile.

If we were looking at 1GHZ as max clock before.....

1GHZ = Around 3 Tflop
1.2GHZ = 3.7 Tflop
1.4 GHZ = 4.3 Tflop
1.5GHZ = 4.6 Tflop

I think 1.2 GHZ should be perfectly feasible going from Samsung 8nm to Samsung 5nm.

 

[Kenka]

Drake launch title will be Xenoblade 3.
Mark my words.

Bookmarked.

Spoiler

;-)

 

[Dakhil]

Nvidia has a Samsung 5nm contract, they secured years ago, they also secured 5nm contract for TSMC... Kopite7 has stated that Ada lovelace is built on only TSMC's 5nm node... With Ada Lovelace coming this year, there is no more time for an Ampere Super series, so what is Nvidia using Samsung's 5nm process node they signed a contract for?

Drake is the answer, Nvidia secured TSMC 5nm not that long ago, and in the leak we can see that the codename changed from Dane to Drake, the confirmation of this is that Dane is a banned search word (at least this is what I remember from last week). There is nothing else Nvidia has Samsung's 5nm node as a use for, and offering the node to Nintendo, also solves die size mystery that we have been working with, as this could literally be half the size of 8nm die, and would also solve how this chip is so big and still able to be portable, as we have yet to see any evidence of Thraktor's working theory on disabling SMs, (there is nothing listed in the document for the GPU other than 12SM, which is odd if the portable mode does disable them, that is a huge issue for developers trying to make games for the platform).

I think Samsung's 5LPE process node's probably being used if Nintendo and Nvidia are indeed using one of Samsung's 5 nm** process nodes for the fabrication of Drake. Samsung's 5LPE process node seems decent enough where Qualcomm used Samsung's 5LPE process for the fabrication of the Snapdragon 888 and the Snapdragon 888+ for the entire lifecycle of the Snapdragon 888 and the Snapdragon 888+. And Qualcomm's currently using Samsung's 5LPE process node for the fabrication of the Snapdragon 8cx Gen 3. Assuming the rumour about Qualcomm using a custom variation of Samsung's 5LPP process node called Samsung's 4LPX process node for the fabrication of the Snapdragon 8 Gen 1 is true, the rumoured yield rate for Samsung's 5LPP process node is probably too problematic for Nintendo and Nvidia to consider using Samsung's 5LPP process node for the fabrication of Drake.

I think there's a possibility Nvidia also uses Samsung's 5LPE process node for the fabrication of Atlan alongside the fabrication of Drake, considering that the sampling of Atlan is planned for 2023. And I imagine when Atlan's being sampled is when Nvidia needs to make a final decision on which process node to use for the fabrication of Atlan, especially since being certified to be in compliance with automotive safety standards, such as ISO 26262, is an extremely time consuming process.

 

[Z0m3le]

This would be an enormous leap over the CPU in the current switch, exciting if true. I wonder how close an 8 Core A78C setup at 2GHZ would get to the 8 Core 3.5GHZ setup in the ps5? Are we now talking somewhere around 50% of the power?

I can see this being great for facilitating simultaneous software releases across all platforms and might help mitigate the "late port" syndrome that Nintendo suffers.

Edit: Actually, this also could change what the max possible maxed dock clock of the GPU is now too. Could we be getting a 3-4 TFLOP docked performance profile.

If we were looking at 1GHZ as max clock before.....

1GHZ = Around 3 Tflop
1.2GHZ = 3.7 Tflop
1.4 GHZ = 4.3 Tflop
1.5GHZ = 4.6 Tflop

I think 1.2 GHZ should be perfectly feasible going from Samsung 8nm to Samsung 5nm.

Per core performance would be much higher than 50% of PS5's CPU I think, Thraktor or someone else more familiar with the benchmarks can detail it, but I think it would be safer to assume a 65-70% of the CPU single thread performance.

And I think its better to think of minimum docked clocks at 1GHz for the GPU than expecting 1.2GHz, but it is reasonable if 5nm is indeed used, we should be hearing a lot more about Drake this year, GDC is this month and with this leak, people will talk, so rumors will start popping up, Nate just last week had no updates and now after this leak, he has a big video coming with more information, so not only is he able to confirm what he has been saying, but he seemingly has found more information.

I'll be doing an update on Drake later this week on my youtube channel, it won't have any insider information, that is something I want to avoid people expecting from me, but I'll explain known specs, features and go over what we can understand about the performance we are starting to see out of Drake. I should also have a DLSS/Tensor core tech video up this week... This is going to be an exciting month for Nintendo hardware enthuasts.

 

[Thraktor]

So timeline wise, Nvidia trademarked hopper in 2019 and has been in a legal battle with dish network ever since. We know Hopper is a 5nm chip built for their data centers (like GA100, this is GH100). We also know that Nvidia didn't secure 5nm TSMC until 2021, so the 5nm they had secured in 2019 was Samsung's. Now Ada is a 2022 desktop release using TSMC's 5nm and Hopper has been delayed outside of 2022, the contract with Samsung should still exist and these contracts are expensive, can be 100s of millions of dollars to break, Dane also had some power consumption issues rumored, and the size of what Drake is revealed to be is far too big for 8nm without causing some questions about its ability to fit in the form factor, and run a low enough power draw for portable mode. 5nm Samsung solves this issue, as it could have about twice the transistors in the same mm^2 as Samsung's 8nm and offer a much lower voltage, bringing in portable power consumption, which if this theory pans out, would solve Thraktor's questions without disabling SMs (something not seen in the leaked documents and thus unlikely IMO), as developers would have to be made aware of such a thing.

I'm suggesting 5nm solves the mysteries we've had this week, and while this is just a theory, we should start hearing rumors about a die shrink soon if it did happen, the leak has seemingly compelled NDA holders to leak info I believe, and stuff from Nate will likely reflect it this week in his video (I assume it's this week).

I'd be surprised if Hopper was ever intended for anything other than TSMC N5, given the positioning and price they'll be charging for the chip it would have always made sense to use the most advanced node available. I think the reason Nvidia didn't secure TSMC N5 capacity is that they didn't have to. Nvidia have never had to pre-pay TSMC to secure capacity before, and it's only because of the chip shortage that TSMC are now in the negotiating position to demand those kinds of pre-payments.

That being said, I'm definitely not ruling out Samsung's 5nm process for Drake, given the size of the GPU we're looking at. I don't even necessarily think it would have changed during development, it could just be that kopite7kimi knew Orin was on 8nm, knew T239 was part of the same chip family, and was just incorrect on the manufacturing process of T239. I wouldn't even completely rule out TSMC N5 at this point.

Edit:

So have we heard or discussed this feature yet possibly being included?
It was definitely present in A100 GPU but I wonder if this will make its way over to Drake also...




"The NVIDIA Ampere architecture adds Compute Data Compression to accelerate unstructured sparsity and other compressible data patterns. Compression in L2 provides up to 4x improvement to DRAM read/write bandwidth, up to 4x improvement in L2 read bandwidth, and up to 2x improvement in L2 capacity."

NVIDIA Ampere Architecture In-Depth | NVIDIA Technical Blog

Today, during the 2020 NVIDIA GTC keynote address, NVIDIA founder and CEO Jensen Huang introduced the new NVIDIA A100 GPU based on the new NVIDIA Ampere GPU architecture. This post gives you a look…

developer.nvidia.com

I think the reason we haven't heard about it (and the reason it's not in consumer Ampere cards) is in the name; it's only really useful for compute workloads. Nvidia don't seem to provide much details on this anywhere, but my guess is that it's a simple Huffman coding or something like that. For compute workloads this could be very useful, as the data is often sparse and low-entropy, but for gaming it's not that much use, as almost all the data is either high-entropy and/or already compressed. Texture data, for example, is already compressed. Framebuffer objects are also compressed on more recent Nvidia GPUs (and probably high-entropy anyway, in the scheme of things). Basically there isn't a lot of scope left for compressing the kind of graphics data that's stored in memory in a gaming use-case.

 

[SiG]

So, uh...

Does that mean that the whole Switch successor switching nodes thing was possible because Nvidia wanted to secure fab contracts? Or was it because of the untimely chip shortage that forced Nvidia, and thus Nintendo, to jump nodes (in order to have more availability)?

 

[Z0m3le]

I'd be surprised if Hopper was ever intended for anything other than TSMC N5, given the positioning and price they'll be charging for the chip it would have always made sense to use the most advanced node available. I think the reason Nvidia didn't secure TSMC N5 capacity is that they didn't have to. Nvidia have never had to pre-pay TSMC to secure capacity before, and it's only because of the chip shortage that TSMC are now in the negotiating position to demand those kinds of pre-payments.

That being said, I'm definitely not ruling out Samsung's 5nm process for Drake, given the size of the GPU we're looking at. I don't even necessarily think it would have changed during development, it could just be that kopite7kimi knew Orin was on 8nm, knew T239 was part of the same chip family, and was just incorrect on the manufacturing process of T239. I wouldn't even completely rule out TSMC N5 at this point.

Fair, though Ada has been in the works for a while and Nvidia had problems securing TSMC for Ampere and went with Samsung, (at least it was reported that AMD bought up all of TSMC's remaining 7nm node for the console launches before the pandemic chip shortage began.

I am leaning to 5nm for the same reasons you mention, but I do think Nvidia has probably had Samsung 5nm allocation incase they weren't able to secure TSMC, also just moving away from Samsung with Ada is going to leave a hole in Samsung's chip manufacturing, where I'd expect them to have hopes to secure Ada on their process, so Drake could have just been a happy accident there. I do remember some power issues being brought up too, which would have been solved with 5nm, and we still know the codename changed over the past two years as a data leak ban search word is Dane, which was identified to be T239 by Kopite7 around this time last year. It will be interesting to see this chip revealed.

 

[Dakhil]

I wouldn't even completely rule out TSMC N5 at this point.

I don't think TSMC's N6 process node can be ruled out either, especially since I imagine demand for TSMC's N6 process node isn't as ridiculously high as demand for TSMC's N5 process node.

So, uh...

Does that mean that the whole Switch successor switching nodes thing was possible because Nvidia wanted to secure fab contracts? Or was it because of the untimely chip shortage that forced Nvidia, and thus Nintendo, to jump nodes (in order to have more availability)?

Probably leaning more towards the former since changing which process node to use is not a process that can be done in a matter of a couple of months. And I imagine having enough capacity isn't really a problem with Samsung's 8N process node.

 

[ILikeFeet]

I think it's important to remember a lot of what kopite7kimi said about the t239 was in respect to Orin. It's possible that includes the node.

As far as NateDrake's sources goes, node isn't something they need to know so they could be regurgitating surface info, like "ampere = 8nm".

 

[SiG]

What if everyone is wrong and this isn’t a new Switch but rather a portable Wii U :shock:

That's...what they said about the Switch originally when it debuted.

They said Nintendo was doomed and everything.

Instead Nintendo got DOOM and the rest was history.