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

[Giancarlo]

I really don't understand where does this talk or fear of Nintendo not being prepared for next-gen comes? They way Nintendo develop their games and the assets or effects in use in their games are in the same ballpark of quality and features as PS4 or XOne games. Just because Nintendo has a lower performance platform doesn't mean they're using or being used to outdated development practices. The jump from Switch development to Switch 2 development isn't a big and wide jump like going from SD development (Wii) to HD development (Wii U).

I suspect a lot of this bias and fear come from western development, whose pipelines takes way longer, are very expensive and have set the standard of high fidelity visuals.

However, if we look at a more comparable, to Nintendo, AAA studio from Japan, Capcom, we see that their games look just as good as most AAA western games while costing a fraction of said western games and coming out very quickly, without much development issues.

My point is: Just because Switch 2 will present a big tech leap, that doesn’t mean that Nintendo pipeline will suddenly stall. Nor does it mean Nintendo will need to re-learn their development practics because everything became outdated or that games will take a long time to be made and will be way more expensive.

The people at Nintendo are professionals and are at the forefront of technology R&D. Whatever comes with new tech leap won't take them by surprise and mess up their development environment.

Nintendo struggled when they went to HD development on Wii U, now the same could be happening(struggling with 4K development),

 

[Discostew]

Dunno if anyone here have or access to a laptop that have a 25W RTX 2050, but having played a little bit with my friend's Dell Vostro RTX 2050, plus some quick dirty napkin math (let's pretend everything here scales linear) I manage to devise some numbers.
At max load (I'm using Plague Tale Requirem here with everything set to Medium to avoid VRAM thrasing and no upscaling at 1080p), the RTX 2050 25W consume around 30W (not sure if this include the GDDR6, it was the number reported by RivaTuner statistics, let's pretend around 5W of that is going to the G6 modules), and clock speed were hovering around 1267MHz max.
The T239 GPU, assuming on 8N as the 2050, is very roughly 3/4 of that GPU (just roughly based on pure compute hardware). Let's keep the clock speed of 1.27GHz, we have 30W on the RTX 2050 minus 5W of the GDDR6, multiply by 0.75, then the T239 GPU will consume 18.75W. If we the multiply by 1/1.267, we have the T239 GPU at 1GHz consume ~15W, just the GPU alone.
I don't know if this have any meaning, but looking at the numbers, I'm kinda on the fence whether T239 will be on 8N or 4N. On one hand if we ignore all other factors that increase the power budget of the chip (the CPU, the FDE, memory controllers...) then 15W seems pretty reasonable. Hell, with some optimizations that Nvidia brings from Ada we can get lower than that, and like I said above the T239 is not strictly 3/4 of the 2050 (for example T239 only have half the ROPs of the 2050, and half the L2 cache too), so there's some wiggle room there. But that's not the case at all, because that's just the GPU, add everything else then we have a problem. Unless Nintendo clock this thing low as hell.
Again this is just pure curiosity that make me do this research myself, and a lot of in between things can screw up the number pretty hard. Let me know what do you guys think.

Just a reminder that clock frequency and power consumption do not scale 1:1. Going from 500Mhz at X watts to 1Ghz requires more than 2 * X watts. As clocks increase linearly, power consumption curves upward. Or, if power consumption increases linearly, clock frequency shifts closer to flatlining.

 

[Dakhil]

CPU at 2.5GHz will be more than perfect since A78C is really good

I don't think there's any guarantee of Drake's CPU running at a frequency as high as 2.5 GHz, especially if Drake's CPU has to run on the same frequency in both handheld mode and TV mode, similar with the Nintendo Switch, due to how difficult scaling down the CPU frequency is compared to scaling down the GPU frequency.

 

[kingman]

4K and DLSS was never confirmed for Zelda, just higher framerate and resolution, and one source said it was focused on load times

Sources: Nintendo showed Switch 2 demos at Gamescom | VGC

The company is understood to have held private tech demos, ahead of an expected 2024 launch…

www.videogameschronicle.com

...the hell? I'm having a helluva Mandela Effect moment cause I could have sworn those original Eurogamer/VGC reports confirmed 60 fps/4k. D:

Regardless, we did hear subsequent reports saying it was indeed 60 fps/4k, and DF even did their T239 tests with that in mind. It also makes fairly logical sense that even if the reports only said "higher resolution and frame rates" we can deduce that it would obviously mean 60 fps for sure (very doubtful Nintendo would bother with a weird number like 40 fps) , and just going up to 1440p would also be a weird number, leaving 4k as the only proper choice.

 

[Mr Swine]

I still don’t know why Nintendo is (possibly) going to announce Switch 2 in June and then release it 9 months later. Wouldn’t it be better to do it in October since it would be 6 months from there?

 

[Dakhil]

...the hell? I'm having a helluva Mandela Effect moment cause I could have sworn those original Eurogamer/VGC reports confirmed 60 fps/4k. D:

NateDrake did mention that he heard about the Breath of the Wild demo at Gamescom 2023 running at 4K at 60 fps via DLSS.

So I think people could be conflating what NateDrake said with what Video Games Chronicle and Eurogamer said.

 

[CeramicPigeon]

Nintendo is probably willing to do so, however I want to say these are the only two examples I can find. There's probably a chance that Nintendo has another one in the past, but there's a case to be made that they might be a bit more unwilling in this specific case due to how disastrous the Wii U went. Really depends on how confident they are with the Switch 2's success.

The Wii U failing had nothing to do with it being sold at a loss though. It certainly didn't help things, but it wasn't a paramount reason as to why it failed.

 

[darthdiablo]

I don't think there's any guarantee of Drake's CPU running at a frequency as high as 2.5 GHz, especially if Drake's CPU has to run on the same frequency in both handheld mode and TV mode, similar with the Nintendo Switch, due to how difficult scaling down the CPU frequency is compared to scaling down the GPU frequency.

Thanks for linking the Dec 2016 Eurogamer article. I haven't seen it before till now.

Fascinating retrospective read from our past, during the period between October 2016 announcement and March 2017 launch.

Edit: LOL some of the article comments are funny to read now:

I dont understand how such a weak device is running Zelda BotW.

Fuck Switch, I'm keeping my Wii U

 

[SoldierDelta]

The Wii U failing had nothing to do with it being sold at a loss though. It certainly didn't help things, but it wasn't a paramount reason as to why it failed.

That wasn't really what I was going for but good point.

I was mainly thinking that a console is a big risk by itself when selling at a loss, even if you manage to succeed. Nintendo learned first-hand what the punishment for selling a console at a loss was, so there's a point to be made (one that I sadly failed to convey well) that they could suffer similar consequences if the Switch 2 (for whatever reason) also suffers.

Regardless, the ultimate statement is fairly clear... Nintendo has made a console in the past that sold at a loss. Hell, Nintendo deliberately cut the price of the 3ds, in spite of the loss-making of this move, just so the console didn't die at launch due to the price tag, this isn't something Nintendo is too afraid of.

 

[HankyPanky]

I still don’t know why Nintendo is (possibly) going to announce Switch 2 in June and then release it 9 months later. Wouldn’t it be better to do it in October since it would be 6 months from there?

all comes down to when production kicks off. having said that June still feels early.

 

[ILikeFeet]

Nintendo struggled when they went to HD development on Wii U, now the same could be happening(struggling with 4K development),

I think there are considerable differences between then and now. 4K development is largely HD development with higher resolution assets. the assets alone aren't the problem, but the sheer volume of assets for that "AAA" feel.

 

[CeramicPigeon]

That wasn't really what I was going for but good point.

I was mainly thinking that a console is a big risk by itself when selling at a loss, even if you manage to succeed. Nintendo learned first-hand what the punishment for selling a console at a loss was, so there's a point to be made (one that I sadly failed to convey well) that they could suffer similar consequences if the Switch 2 (for whatever reason) also suffers.

Regardless, the ultimate statement is fairly clear... Nintendo has made a console in the past that sold at a loss. Hell, Nintendo deliberately cut the price of the 3ds, in spite of the loss-making of this move, just so the console didn't die at launch due to the price tag, this isn't something Nintendo is too afraid of.

They have even less reason to be afraid of it now considering how diversified their business has become since the Wii U days. They have plenty of avenues to absorb any reasonable loss they could sell the Switch 2 at. Not to mention that over time, manufacturing will become easier and cheaper and they will likely plan out a timeline to expect manufacturing to turn a profit, as is common.

 

[darthdiablo]

all comes down to when production kicks off. having said that June still feels early.

It does seem a bit early to me too.

Curious if anyone happen to remember if Switch was mentioned at all to shareholders before October 2016 announcement, considering the October 2016 -> March 2017 is during FY2017 (which begins April 1, 2016). Did the original FY2017 forecast include mentions of "new hardware" at all?

 

[Dimensio]

I still don’t know why Nintendo is (possibly) going to announce Switch 2 in June and then release it 9 months later. Wouldn’t it be better to do it in October since it would be 6 months from there?

Many people are expecting at least a simple PR announcement between March-May similar to the 3DS one in March 2010 or the Switch one in April 2016 because Nintendo will have to give the Forecast for the April 2024-march 2025 fiscal year.

It does seem a bit early to me too.

Curious if anyone happen to remember if Switch was mentioned at all to shareholders before October 2016 announcement, considering the October 2016 -> March 2017 is during FY2017 (which begins April 1, 2016). Did the original FY2017 forecast include mentions of "new hardware" at all?

I don't remember if they talked about it in any investors QA or forecast at the time, but there was this tweet in April 2016:

 

[Dakhil]

MILD travel to other universe in his new Q&A, he said there is exist AMD APU for Switch 2 on 6nm XD

I'm sorry about talking about Moore's Law is Dead, but I find his claims with respect to AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware, very hard to believe.
One, I believe ~99.9% backwards compatibility with the Nintendo Switch is impossible without spending a ridiculous amount of money for a 6 nm* Arm based SoC from AMD.
Two, why talk now about AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware? Why not much earlier, especially since there are plenty of AMD insiders (e.g. Kepler, etc.) more than willing to talk? I believe the simple answer is that those claims were never really reliable to begin with.
* → a marketing nomenclature used by all foundry companies

 

[Gerald]

I don't think there's any guarantee of Drake's CPU running at a frequency as high as 2.5 GHz, especially if Drake's CPU has to run on the same frequency in both handheld mode and TV mode, similar with the Nintendo Switch, due to how difficult scaling down the CPU frequency is compared to scaling down the GPU frequency.

Yeah im mean it will be in both modes, CPU can’t be downclocked, and these clocks im mean on TSMC 4N

 

[Gerald]

I'm sorry about talking about Moore's Law is Dead, but I find his claims with respect to AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware, very hard to believe.
One, I believe ~99.9% backwards compatibility with the Nintendo Switch is impossible without spending a ridiculous amount of money for a 6 nm* Arm based SoC from AMD.
Two, why talk now about AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware? Why not much earlier, especially since there are plenty of AMD insiders (e.g. Kepler, etc.) more than willing to talk? I believe the simple answer is that those claims were never really reliable to begin with.
* → a marketing nomenclature used by all foundry companies

Kepler will ofc said about if if it will exist in some stage, MILD is just full of BS

 

[Hermii]

I'm sorry about talking about Moore's Law is Dead, but I find his claims with respect to AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware, very hard to believe.
One, I believe ~99.9% backwards compatibility with the Nintendo Switch is impossible without spending a ridiculous amount of money for a 6 nm* Arm based SoC from AMD.

I mean even if they were 95% sure they would go with Nvidia anyway, it doesn't hurt to see what other providers would offer. If only to keep Nvidia on their toes.

Two, why talk now about AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware? Why not much earlier, especially since there are plenty of AMD insiders (e.g. Kepler, etc.) more than willing to talk? I believe the simple answer is that those claims were never really reliable to begin with.
* → a marketing nomenclature used by all foundry companies

Did he ever say a chip physically existed? They chip woudnt have to be anything more than theoretical for AMD to make a bid.

 

[darthdiablo]

Many people are expecting at least a simple PR announcement between March-May similar to the 3DS one in March 2010 or the Switch one in April 2016 because Nintendo will have to give the Forecast for the April 2024-march 2025 fiscal year.

I don't remember if they talked about it in any investors QA or forecast at the time, but there was this tweet in April 2016:

Thanks. I did find a mention of "NX" in their financial IR release for FY2016 end and FY2017 forecast:

However I know NX is a bit different because on couple of occasions Nintendo heads felt compelled to reveal thing a bit more & earlier than usual because Wii U was performing extremely poorly up to that point and there were public fear that Nintendo was entering the mobile gaming app business or going 3rd party, etc.

The FY2016 year end and FY2017 forecast (both in one financial IR release) was released April 28, 2016.

So that tweet you shared either came a day before the financial IR release, or on the same day.

 

[Gerald]

Also Jensen said that they co-operation with Nintendo will exist for more than 20 years in 2016

 

[SoldierDelta]

Thanks. I did find a mention of "NX" in their financial IR release for FY2016 end and FY2017 forecast:



However I know NX is a bit different because on couple of occasions Nintendo heads felt compelled to reveal thing a bit more & earlier than usual because Wii U was performing extremely poorly up to that point and there were public fear that Nintendo was entering the mobile gaming app business or going 3rd party, etc.

It's always worth reminding people about how much the Wii U was "beyond redeeming" at that stage. Not surprising at all that Nintendo essentially said "We tried and we failed, we're making a new system". Nintendo trimming down the fat of the initial announcement to final release is not a shock in this case. An initial tease in April with a full release in Q1 next year is not a shock to anyone unless they've never heard of a "console generation" anywhere in their life.

 

[JoshuaJSlone]

Curious if anyone happen to remember if Switch was mentioned at all to shareholders before October 2016 announcement, considering the October 2016 -> March 2017 is during FY2017 (which begins April 1, 2016). Did the original FY2017 forecast include mentions of "new hardware" at all?

Just looking in the "Consolidated Financial Forecast" (or equivalent) section of their main quarterly release, here's how mention of it changed over time.

February 2016:
(nothing)

April 2016:

For our dedicated video game platform business, Nintendo is currently developing a gaming platform codenamed “NX” with a brand-new concept. NX will be launched in March 2017 globally.

July 2016:

The dedicated gaming platform, code-named “NX,” which is currently under development and incorporates a brand-new concept, is scheduled for launch in March 2017 globally.

October 2016:

We have also revealed that our new-concept gaming system, previously referred to using the development codename NX, will be released under the official name of Nintendo Switch. Nintendo Switch offers a wholly new concept as a home gaming system that can switch between different styles of play. We are planning to release it worldwide in March 2017.

January 2017:

We will launch Nintendo Switch, a new home gaming system that diversifies the ways you can play video games, on March 3, 2017 in Japan and overseas. Our launch lineup includes first-party software such as The Legend of Zelda: Breath of the Wild and 1-2-Switch as well as strong third-party titles. The Legend of Zelda: Breath of the Wild will release for Wii U on the same day.

In the section where they usually give hardware shipment numbers and forecasts they never gave an explicit forecast for shipment numbers, but through this period included statements like

The dedicated video game hardware codenamed "NX" is scheduled for launch in March of 2017. Its estimated hardware and software sales units are reflected in the financial forecast ending March 31, 2017, however, we are not disclosing the specific information as it is provisional.

or

A new home gaming system "Nintendo Switch" is scheduled for launch in March of 2017. Its estimated hardware and software sales units are reflected in the financial forecast ending March 31, 2017.

 

[Dakhil]

I mean even if they were 95% sure they would go with Nvidia anyway, it doesn't hurt to see what other providers would offer. If only to keep Nvidia on their toes.

I don't doubt that Nintendo's always looking at what other companies offer. What I'm trying to say is that I believe the simplicity of incorporating backwards compatibility is one of the main factors Nintendo's looking for from other companies. And I think there are huge challenges with incorporating backwards compatibility with AMD, especially since I don't believe AMD's as good as Nvidia when software's concerned. But anyway, I consider looking at what other companies offer to be very different from AMD actually bidding to be Nintendo's SoC provider for Nintendo's next hardware.

Did he ever say a chip physically existed? They chip woudnt have to be anything more than theoretical for AMD to make a bid.

I didn't watch Moore's Law is Dead livestream yesterday. But I believe that's what Moore's Law is Dead said, going by what Gerald said.

 

[darthdiablo]

It's always worth reminding people about how much the Wii U was "beyond redeeming" at that stage. Not surprising at all that Nintendo essentially said "We tried and we failed, we're making a new system". Nintendo trimming down the fat of the initial announcement to final release is not a shock in this case. An initial tease in April with a full release in Q1 next year is not a shock to anyone unless they've never heard of a "console generation" anywhere in their life.

Less about being shocked, more about if we think we'll see a pattern emerge here. March 2025, the rumored Switch 2 launch, if it holds, is in the upcoming fiscal year (FY2025).

I wouldn't be shocked if Nintendo mentions something because, well, it's part of FY2025 forecast, even if only to mention "new hardware", nothing more.

But I also wouldn't be shocked if Nintendo doesn't mention anything either.

It's an informative datapoint. If the former happens - this would mean Nintendo would tend to be open about the upcoming fiscal year (something we can use as a datapoint going forward). If the latter happens instead, then that would mean they mentioned NX/new hardware mostly in an attempt to quell the shareholder & public fears.

We will find out in a few months (around end of April, or whenever the FY2024 year-end and FY2025 forecast financial IR release comes out)

 

[Gerald]

I don't doubt that Nintendo's always looking at what other companies offer. What I'm trying to say is that I believe the simplicity of incorporating backwards compatibility is one of the main factors Nintendo's looking for from other companies. And I think there are huge challenges with incorporating backwards compatibility with AMD, especially since I don't believe AMD's as good as Nvidia when software's concerned. But anyway, I consider looking at what other companies offer to be very different from AMD actually bidding to be Nintendo's SoC provider for Nintendo's next hardware.

I didn't watch Moore's Law is Dead livestream yesterday. But I believe that's what Moore's Law is Dead said, going by what Gerald said.

First mention about NVN2 was from 2019 some people said, so there is no way AMD work on something for Nintendo when Nvidia aleardy start work

 

[Melpomene]

If the latter happens instead, then that would mean they mentioned NX/new hardware mostly in an attempt to quell the shareholder & public fears.

In this vein, I wouldn’t be surprised if they mentioned new hardware in this case for similar reasons (though obviously much less pressing ones, since we don’t have a failing console on our hands). It’s possible all the unofficial info swirling around might prompt them to concede a little in an official capacity to placate shareholders who are presumably quite interested in the validity of that information. But I could of course also see them not caring at all and carrying on like nothing’s happened.

 

[Gingerbread Man]

My two cents: Nintendo will talk about the new hardware only after the DK theme park has been opened and its hype cycle is over. So, late-May at best. That has always been the plan, even before Switch 2's delay.

 

[SoldierDelta]

In this vein, I wouldn’t be surprised if they mentioned new hardware in this case for similar reasons (though obviously much less pressing ones, since we don’t have a failing console on our hands). It’s possible all the unofficial info swirling around might prompt them to concede a little in an official capacity to placate shareholders who are presumably quite interested in the validity of that information. But I could of course also see them not caring at all and carrying on like nothing’s happened.

Here's some fun advice if you ever take up a political or business career that involves talking to journalists of any kind, and is something that a lot of people follow (note: this isn't trying to be condescending, i just think it's a bit funny as a piece of advice). If something crops up in the news that a select group of nerds, activists, reporters or a combination of the three pick up on, the absolute worst thing you can do is acknowledge it. At any point, acknowledging news is a bad choice because it means you'll be tied to it and have more articles written up about it.

Nintendo remaining quiet about the rumors in this case is a completely reasonable thing to do, as it makes the majority of casual shareholders and Nintendo fans ignore the signs and buy the next Mario remake without question. As long as people don't know or hear about it, they likely won't care and dismiss it as a rumor or as untrue. People tend to only hear things from the horse's mouth, so it only becomes a problem to them when Furukawa walks out on stage and says "Switch 2" and "delayed" in the same sentence.

 

[darthdiablo]

Here's some fun advice if you ever take up a political or business career that involves talking to journalists of any kind, and is something that a lot of people follow (note: this isn't trying to be condescending, i just think it's a bit funny as a piece of advice). If something crops up in the news that a select group of nerds, activists, reporters or a combination of the three pick up on, the absolute worst thing you can do is acknowledge it. At any point, acknowledging news is a bad choice because it means you'll be tied to it and have more articles written up about it.

Nintendo remaining quiet about the rumors in this case is a completely reasonable thing to do, as it makes the majority of casual shareholders and Nintendo fans ignore the signs and buy the next Mario remake without question. As long as people don't know or hear about it, they likely won't care and dismiss it as a rumor or as untrue. People tend to only hear things from the horse's mouth, so it only becomes a problem to them when Furukawa walks out on stage and says "Switch 2" and "delayed" in the same sentence.

While I agree in general sense, it’s IR which is different than PR.

It depends on what Nintendo feels they are compelled to report in their FY2025 forecast numbers that is shared to shareholders. It’s not about acknowledging the rumors.

 

[Hermii]

I didn't watch Moore's Law is Dead livestream yesterday. But I believe that's what Moore's Law is Dead said, going by what Gerald said.

I just watched it now, he never said AMD had a physical design.

Otherwise he doesn't make to much sense though.

 

[Anatole]

The strangest thing about existing in this thread for years is watching the same misinformation spread again and again as the same topics loop back around. There’s always a half-remembered definitive post by one of the really excellent regulars buried somewhere among that hundred thousand, but I don’t have the vigilance in me to search for it.

Someone will correct the info, or they won’t.

Someone else, who felt ignored their opinion was unfairly ignored the last time, will latch on to the opportunity to bring out their bugbear again, and they’ll get a lot of Yeahs for it.

One, three, six months from now, it will happen again. 150k posts, 200k posts. The same YouTubers. The same old leaked info. The same personalities. The same arguments.

The Switch 2 is always 2 years away.

I think I’m depressed.

 

[bmfrosty]

It's definitely going to be 8 GB of RAM. Look at Nvidia's smaller GPUs, all much stronger than this will be (2050 Mobile; 1650 Mobile Max-Q; 3050 Mobile), they have 4 GB, rarely 6 GB. The only saving grace is that NV GPUs can deal with lower GRAM much better than AMDs.

And production chips (especially on the lower cost end), never use the full amount of execution units they have by design. For yields and higher production numbers, they always leave one or two units unused in production design form. Full units are only used for high end models which really just means selling the few percent that come out perfect for a premium and to expand the market upwards. This isn't that. This is supposed to be a mass market product with millions of units produced every month. Look at the Orin chip, only the top model uses the full chip, 99% of units sold only use a part of the chip. Chips that don't come out perfect are sold as lower tier products instead of throwing them away. The GPU is the biggest single part and will be more susceptible to damage.

Leaks said that T239 has 12 SMs by Design? Production chips will use 8 or at best 10.

The chip was done in 2022? It's not going to use a high end process from 2024. It's going to use whatever its sister-chips of the same architecture are being made on, which started sales in the last year or two.

Nintendo is selling these things for 100 - 150, maybe up to 200 for the NS2. Everything above that is transport, taxes, retail costs and retail margin. If it's sold for 399, in many parts of Europe the taxes alone are 80, leaving just 320 to the retailer, take out a margin, then take out the cost of running the store, then take out storage and shipping across several locations across the world, what is left is the price Nintendo sells it for. Then take out Nintendos margin, their costs of developing the thing and all the design work from cooperating companies, and then you have production costs. This has to be cheap because it is sold cheap.

We have seen this movie 8 years ago, and 5 years before that, and 5 years before that... We were always left disappointed by crippled hardware. Be it non-standard storage mediums or obsolete chips - it always made it harder, not easier, for third party developments to be brought to the Nintendo platform. Now that I write this, that may not be an oversight.

There's a handy Nvidia Orin chart out there that shows the different configurations with TDP at the bottom. We roughly know (look at the OG Switch) the TDP limits the hardware form factor imposes on the chip alone which might raise a bit in the new gen but that's just hoping and probably won't happen (3-5 W handheld, 10 - 15 Watt docked). Don't dream, be honest with yourself, take a look, and be prepared to be disappointed. Because that is the only option. Better now than sitting around for another year hoping for a miracle that you can already know today won't come.

This thing is almost guaranteed to end up looking something like this:

8nm process
8GB RAM
8 SM GPU with 1024 cores clocked somewhere around 625 - 765 MHz docked, around 382 - 465 Mhz handheld
6 - 8 A78C cores running at 1.5 GHz
128 GB storage

Before you lash out because you don't like the numbers you are facing, remember this isn't my opinion, these are the limits which the TDP and production process are imposing - as per Nvidia themselves.

Everything outside of that (RAM + Storage) is just economics, in two ways:
Firstly the sales price. Don't show me your Android phone with 8 Cores, 8 GB and 256 GB for 199, I know that and I have one as well, but that's a different thing, the SoC/GPU is much smaller and thus cheaper and not on an advanced node like what everyone wants from the Switch, and the development of the device and Software are basically free compared to Switch OS, Joycons, each with their own battery, HD rumble, IR sensor, dock, etc.- you are not just buying the three hardware parts (CPU, RAM, Storage) and overpaying for them, the Switch is a whole device and ecosystem and Nintendo aims to turn a profit, and as much of it as possible. Chinese Android makers are basically giving them away at cost to gain market share and a foothold to introduce more premium devices afterwards. Nintendo has no prospects of coming out with Premium $1099 Switches and no venture funds covering the costs until then, who are betting on them taking out the competition and raking it in later when they have a monopoly, also referred to as "disruptive technology".
Nintendo also mainly uses traditional retail, driving down their own sales price to leave headroom for the retailer, cheap Androids are mostly sold online by webshops and the manufacturer sales price isn't as different as the end buyers price might suggest, i.e. the Switch is probably sold by Nintendo for not much more than those Androids, even if the final listed sales price is 50% higher. It has hardware like cheap Android devices because it is a cheap device, that's the aim. There is no market for a $599 Switch, let alone a $899 one. There also isn't a market for a high monthly subscription rate to offset a Switch sold at a loss.

Secondly, if 5 years ago you might have hoped for a higher powered Switch, the success in the last 4 years has guaranteed that you won't get one. Because it has proven two crucial things to Nintendo:

A) That there is a huge market that's open to Nintendo. Bigger than they themselves or anyone else thought they might still have, when everyone was convinced the handheld market was gone to phones and the TV market was the domain of PC-gaming replacement and yearly shooter and sports franchises.

B) That the weak hardware of the Switch didn't prevent them from selling almost 150 Million devices, gaining that market and making it to the top of the best selling gaming devices ever, guaranteed Top 3 of all time, neck to neck, and outdoing all of their previous efforts, most of them combined. (Obviously except for DS - not yet)

Out of these two lessons they have to draw two conclusions to follow up on the success of the Switch 1:

a) They have to keep access to that monster of a market open, so they can't have a "$599 Dollars!" machine.
Once designed it's impossible to scale back and a more expensive to manufacture device guarantees losses of both sales and on every sale.
To keep access to the market, the console has to be able to access the almost impulse buy territory of adults who aren't gamers, parents who buy it as a family activity, and also the second console territory of all three of these: families with more than one child, gamers with another gaming device, dedicated cheaper handheld device for the car (kids), commute, as a quick present or just as a low entry threshold to dip their toe in and see what it's all about without having to commit too much. Given these objectives, which are covered by pricing, it will need a second, simpler, smaller, lower priced device to complement the upper mid-range main Switch 2 (350-399).
Given that whatever hardware it launches with will set the baseline that a cheaper device can't go under to ensure all the games work, the base hardware, whatever they put into the Switch 2 at launch, has to be able to fit into the "lite" price bracket of a second cheaper device as well (199-249). Which just isn't possible with 16 or even 12 GB of RAM, nor necessary see Nvidia's own GPUs. Don't forget, they aren't just designing a main device, they are also designing a cheap device at the same time. The cuts come from other features, not the hardware platform.

In short they have to keep access to the whole newly confirmed market open, which means a low entry cost, at a profit, which means the lowest necessary components.

b) They don't need strong hardware to sell a lot of units! Read that again! They don't need to participate in a graphical arms race, they don't need to have all kinds of third party blockbuster games. This has just been proven, if they did need any of that they wouldn't have sold more than 20-50 million. The market has just confirmed to Nintendo that they don't need strong hardware. So they won't make it because it has been proven that they don't need to.

I am repeating this so often because no one here seems to really have understood that point.

They won't bother with what they don't need to do, and especially something which would compromise the number 1 objective outlined in the above point: To make money by selling the most possible amount of hardware at the highest possible profit. Why would they do anything that would cut into that profit and raise the price, and thus decrease the market and income, unless it's absolutely necessary, when they don't need to?!? They wouldn't, they are not insane.
And they are not going to risk losing out on literal tens of Billions to satisfy the small number of Nintendo enthusiasts who dream of a Nintendo console that can be everything and replace all their other devices for gaming and streaming by being able to play all third party titles well, as well as Nintendo games.
No, they will do only what is absolutely necessary - and hope to repeat their success.

We are still discussing the hardware and GPU and flops of this thing when they have been confirmed to Nintendo to be irrelevant. They will have sold 150 Million devices at an average of around 300,- with an abysmal 190-390 Gflops, which was too little even before it released.
Back then everyone hoped for at least 512 Cores at 1Ghz, we got only 256 Cores at 307-765Mhz, a Quad ARM at just 1 GHz like it's 2010, 4GB and 32GB MMC, a gut punch, almost a decade ago!
And it's still going strong! Still! It will still outsell the Wii U from today until it's pulled. The N64 and GC in it's final 2-3 years. You have to understand what these numbers mean, especially to Nintendo.

It means they don't need more. Of aything. They don't need to, anything is good enough. That's why you won't get anything from them, no compromise, no token of appreciation, in the form of more RAM or better node for higher fps or more resolution in 3rd party ports, which is basically what y'all are hoping for. But they don't need to. And they won't. Only whatever is the bare minimum necessary for the benchmark they have set for their own games, and that is all. They don't need to care about 3rd party ports, that's the 3rd party developer's problem if he wants to make money off of Nintendo's customers.

They don't need to try to compete with others any more, they are and can be their own market. They will make a dedicated Nintendo machine because that is what the people/market want.

A toy to play Nintendo video games on. It doesn't need or aim to provide High-End Computing Power Ray-Tracing Graphics Narrative Driven Engaging Hardcore Gaming Experiences.

They're not looking to be HBO or AMC, they are the Disney and Pixar of video games.

The hundred millions of gamers will not replace their PC/PS with a Nintendo device no matter how powerful they make it, because those will always have an edge over Nintendo in having every and all third party games and their own exclusives. So there is no sense in trying to replace those devices, you won't win them over. But some of them might buy a Switch in addition. Some people don't want to spend 600+ on serious dedicated gaming hardware and just want something affordable and easy, to unwind from time to time or to have fun as a family. Parents want something colorful, fun, safe, uncomplicated, trustworthy and affordable for their children.
That's Nintendo.

Ain't no one installing steam stores and shady emulators on weird Switch knockoffs while looking for illegal game copies, when he can get the original for half the price with zero hassle.

And no one's counting frames per second. I'd like 60, but the market doesn't seem to mind even 25 from time to time.

If third parties want to release their games on it to try and get a piece of that pie, they can, it's going to be capable enough for roughly a notch below Xbox One graphics in handheld and PS4 Slim when docked, seeing what they could put out on PS3/X360 and the last generation, there's really no excuse, all it takes is some effort, but Nintendo is not going to make a device for third party developers' games. They will make a dedicated Nintendo device just like the current Switch, because that has proven to be a home run.

They just need to have lots of good games of their own, and whatever affordable machine they deem necessary to make them look good. And that is all they are going to do.




That's why, again, this thing is almost guaranteed to end up looking something like this, at best:

8 nm process
8 GB RAM
8 SM GPU with 1024 cores clocked somewhere around 625 - 765 MHz docked, around 382 - 465 Mhz handheld
6 - 8 A78C cores running at most at 1.5 GHz
128 GB storage

That was such a a long post in reply to my little post that I declared it tl;dr and went on with my day.

 

[Hermii]

That was such a a long post in reply to my little post that I declared it tl;dr and went on with my day.

Good decision.

 

[Dakhil]

So MLID is back again and he is doubling down on his Switch 2 will be 8nm information in his live-stream...
He also reiterated that his sources told him that there was a 5nm version but Nintendo still went with the 8nm Samsung production.

He even goes on to claim (much similar to our specs here) that Nvidia are able to achieve better than Steamdeck performance in handheld (by 20-30% no less @ 500-800MHz GPU clock) and he's expecting 3-4TFlops docked 900Mhz-1.3Ghz GPU clock(on 8nm).

He talks about the CPU only being comparable to XboxOne X(Lol) and that this new Switch will be able to achieve these handheld performances for the low price of 5-8w... Please tell me what kind of Blackmagic has Nvidia and Samsung figured out since the Ampere cards?

Switch 2 stuff is closer to the beginning of his stream.

Moore's Law is Dead also said at 6:43 that there's a big misconception about how 1536 CUDA cores, which he claimed to have leaked, and that other people have leaked, and are on the Nvidia leaks years ago, on the GPU would've been retained, and the GPU frequency would've increased, if fabricated using TSMC's 5 nm* process node, which is wrong. He claims Nintendo would've reduced the number of CUDA cores on the GPU from 1536 CUDA cores to 1024 CUDA cores and increased the GPU frequency if fabricated using TSMC's 5 nm* process node.

He said on 4:58 that he doesn't agree with the data or performance curves from people online who think that achieving comparable or slightly better performance than the Steam Deck, with the GPU on Nintendo's new hardware being stronger than the Steam Deck's GPU, with Samsung's 8N process node, is not possible. And that people who disagree need to drive to Nvidia and say "Jensen, I'm smarter than you" since the people disagreeing apparently don't think achieving comparable or slightly better performance than the Steam Deck is possible (with Samsung's 8N process node), but Nvidia does.

---

I just watched it now, he never said AMD had a physical design.

Otherwise he doesn't make to much sense though.

At 7:33, Moore's Law is Dead said that the reason Nvidia won the contract is that AMD couldn't provide what Nintendo wanted in handheld mode at the price Nintendo's asking for. And at 7:59, he said he's been told that although the AMD version of Nintendo's new hardware outperforms the Nvidia version when in TV mode at 30 W, the AMD version couldn't deliver the level of performance Nintendo wanted in handheld mode unless a more advanced process node (he defines a more advanced process node in 7:52 as TSMC's 6 nm* process node or more advanced) is used, which is more expensive. I think that's where Gerald thought Moore's Law is Dead's saying a physical design from AMD exists. But yeah, TSMC's 6 nm* process node sounds more like his own speculation than a claim.

---

* → a marketing nomenclature used by all foundry companies

 

[Hermii]

Moore's Law is Dead also said at 6:43 that there's a big misconception about how 1536 CUDA cores, which he claimed to have leaked, and that other people have leaked, and are on the Nvidia leaks years ago, on the GPU would've been retained, and the GPU frequency would've increased, if fabricated using TSMC's 5 nm* process node, which is wrong. He claims Nintendo would've reduced the number of CUDA cores on the GPU from 1536 CUDA cores to 1024 CUDA cores and increased the GPU frequency if fabricated using TSMC's 5 nm* process node.

He said on 4:58 that he doesn't agree with the data or performance curves from people online who think that achieving comparable or slightly better performance than the Steam Deck, with the GPU on Nintendo's new hardware being stronger than the Steam Deck's GPU, with Samsung's 8N process node, is not possible. And that people who disagree need to drive to Nvidia and say "Jensen, I'm smarter than you" since the people disagreeing apparently don't think achieving comparable or slightly better performance than the Steam Deck is possible (with Samsung's 8N process node), but Nvidia does.

Yep, all this is ridiculous. That's what I mean by not making to much sense, Im just to lazy to make a summary.

 

[Marce-chan]

I suspect that EPD is targeting higher fidelity than they can currently handle

I wouldn't say it's beyond what they can handle, just that it's an increasing development time that we already saw between late Switch games(Splatoon 3, Tears of the Kingdom, Super Mario Bros Wonder) against early Switch titles (Splatoon 2, BotW, Odyssey).

I believe Switch itself was delayed from October 2016 to March 2017 because Nintendo was initially banking that the launch games would take about the same time as their Wii U predecessors to make, which didn't happen (Odyssey for example definitely wouldn't be able to release no matter what in late 2016, but well, Nintendo thought they could get BotW done in less than 3 years lol).

That's a matter of choice. If you want to do, say, 4K60 with little other changes, the fidelity of software targeting a T239 platform is... Nintendo Switch?

Plus there's the other way, too, Horizon Forbidden West has PS4 as its ground floor and was one of the most expensive games ever made, as was The Last of Us Part 2.

I believe with Horizon it's more of a game targeting PS5 with a PS4 version instead of the other way around (GoW Ragnarok). But yeah, late PS4 games were taking what was once enough to enter the top 10 most expensive games ever made as the norm.

I think it's more on Sony and other publishers that put like 2,000 people working on a game to get it done faster, something that Nintendo really never does. Which is also probably one of the reasons their games are taking longer and longer.

I really don't understand where does this talk or fear of Nintendo not being prepared for next-gen comes? They way Nintendo develop their games and the assets or effects in use in their games are in the same ballpark of quality and features as PS4 or XOne games. Just because Nintendo has a lower performance platform doesn't mean they're using or being used to outdated development practices. The jump from Switch development to Switch 2 development isn't a big and wide jump like going from SD development (Wii) to HD development (Wii U).

I suspect a lot of this bias and fear come from western development, whose pipelines takes way longer, are very expensive and have set the standard of high fidelity visuals.

However, if we look at a more comparable, to Nintendo, AAA studio from Japan, Capcom, we see that their games look just as good as most AAA western games while costing a fraction of said western games and coming out very quickly, without much development issues.

My point is: Just because Switch 2 will present a big tech leap, that doesn’t mean that Nintendo pipeline will suddenly stall. Nor does it mean Nintendo will need to re-learn their development practics because everything became outdated or that games will take a long time to be made and will be way more expensive.

The people at Nintendo are professionals and are at the forefront of technology R&D. Whatever comes with new tech leap won't take them by surprise and mess up their development environment.

I don't think that just because the next system will be a big leap most games will be, either. I'd say that anyone expecting more games than Mario and Mario Kart at first to really feel like a next gen thing that pushes the hardware is setting themselves up for disappointment. I'd expect mostly what we got on mid/late Switch titles but with higher resolution and frame rate.

It's not about them not being ready, but aiming for higher fidelity and bigger scope definitely makes development get longer. It isn't something exclusive to Wii -> Wii U, but Wii U -> Switch and even early Switch to late Switch games. Super Mario 3D World took about just two years to make, Odyssey took four, Tears of the Kingdom took something between 5-6 years while BotW took 4, etc.

Nintendo largely does not pump out same system, same engine sequels (outside of Pokémon) like Capcom does and that raises significant development risks. There’s already speculation that they delayed their entire console because 3D Mario (likely had reboots but was in some form of development since 2017) and maybe Prime 4 (had public reboots but was in development since at least 2017) weren’t ready to launch by holiday 2024. Dev times for Nintendo this gen will be long as they’re already long.

I don't think gigantic development times like what 3D Mario and Metroid Prime 4 are getting and Tears of the Kingdom got will be the norm. Mostly because I don't see most Nintendo franchises going for bigger scopes and higher fidelity than they're already getting on Switch.

I also don't see 3D Mario taking that long as any kind of catastrophe like many are taking. They probably took longer in pre-production and didn't start full development until early 2019 or something, in which an early 2025 release is just about what we'd expect from a prestige franchise that was already having bigger development cycles but this time they didn't have to rush anything since the next generation would take awhile back when it started development.

I don't think 3D Mario is the sole reason for the delay tho. But well BotW was the biggest reason for the Switch delay so it wouldn't be unheard of.

Isn’t MLIS infamously more often wrong than right?

I don't know about the track record but what he said just screams BS from everything we're hearing. He says the hardware and software are ready and have been for a while, which makes literally no sense. We got reports that it was recently delayed 3-6 months because they wanted to ensure a steady flow of software. That vastly contradicts his claims of it being ready for awhile.

I still don’t know why Nintendo is (possibly) going to announce Switch 2 in June and then release it 9 months later. Wouldn’t it be better to do it in October since it would be 6 months from there?

I often think of the original Switch being a brand new concept and having just 4 months from full reveal to release as a reason to take the reports about reveals 6-9 months before release with a grain of salt.

But I think the fact that they mentioned that they were producing new hardware back in early 2015, then dating it for March and announcing Zelda to come as well in early 2016, having that full Zelda blowout and everything made it already have a hype behind when it was announced.

They had Zelda fully revealed 9 months before release and that was the launch title. The Super Mario game was formally announced only in January, 9 months before its release that holiday season. Maybe they don't want to advertise the next Mario which is presumably the launch title for only 4 months?

 

[Z0m3le]

What I find really amazing about his speculation is that he still thinks the GPU is between 3 to 4 TFLOPs, but that the CPU is on par with the XB1's CPU... Which Switch's CPU at 1GHz is already ~2/3rds of the performance, while A78 cores are 3 times the performance per clock than the cores inside Switch... It's amazing how his logic block here looks like a sponge.

 

[Hermii]

What I find really amazing about his speculation is that he still thinks the GPU is between 3 to 4 TFLOPs, but that the CPU is on par with the XB1's CPU... Which Switch's CPU at 1GHz is already ~2/3rds of the performance, while A78 cores are 3 times the performance per clock than the cores inside Switch... It's amazing how his logic block here looks like a sponge.

What I found the most amazing is that he said a 5nm soc would have fewer coda cores than a 8nm soc. Like doesn't he know how densities/ power curves work?

 

[Gerald]

What I found the most amazing is that he said a 5nm soc would have fewer coda cores than a 8nm soc. Like doesn't he know how densities/ power curves work?

he said many things that Xbox have contract with Intel, when we know from Kepler from last year that Xbox work with AMD on APU Zen5/RDNA5 and Next Xbox will release in Late 2026

 

[Z0m3le]

What I found the most amazing is that he said a 5nm soc would have fewer coda cores than a 8nm soc. Like doesn't he know how densities/ power curves work?

No he doesn't, that is the problem... It's also probably why tech people feel free to tell him stuff, they know it's cryptic enough just telling him the info directly.

 

[omniryu]

The Switch 2 is always 2 years away.

I think I’m depressed.

Buddy, there is nothing to be depressed about. You, who happen to be a Nintendo fan, got other things to do then be worried. I know you haven't played every published Nintendo game. Try some out, even then that's only Nintendo. You have some much stuff to consume.

Moore's Law is Dead said...

That's what I want to know. Where do they get this stuff? Here's what bothers me. Let's give MLID the benefit of the doubt. He knew what Nvidia said. Yet, he knows the intention of what Nintendo's plan for the Switch this year.
Like who is his connection? There's no way you connections no Nintendo's plan for the year and is a deep Nvidia insider or he GPU insider.

 

[Darknut85]

Buddy, there is nothing to be depressed about. You, who happen to be a Nintendo fan, got other things to do then be worried. I know you haven't played every published Nintendo game. Try some out, even then that's only Nintendo. You have some much stuff to consume.

Exactly. Nintendo might be be doomed, but we‘ve at least plenty games to play (unlike fans of some other consoles)./s

 

[Maxrunner]

I was told this by an insider, I don't have direct info from anyone in NDA.

So its probably 12gb then.

 

[ILikeFeet]

That's what I want to know. Where do they get this stuff? Here's what bothers me. Let's give MLID the benefit of the doubt. He knew what Nvidia said. Yet, he knows the intention of what Nintendo's plan for the Switch this year.
Like who is his connection? There's no way you connections no Nintendo's plan for the year and is a deep Nvidia insider or he GPU insider.

there is no connection. he makes shit up to fill his hour-long podcast

 

[oldpuck]

@miamoto - are you okay? You seem extremely angry? I don't know your post history - have you received any sort of lash out from your opinions in the past? Like, I think some of your analysis doesn't hold up, partially because it seems based on misunderstanding some facts, but I don't think your position is unreasonable. I don't think it's even extreme for the thread, excepting some extreme optimists.

I'm gonna touch on some things where I think you are factually inaccurate, and how I think that puts your analysis on the wrong path, but I don't want to get into a pileup here. You seem upset, and I don't want to add to that.

The only saving grace is that NV GPUs can deal with lower GRAM much better than AMDs.

This isn't true. Memory usage is almost entirely dictated by geometry and texture size, not architecture. Nvidia doesn't have a substantial advantage here.

The RAM differences between AMD and Nvidia cards is a side effect of how the two companies have chosen to bin their dies. AMD's biggest 6000 series card only has 16 GB of memory in it, Nvidia's has much more, 24 GB.

However, Nvidia continues to use a floorswept version of that largest die for many more cards than AMD does. As they drop GPCs from the die, they also lose a certain number of memory controllers. GDDR is only produced in certain quantities, without custom orders. With only so many slots and so many DIMMs available, Nvidia is placed in a position where they they either have to way over provision RAM or way under provision RAM, especially on their midrange cards

This is why the 4060 Ti was such a disappointment. The 4060 started out with 8 GB of RAM, but that's not enough for the level of power on the card offers. The Ti offered 16GB of RAM at 100 dollars more - which is too goddamn much. That's because the hardware for a 12 GB card simply doesn't exist.

And production chips (especially on the lower cost end), never use the full amount of execution units they have by design.

This is not true either. Mid-range cards are often floorswept versions of the high end cards, yes, but low-end cards are not. The AD-107 Die, the basis for the bottom of the RTX 40 range is sold as a full die, it's not intentionally over designed and then cut down. Initially it was the same thing for GA-107, the RTX 3050 die. Years later, Nvidia would start selling GA-107 dies with cores disabled, but it seems unlikely that yields suddenly dropped. It seems much more likely that Nvidia needed an even cheaper lower end card, and needed to manually disable some working cores to create product differentiation.

This is actually quite important to understand. It makes sense to overdesign big chips, but not to overdesign small chips.

Chunks of a wafer have errors. Those errors will result in dies that don't work fully. The bigger a chip, the more likely it will have errors. Some parts of a chip can be duplicated, mainly cores. Some parts cannot, like L2 cache, uncore, IO interconnects, etc.

If you have a larger chip (like an Xbox Series X APU, for example), you can make the GPU 30% bigger but only make the chip, say, 15% bigger. That means you get a 15% decrease in the number of chips per wafer, but you have a 30% buffer on your GPU where you can work around errors. That results in more chips per wafer.

But that breaks down for smaller chips. As a chip gets smaller, it becomes less and less likely that you can survive an error, because a same-size error on a same-size wafer hits a larger percentage of the chip. A 30% increase in the GPU size of a small chip means you only get like a 10% increase in the chip size (because uncore is a much larger percentage of the chip), but something like a 0-5% increase tin the number of errors you can survive (because you can never survive errors in the uncore, and the chip is so small that you can only survive tiny errors in the cores). This is just fewer chips per wafer.

For yields and higher production numbers, they always leave one or two units unused in production design form. Full units are only used for high end models which really just means selling the few percent that come out perfect for a premium and to expand the market upwards. This isn't that. This is supposed to be a mass market product with millions of units produced every month. Look at the Orin chip, only the top model uses the full chip, 99% of units sold only use a part of the chip. Chips that don't come out perfect are sold as lower tier products instead of throwing them away. The GPU is the biggest single part and will be more susceptible to damage.

...but this is a highly custom chip with no market for binned/floorswept units. Being able to use binning to create multiple product lines doesn't actually work here. The most cost effective strategy is not to maximize the number of chips that work, period, and then sell off chips at different price/performance tiers (like Orin, or a GPU) but to maximize the number of chips that hit the highest end performance target.

Leaks said that T239 has 12 SMs by Design? Production chips will use 8 or at best 10.

No, that is not what the leak said. There are no chip design leaks that I am aware of. There are Nintendo software leaks and that software says that production chips have 12SMs.

The chip was done in 2022? It's not going to use a high end process from 2024. It's going to use whatever its sister-chips of the same architecture are being made on, which started sales in the last year or two.

Yeah, I buy this.

There's a handy Nvidia Orin chart out there that shows the different configurations with TDP at the bottom. We roughly know (look at the OG Switch) the TDP limits the hardware form factor imposes on the chip alone which might raise a bit in the new gen but that's just hoping and probably won't happen (3-5 W handheld, 10 - 15 Watt docked). Don't dream, be honest with yourself, take a look, and be prepared to be disappointed. Because that is the only option. Better now than sitting around for another year hoping for a miracle that you can already know today won't come.

I think you are roughly correct in power draw. However, to get down to 3-5 W on Orin, you actually need to spec it down below the current Switch. So I suspect the Orin power profiles don't actually apply.

This thing is almost guaranteed to end up looking something like this:

8nm process

Yeah, I don't think you're in an extreme position there

8GB RAM

This is an extreme position. Remember what I said about memory controllers dictating the RAM options available? Well, obviously, T239 is going to use a different memory controller, LPDDR based instead of GDDR. Which changes the economics of what's available.

No one manufactures 4GB 64-bit LPDDR5 modules. They're useless in the market. Unless Nintendo and Nvidia want to pay extra money for custom RAM, this isn't an option. And as we saw with the 4060Ti above, it's probably not cost effective to pay extra money for custom RAM.

From a technological point of view, Nintendo will need some RAM for, you know, games. It's not just GPU RAM. Nvidia cards are notorious for being underspecced in RAM, and Nintendo needs more RAM than just RAM for the GPU. 8GB seems extremely unlikely.

8 SM GPU with 1024 cores

I would agree with you, if the NVN2 leak hadn't happened. I understand that many people misunderstand the leak to think that it shows the number of SMs on die, which they can then assume might be floorswept off. But that's not what the leak says. The leaks says that Nintendo's operating system and games expect 12 SMs to be there, in production.

clocked somewhere around 625 - 765 MHz docked, around 382 - 465 Mhz handheld

I'm not disputing your handheld clocks, but your docked clocks seem like you're just trying to create controversy? Like what is your rationale here?

You're driving a much larger screen with 4x the pixels, but you're looking at as little as a 30% increase in GPU performance? Why? You're not power constrained, you're only constrained by the fan's ability to dissipate heat. The Switch was originally clocked at over double in docked mode, and the docked clock was on the lower end of what other Maxwell GPUs had (like the 800M mobile chip).

You think that the RTX design is so power inefficient at clocks this low that you can't double the clock speeds even at 15W up from 3W?

6 - 8 A78C cores running at 1.5 GHz

I mean, again, this isn't even the leak. Nvidia's Linux driver says it will be 8 cores. But yeah, that's about where I expect the clocks to be. Not crazy.

128 GB storage

Yeah, that tracks.

Before you lash out because you don't like the numbers you are facing, remember this isn't my opinion, these are the limits which the TDP and production process are imposing - as per Nvidia themselves.

Not lashing out, it just seems like you're making up some limits that don't have anything to do with TDP and production processes, like RAM and Storage.

I don't think 8nm is an unreasonable node, and I think your analysis on the power limits is spot-fucking on. But that's actually the reason that some folks tend to not think 8nm is the node - because your power analysis is exactly right, and it does come from Nvidia but the chip size (SMs, CPU cores) also comes from Nvidia and are production numbers. You seem to have not known that the numbers weren't from a spec sheet or a chip design, but from software, and assumed that Nvidia would need to floorsweep the chip to hit the power targets.

That makes sense, except that it takes Orin, which is a different design, as a Absolute Accurate Reference for the power consumption of T239, while thinking that leaks about T239 itself are Squishy Malleable References for T239's chip size. There are three majory possibilities.

The power numbers apply to T239, and the chip size is right, so it's not 8nm. This is maybe the loudest contingent of the thread. If you put a gun to my head, this is where I fall, but please don't put a gun to my head.

The power numbers apply to T239, and it's on 8nm, so the chip size is wrong. This is your position, and a pretty common position in other parts of the internet. As I have tried to show here, that doesn't actually jibe with what leaked, just with what people assume leaked, and it doesn't really jibe with how Nvidia handles their small dies. That's why this is not my position, and I find it the least likely.

It's on 8nm for all the reasons you describe, the chip size is right for all the reasons I describe, then Orin's power limits must not apply. There is substantial evidence for this as well. In both the Nvidia hack and in the Linux drivers there is evidence of power saving tech that Orin doesn't have. And AMD was able to increase Steam Deck's power efficiency just by tuning power curves for the chip. This has been my position historically, and something I still think has lots of merit. It's just that the amount of savings required seem pretty high.

By way of comparison, I'll put down my current predictions as well.

12 SM GPU
Just straight out of the leak. If this is floorswept, then I'll count this as a serious failure of analysis on my part. I'll take the L.
500Mhz handheld, 1GHz docked.
Not much higher than your numbers, and not much higher than the current Switch. Thraktor has made a really good argument about likely clock speeds, but I still tend toward some slight pessimism here.
12 GB of RAM.
Rumors have pushed this up to 16 GB, but I'm still holding to the older 12 GB prediction. We've gone round and round 1000 times about the availability of 64bit 4GB modules, they just aren't out there. Nintendo is not historically conservative about RAM - but RAM isn't getting cheap as fast as it once did. 12 GB is still really nice compared to last gen, I think it's a good bang-for-the-buck cost savings.
8 A78C cores
Straight out of Nvidia's Linux drivers. I think it's funny that no one suggests floor sweeping CPU cores, even though the logic would be the same.
1.5GHz
I think this is one of the more flexible specs here, and I wouldn't be surprised by something wildly different. CPU power is less at a premium, but also, the ARM cores are super efficient at these clocks. So it really depends on how much room there is in the power budget. I wouldn't be completely surprised by 2.0GHz, or as low as 1.2GHz.

You'll note I don't list the node. That's because I don't much give a shit, and because, as I've said, I'm really flaky on that part. I'm more confident about the performance target based on the hardware leaks than I am in how they achieve it. I can give more analysis if you like, but I feel like I've gone on too long already. I just want it to boil down to - I actually agree with most of what you're saying, but a small change in data (the nature of the leak, the odds of binning/floorsweeping) takes me in a substantially different direction.

 

[Tof]

@miamoto - are you okay? You seem extremely angry? I don't know your post history - have you received any sort of lash out from your opinions in the past? Like, I think some of your analysis doesn't hold up, partially because it seems based on misunderstanding some facts, but I don't think your position is unreasonable. I don't think it's even extreme for the thread, excepting some extreme optimists.

I'm gonna touch on some things where I think you are factually inaccurate, and how I think that puts your analysis on the wrong path, but I don't want to get into a pileup here. You seem upset, and I don't want to add to that.



This isn't true. Memory usage is almost entirely dictated by geometry and texture size, not architecture. Nvidia doesn't have a substantial advantage here.

The RAM differences between AMD and Nvidia cards is a side effect of how the two companies have chosen to bin their dies. AMD's biggest 6000 series card only has 16 GB of memory in it, Nvidia's has much more, 24 GB.

However, Nvidia continues to use a floorswept version of that largest die for many more cards than AMD does. As they drop GPCs from the die, they also lose a certain number of memory controllers. GDDR is only produced in certain quantities, without custom orders. With only so many slots and so many DIMMs available, Nvidia is placed in a position where they they either have to way over provision RAM or way under provision RAM, especially on their midrange cards

This is why the 4060 Ti was such a disappointment. The 4060 started out with 8 GB of RAM, but that's not enough for the level of power on the card offers. The Ti offered 16GB of RAM at 100 dollars more - which is too goddamn much. That's because the hardware for a 12 GB card simply doesn't exist.


This is not true either. Mid-range cards are often floorswept versions of the high end cards, yes, but low-end cards are not. The AD-107 Die, the basis for the bottom of the RTX 40 range is sold as a full die, it's not intentionally over designed and then cut down. Initially it was the same thing for GA-107, the RTX 3050 die. Years later, Nvidia would start selling GA-107 dies with cores disabled, but it seems unlikely that yields suddenly dropped. It seems much more likely that Nvidia needed an even cheaper lower end card, and needed to manually disable some working cores to create product differentiation.

This is actually quite important to understand. It makes sense to overdesign big chips, but not to overdesign small chips.

Chunks of a wafer have errors. Those errors will result in dies that don't work fully. The bigger a chip, the more likely it will have errors. Some parts of a chip can be duplicated, mainly cores. Some parts cannot, like L2 cache, uncore, IO interconnects, etc.

If you have a larger chip (like an Xbox Series X APU, for example), you can make the GPU 30% bigger but only make the chip, say, 15% bigger. That means you get a 15% decrease in the number of chips per wafer, but you have a 30% buffer on your GPU where you can work around errors. That results in more chips per wafer.

But that breaks down for smaller chips. As a chip gets smaller, it becomes less and less likely that you can survive an error, because a same-size error on a same-size wafer hits a larger percentage of the chip. A 30% increase in the GPU size of a small chip means you only get like a 10% increase in the chip size (because uncore is a much larger percentage of the chip), but something like a 0-5% increase tin the number of errors you can survive (because you can never survive errors in the uncore, and the chip is so small that you can only survive tiny errors in the cores). This is just fewer chips per wafer.



...but this is a highly custom chip with no market for binned/floorswept units. Being able to use binning to create multiple product lines doesn't actually work here. The most cost effective strategy is not to maximize the number of chips that work, period, and then sell off chips at different price/performance tiers (like Orin, or a GPU) but to maximize the number of chips that hit the highest end performance target.


No, that is not what the leak said. There are no chip design leaks that I am aware of. There are Nintendo software leaks and that software says that production chips have 12SMs.


Yeah, I buy this.


I think you are roughly correct in power draw. However, to get down to 3-5 W on Orin, you actually need to spec it down below the current Switch. So I suspect the Orin power profiles don't actually apply.

Yeah, I don't think you're in an extreme position there

This is an extreme position. Remember what I said about memory controllers dictating the RAM options available? Well, obviously, T239 is going to use a different memory controller, LPDDR based instead of GDDR. Which changes the economics of what's available.

No one manufactures 4GB 64-bit LPDDR5 modules. They're useless in the market. Unless Nintendo and Nvidia want to pay extra money for custom RAM, this isn't an option. And as we saw with the 4060Ti above, it's probably not cost effective to pay extra money for custom RAM.

From a technological point of view, Nintendo will need some RAM for, you know, games. It's not just GPU RAM. Nvidia cards are notorious for being underspecced in RAM, and Nintendo needs more RAM than just RAM for the GPU. 8GB seems extremely unlikely.


I would agree with you, if the NVN2 leak hadn't happened. I understand that many people misunderstand the leak to think that it shows the number of SMs on die, which they can then assume might be floorswept off. But that's not what the leak says. The leaks says that Nintendo's operating system and games expect 12 SMs to be there, in production.


I'm not disputing your handheld clocks, but your docked clocks seem like you're just trying to create controversy? Like what is your rationale here?

You're driving a much larger screen with 4x the pixels, but you're looking at as little as a 30% increase in GPU performance? Why? You're not power constrained, you're only constrained by the fan's ability to dissipate heat. The Switch was originally clocked at over double in docked mode, and the docked clock was on the lower end of what other Maxwell GPUs had (like the 800M mobile chip).

You think that the RTX design is so power inefficient at clocks this low that you can't double the clock speeds even at 15W up from 3W?


I mean, again, this isn't even the leak. Nvidia's Linux driver says it will be 8 cores. But yeah, that's about where I expect the clocks to be. Not crazy.


Yeah, that tracks.


Not lashing out, it just seems like you're making up some limits that don't have anything to do with TDP and production processes, like RAM and Storage.

I don't think 8nm is an unreasonable node, and I think your analysis on the power limits is spot-fucking on. But that's actually the reason that some folks tend to not think 8nm is the node - because your power analysis is exactly right, and it does come from Nvidia but the chip size (SMs, CPU cores) also comes from Nvidia and are production numbers. You seem to have not known that the numbers weren't from a spec sheet or a chip design, but from software, and assumed that Nvidia would need to floorsweep the chip to hit the power targets.

That makes sense, except that it takes Orin, which is a different design, as a Absolute Accurate Reference for the power consumption of T239, while thinking that leaks about T239 itself are Squishy Malleable References for T239's chip size. There are three majory possibilities.

The power numbers apply to T239, and the chip size is right, so it's not 8nm. This is maybe the loudest contingent of the thread. If you put a gun to my head, this is where I fall, but please don't put a gun to my head.

The power numbers apply to T239, and it's on 8nm, so the chip size is wrong. This is your position, and a pretty common position in other parts of the internet. As I have tried to show here, that doesn't actually jibe with what leaked, just with what people assume leaked, and it doesn't really jibe with how Nvidia handles their small dies. That's why this is not my position, and I find it the least likely.

It's on 8nm for all the reasons you describe, the chip size is right for all the reasons I describe, then Orin's power limits must not apply. There is substantial evidence for this as well. In both the Nvidia hack and in the Linux drivers there is evidence of power saving tech that Orin doesn't have. And AMD was able to increase Steam Deck's power efficiency just by tuning power curves for the chip. This has been my position historically, and something I still think has lots of merit. It's just that the amount of savings required seem pretty high.

By way of comparison, I'll put down my current predictions as well.

12 SM GPU
Just straight out of the leak. If this is floorswept, then I'll count this as a serious failure of analysis on my part. I'll take the L.
500Mhz handheld, 1GHz docked.
Not much higher than your numbers, and not much higher than the current Switch. Thraktor has made a really good argument about likely clock speeds, but I still tend toward some slight pessimism here.
12 GB of RAM.
Rumors have pushed this up to 16 GB, but I'm still holding to the older 12 GB prediction. We've gone round and round 1000 times about the availability of 64bit 4GB modules, they just aren't out there. Nintendo is not historically conservative about RAM - but RAM isn't getting cheap as fast as it once did. 12 GB is still really nice compared to last gen, I think it's a good bang-for-the-buck cost savings.
8 A78C cores
Straight out of Nvidia's Linux drivers. I think it's funny that no one suggests floor sweeping CPU cores, even though the logic would be the same.
1.5GHz
I think this is one of the more flexible specs here, and I wouldn't be surprised by something wildly different. CPU power is less at a premium, but also, the ARM cores are super efficient at these clocks. So it really depends on how much room there is in the power budget. I wouldn't be completely surprised by 2.0GHz, or as low as 1.2GHz.

You'll note I don't list the node. That's because I don't much give a shit, and because, as I've said, I'm really flaky on that part. I'm more confident about the performance target based on the hardware leaks than I am in how they achieve it. I can give more analysis if you like, but I feel like I've gone on too long already. I just want it to boil down to - I actually agree with most of what you're saying, but a small change in data (the nature of the leak, the odds of binning/floorsweeping) takes me in a substantially different direction.

Thank you Old Puck for another in depth educational response. When I first read miamoto’s post I could acknowledge the time put into it but I could also see that a lot of incorrect assumptions were made and many people (not everyone as I saw just as many genuine responses) here jumped on that as another “because Nintendo” post. But your response here and others who also replied in good faith will not only help educate regulars but also newcomers and lurkers who might have a similar stance to miamoto which will lead to more productive discussion. I think many of us always assume that everyone should be up to date with the information that is out there on the NS2 but even with page 1 of this thread, sometimes it really is easier to just respond to people’s input in this thread as it specifically targets common misconceptions rather than assuming everyone has 100% understanding of the information already presented

 

[PedroNavajas]

I'm sorry about talking about Moore's Law is Dead, but I find his claims with respect to AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware, very hard to believe.
One, I believe ~99.9% backwards compatibility with the Nintendo Switch is impossible without spending a ridiculous amount of money for a 6 nm* Arm based SoC from AMD.
Two, why talk now about AMD bidding to be the SoC provider for Nintendo's new hardware, and the existence of a 6 nm* Arm based SoC from AMD for Nintendo's new hardware? Why not much earlier, especially since there are plenty of AMD insiders (e.g. Kepler, etc.) more than willing to talk? I believe the simple answer is that those claims were never really reliable to begin with.
* → a marketing nomenclature used by all foundry companies

While I seriously doubt there was any serious bid from AMD to Nintendo, I do not have hard believing that there might be someone in AMD pitching chip designs to several potential consumers, including Nintendo. Also, I do not think is a stretch that there's someone in Nintendo in communication with several different chip vendors in order to explore their options, even if is just to have some cards to play when negotiation prices with Nvida. I mean, it would be negligent for Nintendo is they are not exploring what other chips design are there that could fit their purposes.

That AMD pitched an ARM design part is hard to believe. Also, Is doubtful that MLiD is privy to details of such communications.

 

[Hermii]

wall of text

miamoto said:
The chip was done in 2022? It's not going to use a high end process from 2024. It's going to use whatever its sister-chips of the same architecture are being made on, which started sales in the last year or two.


You buy this? Those Lovelace chips, also taped out at the same time as Drake. And its a handheld where performance per watt is a lot more important, than even laptop chips.

I mean, I would buy the basic logic about sister chips (not timing, that seems more suggestive of 5nm), if we didn't have info about 8 a78, 12 SM but we do.

 

[Hermii]

While I seriously doubt there was any serious bid from AMD to Nintendo, I do not have hard believing that there might be someone in AMD pitching chip designs to several potential consumers, including Nintendo. Also, I do not think is a stretch that there's someone in Nintendo in communication with several different chip vendors in order to explore their options, even if is just to have some cards to play when negotiation prices with Nvida. I mean, it would be negligent for Nintendo is they are not exploring what other chips design are there that could fir their purposes.

That AMD pitched an ARM design part is hard to believe. Also, Is doubtful that MLiD is privy to details of such communications.

Not really, anyone can license arm. That's how ARMs business mode works. Also, since this is a handheld and a successsor to an arm console, it would make more sense to stick with arm, no matter who the vendor is.

 

[Dakhil]

You buy this? Those Lovelace chips, also taped out at the same time as Drake. And its a handheld where performance per watt is a lot more important, than even laptop chips.

I mean, I would buy the basic logic about sister chips (not timing, that seems more suggestive of 5nm), if we didn't have info about 8 a78, 12 SM but we do.

I assume oldpuck thought that miamoto's talking about Nvidia's flavour of TSMC's N3E process node, especially if Hopper Next's still planned for launch in 2H 2024, and TSMC's N3E process node should start high volume manufacturing (HVM) this year, which is correct, not about how implausible the possibility of Nintendo and Nvidia using TSMC's 4N process node to fabricate Drake is, which I've already debunked here.