Support NeoGAF

[th3sicknness]

X1 would be used in the dev kit even if it's 28nm.

So they are going to fab a chip at 28nm that they were already making at 20nm?? How does that make more sense? Why on Earth would Nvidia let them move backwards with their name on this device?

 

[MuchoMalo]

Yeah but you PC players are a weird bunch Seriously though of course a person invested in a gaming PC is not going to bother with any console really, unless there's exclusives. However there are 10's of millions of console gamers that could be persuaded. That's who Nintendo's potential market is.

And someone with a PS4 or Xbone isn't gonna buy a new Nintendo console for third-party ports no matter how beefy it is. So, really, if Switch's success is based on ports as so many people believe, it was DOA before they even started working on it.

So they are going to fab a chip at 28nm that they were already making at 20nm?? How does that make more sense? Why on Earth would Nvidia let them move backwards with their name on this device?

Because it's cheaper. Both the A57 and Maxwell have already been used on 28nm so it costs very little in terms of development to create the chip, while actual production would be significantly cheaper.

 

[ChrisMahoney]

28nm is still a very commonly used process and and very cheap due to being tapped-out. It's also not going away anytime soon. It's what makes the most sense given what we know about the Nintendo of today. Before Wii U I'd say that they wouldn't use an old process as even Wii used the same process as PS3, but the Nintendo of today only cares about keeping costs down. They don't give a shit about efficiency. 28nm is definitely what we're getting. 20nm is a dead-end process, and 16nmFF wouldn't need a fan running in portable mode even it if has 3 SMs. There's just no way that it's anything other than 28nm.



The K1 is based on an older architecture and is slower than Switch's docked mode. X1 would be used in the dev kit even if it's 28nm.

We don't know if there's a fan running in the retail version of portable mode.

 

[MuchoMalo]

We don't know if there's a fan running in the retail version of portable mode.

The recent patent indicates that it is IIRC.

 

[th3sicknness]

Because it's cheaper. Both the A57 and Maxwell have already been used on 28nm so it costs very little in terms of development to create the chip, while actual production would be significantly cheaper.

Arent both iterations of Maxwell based Tegra discontinued (Pixel C and Shield Tablet)? Also if Nvidia is already producing Pascal based Tegra for automotive wouldnt it be just as likely to use 16nm because they are already at scale?

They are going to go backwards to get a cheaper price, just doesnt seem likely?

 

[Gotdatmoney]

So 28nm looks pretty realistic now. Ihat would point towards 2 SM due to size wouldn't it? Though I wonder. How much heat are we looking at on a 28nm process? If it's comparable to what the X1 produces at 20nm do those clocks really still need a fan running at all times?

 

[Schopenhauerian]

Not every Nintendo thread has to become a history lesson does it?

The consistent references, in every Nintendo thread, to Nintendo's negative traits as a company (lack of intelligence, lack of humility, lack of ethics, and so on) are what compel me to investigate the history a bit more (this negativity towards Nintendo within the enthusiast community is a problem for them).

Compared to other threads, however, folks in this thread seem to me much more likely to actually offer argument and evidence in support of their claims, which is why I thought it might be worthwhile to ask some questions (here and here) that seem to me pertinent to the evaluation of those claims (the same claims about Nintendo that are consistently made in every Nintendo thread).

 

[Hermii]

So they are going to fab a chip at 28nm that they were already making at 20nm?? How does that make more sense? Why on Earth would Nvidia let them move backwards with their name on this device?

It's a Nintendo device not a nvidia device. Nintendo would always have the final say.

 

[MuchoMalo]

Arent both iterations of Maxwell based Tegra discontinued (Pixel C and Shield Tablet)? Also if Nvidia is already producing Pascal based Tegra for automotive wouldnt it be just as likely to use 16nm because they are already at scale?

They are going to go backwards to get a cheaper price, just doesnt seem likely?

They aren't using a straight X1, so the X1 being discontinued is completely irrelevant. And we know that it's Maxwell already from multiple sources. And does it really seem unlikely that Nintendo of all companies wouldn't go backwards to cut costs? Even after 3DS and Wii U?

 

[lwilliams3]

If Nintendo is going for a bigger process, there is probably likely another reason besides short-term cost.

 

[ggx2ac]

So 28nm looks pretty realistic now. Ihat would point towards 2 SM due to size wouldn't it? Though I wonder. How much heat are we looking at on a 28nm process? If it's comparable to what the X1 produces at 20nm do those clocks really still need a fan running at all times?

Yes when docked, uncertain about when portable.

Tegra Shield TV runs at 20W and uses active cooling.

The Switch when docked is likely to be in a similar Wattage range so the cooling is to make sure it doesn't overheat.

When the Switch is portable according to the patent, it runs at a lower RPM compared to when it is docked. This is probably for the same reasons to avoid overheating otherwise the CPU and or GPU would throttle.

If Nintendo is going for a bigger process, there is probably likely another reason besides short-term cost.

Yields, which would be tied to cost. It would get expensive having to bin wafers they can't use.

Edit: Maybe this would tie with what NateDrake said about them looking at Pascal but instead went with Maxwell, but that's speculation.

 

[Rodin]

It often pays to be a pessimist with these things, but damnit Nintendo!

Yeah pretty much hahaha

and we will see how many SM it has on the 12th.

We definitely won't. Nintendo will never release a spec sheet with the number of cores.

We don't know if there's a fan running in the retail version of portable mode.

We have a patent showing that the fan spins at lower RPM when undocked instead of shutting down completely.

 

[Gotdatmoney]

Yes when docked, uncertain about when portable.

Tegra Shield TV runs at 20W and uses active cooling.

The Switch when docked is likely to be in a similar Wattage range so the cooling is to make sure it doesn't overheat.

When the Switch is portable according to the patent, it runs at a lower RPM compared to when it is docked. This is probably for the same reasons to avoid overheating otherwise the CPU and or GPU would throttle.

Shoulda specified I meant portable mode. In the patent as you said it says that the system runs the fan at low RPMs. I'm just wondering if the move to a bigger process would make it necessary from a heat generation stand point considering how low the clocks are.

I'm just trying to figure out if at 28nm 3SM would be outta the question due to heat generation and chip size. I imagine hell yess is the answer.

 

[ggx2ac]

Shoulda specified I meant portable mode. In the patent as you said it says that the system runs the fan at low RPMs. I'm just wondering if the move to a bigger process would make it necessary from a heat generation stand point considering how low the clocks are.

I'm just trying to figure out if at 28nm 3SM would be outta the question due to heat generation and chip size. I imagine hell yess is the answer.

Yes and yes, those are the constraints for a 28nm node. It's hard to make the GPU larger while keeping the die size reasonable and it would also increase the thermal output and power consumption even if you were able to do it.

 

[Donnie]

Thraktor actually made a pretty good case for it. It's the only thing that jives with the low clocks speeds and still requiring a fan.

He made the case that the new 28nm processes are very close to 20nm in power usage/efficiency. That's why I asked him why he thought that 28nm was possible (as needing a fan doesn't seem to point to 28HPC+ anymore than it does to 20nm given their similarities).

He did make a more important point regarding the vendors still producing 20nm chips though. But for me die size would be a concern at that process and would seriously limit Nintendo from making the kind of customisations they love to make.

 

[Schnozberry]

So they are going to fab a chip at 28nm that they were already making at 20nm?? How does that make more sense? Why on Earth would Nvidia let them move backwards with their name on this device?

Nvidia didn't sell them the Tegra X1, and they don't make the Tegra X1 anymore. They licensed them their IP and assisted them in engineering a custom design. Nvidia is interested in getting a design win and gaining revenue. They don't care whether Nintendo goes bleeding edge or not.

 

[Donnie]

I think the Wii U being so customized makes it more likely that they would use a more "off the shelf" part this time. They put in a lot of work with Wii U and it didn't pay off at all.

Gamecube was a sales disappointment too, and they basically reused that hardware for Wii. They went bold with the controller and design, but the CPU/GPU were very close to Gamecube. Not trying to predict anything here, just saying I could imagine a similar situation with Switch.

DS was also a custom GPU, so was 3DS and they were very successful. I mean every company can break from tradition I suppose. But until I see it I find it hard to believe they'd break from basically always using custom parts.

 

[Schnozberry]

He made the case that the new 28nm processes are very close to 20nm in power usage/efficiency. That's why I asked him why he thought that 28nm was possible (as needing a fan doesn't seem to point to 28HPC+ anymore than it does to 20nm given their similarities).

He did make a more important point regarding the vendors still producing 20nm chips though. But for me die size would be a concern at that process and would seriously limit Nintendo from making the kind of customisations they love to make.

What type of modifications are you expecting Nintendo to make other than using the more modern USB standard? Some cache customization is all I can think of.

 

[th3sicknness]

Nvidia didn't sell them the Tegra X1, and they don't make the Tegra X1 anymore. They licensed them their IP and assisted them in engineering a custom design. Nvidia is interested in getting a design win and gaining revenue. They don't care whether Nintendo goes bleeding edge or not.

I would imagine they care about a product that is going to have their name on it and growing a business relationship they said could last 10+ years.

 

[Gotdatmoney]

Yes and yes, those are the constraints for a 28nm node. It's hard to make the GPU larger while keeping the die size reasonable and it would also increase the thermal output and power consumption even if you were able to do it.

Do I have a good source I could read about how the differemt processes work? 28nm vs 20nm etc etc? I would love to get a better picture of how the manufacturing actually works.

 

[Raet]

Do we even know that the fan runs in portable mode?

Although if the lower clock speeds from Eurogamer are true for the final version, we can't really explain that unless it's 28 nm, can we?

 

[Donnie]

... Yeah, if that's the case this thing is not gonna be 20nm in any case. Maybe it's 16nm with 3-4SM, maybe it's 28 with 2SM, but it's definitely not 20nm. There's no point.

Personally, i'm team 28nm.

I wouldn't say their's no point in 20nm. I mean the new 28nm processes may be extremely close in energy efficiency but they aren't close in die size. Maybe a X1 chip would be ok on 28HPC+ despite being significantly larger and basically as power efficient as 20nm. But forget about being able to add anything to the die, like the embedded memory pools Nintendo are so fond of.

 

[MuchoMalo]

I would imagine they care about a product that is going to have their name on it and growing a business relationship they said could last 10+ years.

Well, they have two choices: do what Nintendo wants, or lose the contract. Which do you think they'll choose? Hint: They'll make the same choice that AMD, IBM and DMP did last gen.

 

[Schnozberry]

I would imagine they care about a product that is going to have their name on it and growing a business relationship they said could last 10+ years.

How is Nvidia harmed by the Switch not being on the latest process node?

 

[Smurfman256]

Do we even know that the fan runs in portable mode?

Actually, yes. It's right there in the patents.

 

[Donnie]

What type of modifications are you expecting Nintendo to make other than using the more modern USB standard? Some cache customization is all I can think of.

Embedded memory is the most likely, considering they tend to do it with every GPU they produce. Not for a frame buffer in this case of course. But perhaps as a compute buffer, or a shared cache for GPU/CPU.

 

[th3sicknness]

Well, they have two choices: do what Nintendo wants, or lose the contract. Which do you think they'll choose? Hint: They'll make the same choice that AMD, IBM and DMP did last gen.

Couldnt you say that about Nintendo as well? With the Wii U failing and the timing they were boxed in? The architecture that Nvidia is providing is uniquely perfect for the concept Nintendo was pushing.

 

[Donnie]

Because it's less likely to be 20nm at TSMC when clients have abandoned that process node and either went with 16nmFF or 28nm.

Of course TSMC refines their process nodes to improve leakage and power consumption.

http://www.tsmc.com/english/dedicatedFoundry/technology/28nm.htm

They've refined their 28nm and 16nm nodes but have hardly touched the 20nm nodes according to their foundry page.

So as Thraktor stated in a post a couple of pages ago, they have a 28nm HPC+ node which is very close in comparison to the 20nm node.

On the other hand there is 16nmFF+ and 16nmFFC which is obviously better in power consumption and density. The question is, which is cheaper and gives better yields?

Thraktor summarises the above in this post anyway so take a look: http://m.neogaf.com/showpost.php?p=227547584

At the end of the day Nintendo will customise the GPU the way it wants and the process used will be of secondary concern. If they are able to make the changes they want to the Tegra GPU with a new 28nm process then they'll probably use it. If not they'll use 20nm whether other people are using it or not. How many vendors were still using 45nm when Nintendo used it in WiiU? Yet Nintendo used it purely because it was the smallest process available that supported a feature they wanted to use in the GPU's design.

 

[Net]

There doesn't seem to be much of a reason to use 20nm. The price is too close to 16nm and the performance is too close to 28nm. If it weren't for the fact that TX1 is fabbed on 20nm we probably wouldn't even be considering it. That's not to say it's impossible, though. Perhaps TSMC is offering an exceptionally good deal to use up their remaining 20nm capacity.

The clock speeds we are aware of are perfectly doable on 28nm, and if Nintendo's goal is to make Switch as affordable as possible from day one then 28nm would seem like the sensible choice.

Welp, Thraktor has spoken.

28nm it is. What a nightmare.

 

[ChrisMahoney]

The recent patent indicates that it is IIRC.

It could be for hot climates to prevent throttling. In normal conditions it could be off

 

[Donnie]

Since when has any handheld or laptop or whatever ever had a fan just for hot climates?

 

[ggx2ac]

Do I have a good source I could read about how the differemt processes work? 28nm vs 20nm etc etc? I would love to get a better picture of how the manufacturing actually works.

If you want to learn how the actual manufacturing works, wiki would be the first source you can look at to get an idea: https://en.m.wikipedia.org/wiki/Semiconductor_device_fabrication

With regards to 28nm vs 20nm, that was in reference TSMC doing the process and the comparisons Thraktor made that you can look at.

http://www.tsmc.com/english/dedicatedFoundry/technology/28nm.htm

http://m.neogaf.com/showpost.php?p=227547584

The thing is then, you have to look at the Tegra X1 since we're making references to that with its die size and how it's area would increase going from 20nm to 28nm

The following is just for reference of a TX1:

 

[lwilliams3]

At the end of the day Nintendo will customise the GPU the way it wants and the process used will be of secondary concern. If they are able to make the changes they want to the Tegra GPU with a new 28nm process then they'll probably use it. If not they'll use 20nm whether other people are using it or not. How many vendors were still using 45nm when Nintendo used it in WiiU? Yet Nintendo used it purely because it was the smallest process available that supported a feature they wanted to use in the GPU's design.

I agree with that.

Welp, Thraktor has spoken.

28nm it is. What a nightmare.

With what Thraktor said, I don't believe so.

Of course, this is still just a theory.

 

[Schnozberry]

Embedded memory is the most likely, considering they tend to do it with every GPU they produce. Not for a frame buffer in this case of course. But perhaps as a compute buffer, or a shared cache for GPU/CPU.

I would think 1 or 2MB of shared L3 would be enough. It wouldn't take up that much space, even at 28nm.

 

[AzaK]

And someone with a PS4 or Xbone isn't gonna buy a new Nintendo console for third-party ports no matter how beefy it is. So, really, if Switch's success is based on ports as so many people believe, it was DOA before they even started working on it.

I tend to agree. Noone is going to buy a Switch to have another, lesser machine to play third party ports on. PS4/XBO owners will buy it if they really want handheld gaming (I don't think that's big for that audience per se) or it's so damned cheap they pick it up for Nintendo games. The real audience for it is the current 3ds owners upgrading and casual gamers who just play the odd Madden or Fifa etc and don't care about mega horse power.

 

[EDarkness]

I tend to agree. Noone is going to buy a Switch to have another, lesser machine to play third party ports on. PS4/XBO owners will buy it if they really want handheld gaming (I don't think that's big for that audience per se) or it's so damned cheap they pick it up for Nintendo games. The real audience for it is the current 3ds owners upgrading and casual gamers who just play the odd Madden or Fifa etc and don't care about mega horse power.

Ahem...I'm buying a NS for ports. I have a PS4, but would rather play them on the NS if possible.

 

[darkstar0155]

in a discussion with a friend, he was saying the Nintendo should just go all out with the switch and design it to be as powerful as a ps4 pro and take a huge loss (hahah). Ignoring the obvious economic and business reasons why this would be a horrible idea, is this even technologically possible in a handheld form factor? And if so, about what price would something like this be? Just sorta curious.

 

[7roject28]

Wasn't one of the rumored reason that they delayed the Switch was to put in better hardware? With Nvidia's wording, I'm guessing a custom Pascal chip or else I'll eat a 3DS cartridge! And it'll be $399!

 

[Hermii]

in a discussion with a friend, he was saying the Nintendo should just go all out with the switch and design it to be as powerful as a ps4 pro and take a huge loss (hahah). Ignoring the obvious economic and business reasons why this would be a horrible idea, is this even technologically possible in a handheld form factor? And if so, about what price would something like this be? Just sorta curious.

No way in a battery driven device.

However they could have made it a lot better than what we likely end up with.

 

[th3sicknness]

I tend to agree. Noone is going to buy a Switch to have another, lesser machine to play third party ports on. PS4/XBO owners will buy it if they really want handheld gaming (I don't think that's big for that audience per se) or it's so damned cheap they pick it up for Nintendo games. The real audience for it is the current 3ds owners upgrading and casual gamers who just play the odd Madden or Fifa etc and don't care about mega horse power.

I dont think this is the case to be honest. If the Switch got Rocket League and had decent online integration, local multiplayer, and or cross play I would buy it hands down no questions asked regardless of what it looked like. Also this could be an INDIE developers dream. It could run pretty much any indie game out there at close enough graphics fidelity to not bother anyone with added portability.

Also factor in that 3rd party games are costing MORE and MORE money every year to develop and the sales arent getting better. COD , BF1, TFALL, and HALO are all experiencing franchise fatigue so there are less and less heavy hitting 3rd party games. People are also forgetting pricing as well.. How is Nintendo going to price out games for this machine? Are we going to get 29.99 to 39.99 for DS type games or 49.99 to 59.99 for Wii U type games? Maybe to compensate for power we get pricing and portability?

 

[Vampirolol]

Wasn't one of the rumored reason that they delayed the Switch was to put in better hardware? With Nvidia's wording, I'm guessing a custom Pascal chip or else I'll eat a 3DS cartridge! And it'll be $399!

With Nintendo there is only one thing for sure: never underestimate how cheap they can get. Remember what Wii, 3DS and Wii U had inside.

 

[Hermii]

Wasn't one of the rumored reason that they delayed the Switch was to put in better hardware? With Nvidia's wording, I'm guessing a custom Pascal chip or else I'll eat a 3DS cartridge! And it'll be $399!

Be prepared to eat that 3ds cartridge.

 

[Net]

Wasn't one of the rumored reason that they delayed the Switch was to put in better hardware? With Nvidia's wording, I'm guessing a custom Pascal chip or else I'll eat a 3DS cartridge! And it'll be $399!

Record yourself eating the cartridge and put it on Youtube.

 

[blu]

On the other hand there is 16nmFF+ and 16nmFFC which is obviously better in power consumption and density. The question is, which is cheaper and gives better yields?

That's the epitome of a rhetorical question.

 

[Astral Dog]

in a discussion with a friend, he was saying the Nintendo should just go all out with the switch and design it to be as powerful as a ps4 pro and take a huge loss (hahah). Ignoring the obvious economic and business reasons why this would be a horrible idea, is this even technologically possible in a handheld form factor? And if so, about what price would something like this be? Just sorta curious.

No,of course not unless you want to carry a big box that burns your hands and malfunctions. Sorry but your friend is insane.

you can also look at an Xbox One S and PSlim to see even the small energy friendly version are far and away from a handheld friendly form

 

[Rolf NB]

Do we even know that the fan runs in portable mode?

Although if the lower clock speeds from Eurogamer are true for the final version, we can't really explain that unless it's 28 nm, can we?

Actually, yes. It's right there in the patents.

We don't know.

The patent providing a claim does not mean it will.
The USPO is not going to verify all claims in a patent are used to their fullest extent. There is absolutely no downside in making patents broad and overcovering. The patent will be just as valid if several of its claims are never implemented.

Even if we take a patent application as a spec sheet, which we absolutely should not, zero pretty much is a reduced number compared to any positive number.

 

[th3sicknness]

We don't know.

The patent providing a claim does not mean it will.
The USPO is not going to verify all claims in a patent are used to their fullest extent. There is absolutely no downside in making patents broad and overcovering. The patent will be just as valid if several of its claims are never implemented.

Even if we take a patent application as a spec sheet, which we absolutely should not, zero pretty much is a reduced number compared to any positive number.

Agreed. They could have included the fan in the originally filed patent to cover their bases and later in development remove it if headway was made towards not using it.

 

[lwilliams3]

With Nintendo there is only one thing for sure: never underestimate how cheap they can get. Remember what Wii, 3DS and Wii U had inside.

The Wii U, in particular, wasn't cheap. That is what made the Wii U situation such a big problem.

 

[darkstar0155]

No,of course not unless you want to carry a big box that burns your hands and malfunctions. Sorry but your friends is insane.

you can also look at an Xbox One S and PSlim to see even the small energy friendly version are far and away from a handheld friendly form

This was pretty much my main argument (from the tech side). I was like "dude, it would need to be as big as a gaming laptop, and the battery would last 2hrs max" He was all "I disagree, they could do it with the advancements in battery tech, and sell it at a loss for 500$ and sell like crazy" I think I literally smacked my head against my desk lol.

 

[UnemployedVillain]

With Nintendo there is only one thing for sure: never underestimate how cheap they can get. Remember what Wii, 3DS and Wii U had inside.

Neither of those consoles was made cheaply. You could argue Nintendo cheaped out on the 3DS battery or the Wii U screen, but the 3DS screen was state of the art when it was released