Switch 2 Speculation

[Shifty Geezer]

Switch 2 will be playing new games just fine. And it has a docking station for "console" like gaming. It is the same market.

It's clearly not. NSW2 will sell >100 million units. How many will Rog Ally et al sell? Or are you expecting a good 60 million each with the Ally cannibalising Nintendo's market?

NSW is for Nintendo gamers. Portable PCs are for PC gamers. They're both the same market in that they buy and play video games, but they have their distinction audiences and revenue channels.

 

[troyan]

"Portable" PCs playing the same games. You can play Red Dead Redemption 1 on a Switch Lite for $199 or on a Alloy X for $799.

My point is that a better process than 8nm doesnt have enough advantages to offset the >3x higher wafer prices and the higher development costs. And similiar performance is possible with a mobile SoC with a (very) wide GPU.

 

[Subtlesnake]

My point is that a better process than 8nm doesnt have enough advantages to offset the >3x higher wafer prices and the higher development costs. And similiar performance is possible with a mobile SoC with a (very) wide GPU.

You have a source for the claim that 5LPP is >3x more per wafer?

 

[Chiimaero]

You have a source for the claim that 5LPP is >3x more per wafer?

Let's be generous here and ask for proof that 8nm is 3x more expensive than TSMC 4nm!
There were a lot of educated guesses on the Switch 2 famiboard thread which came to the conclusion that the difference in price should actually be pretty minimal. Which I'm inclined to believe given we have Android consoles sporting a high-end Qualcomm chip based on TSMC 4nm, available now, for $379.

Unless one is part of the few people on this world who has access to whatever deal Samsung and NVidia signed for T239, it's simply impossible to compare per chip prices across fabs and nodes.

My point is that a better process than 8nm doesnt have enough advantages to offset the >3x higher wafer prices and the higher development costs. And similiar performance is possible with a mobile SoC with a (very) wide GPU.

Same pb here, we don't really know how large of a gap a move to a vastly better process (N4) would grant.
Say it gives the Switch 2 at least 30% higher clocks... then even if it costs more, having such a perf gap for a device that is supposed to last 8+ years is enough and then some. It could mean:
- 60fps games instead of 40fps one
- 30fps games instead of cannot run at all so won't be ported
- less work for devs to get the same output as less optimisation is needed
I view this as a big deal, and it will be a real bummer if it turns out T239 is indeed a Samsung 8nm chip.

The actual funny thing is that if it's 8nm, it will inevitably have a revision later on down the road (5LPP or 4LPP) and we will then be able to know exactly what the Switch 2 should have been from the start as there is no "rushed to market" excuse this time around.

 

[LiveGamer]

Excuse my ignorance but even if the wafer is more expensive wouldn't a smaller node allow for more SOCs per wafer offsetting the cost?

 

[Chiimaero]

Excuse my ignorance but even if the wafer is more expensive wouldn't a smaller node allow for more SOCs per wafer offsetting the cost?

Sure. Die size is important. Yield is important. Business is important (will you pay per wafer? per functional chip? can you use binned chips?...). I guess politics can be equally important (if the foundry is desperate for customers, it's likely they will offer interesting prices). That's why it's impossible to state that chip X is more expensive than chip Y. Too many variables.

 

[troyan]

Excuse my ignorance but even if the wafer is more expensive wouldn't a smaller node allow for more SOCs per wafer offsetting the cost?

Production costs are skyrocketing with every new node. Density improvements dont offset it. Development costs are higher, too. Overall it only gets worse going forward. And it doesnt matter how old the process is.

AMD wanted to use chiplets to offset these higher 5nm costs. nVidia is trying to build small dies (Blackwell uses GDDR7 instead of a wider memory interface). Nintendo is in a situation that they dont need the latest process node. A unique position allows for a different approach.

 

[Subtlesnake]

In the graphs I have seen here, the cost per transistor is basically flat, not increasing sharply. The reason to go for a newer node in the case of Nintendo would be to meet their power/efficiency targets, and as 5LPP is based on Samsung's 7nm node, the cost per wafer may not have massively increased vs 8nm. Ultimately it's up to Samsung to decide what kind of deal to give Nvidia/Nintendo.

 

[Jubei]

"Portable" PCs playing the same games. You can play Red Dead Redemption 1 on a Switch Lite for $199 or on a Alloy X for $799.

My point is that a better process than 8nm doesnt have enough advantages to offset the >3x higher wafer prices and the higher development costs. And similiar performance is possible with a mobile SoC with a (very) wide

Sure you can play the same games. But not with the same fidelity. I think what people were trying to point out to you is that a family with children looking to buy a 399 USD Switch is not gonna suddenly consider a 799 Rog Ally because it runs Witcher 3 better

They were looking to buy the Switch for Mario and Zelda. And the person who is looking at a Rog Ally is not gonna downgrade for a Switch if all they are about is Call of Duty and Helldivers 2

Rog Ally X will outperform Switch 2 because it has a much larger battery and can afford to run at higher wattage while offering the same battery life. And when ”docked” you can simply connect it to a 4090 egpu. Its not even a matter of debate, if money is no object, PC always wins

As for your second point. Nobody knows the wafer prices for the type of volume a company like Nintendo needs. But regardless if you design wide or fast, TSMC 4nm is gonna offer much better perf/watt. Going with 8nm is about cost, not optimal design

 

[troyan]

The Switch was sold to every demographic. There is a reason why this console has a huge varity in games. It even got games which are not released on the Xbox. I mean the Switch is so big that WB has ported Hogwarts to the system...

At this moment Apple is using the first 5nm A14 chip for the cheapest iPad. So looking at Anandtech numbers the A14 is >2x more efficient than Samsung's 8nm Exynos 9820: https://www.anandtech.com/show/16192/the-iphone-12-review/4
Even for a two full node difference the impact isnt as big as the financial cost. And a bigger mobile SoC on 8nm with lower clocks (Exynos 9820 runs at 704Mhz or so) will increase efficiency further.

 

[Chiimaero]

Sure you can play the same games. But not with the same fidelity. I think what people were trying to point out to you is that a family with children looking to buy a 399 USD Switch is not gonna suddenly consider a 799 Rog Ally because it runs Witcher 3 better

They were looking to buy the Switch for Mario and Zelda. And the person who is looking at a Rog Ally is not gonna downgrade for a Switch if all they are about is Call of Duty and Helldivers 2

Rog Ally X will outperform Switch 2 because it has a much larger battery and can afford to run at higher wattage while offering the same battery life. And when ”docked” you can simply connect it to a 4090 egpu. Its not even a matter of debate, if money is no object, PC always wins

As for your second point. Nobody knows the wafer prices for the type of volume a company like Nintendo needs. But regardless if you design wide or fast, TSMC 4nm is gonna offer much better perf/watt. Going with 8nm is about cost, not optimal design

The Switch was sold to every demographic. There is a reason why this console has a huge varity in games. It even got games which are not released on the Xbox. I mean the Switch is so big that WB has ported Hogwarts to the system...

At this moment Apple is using the first 5nm A14 chip for the cheapest iPad. So looking at Anandtech numbers the A14 is >2x more efficient than Samsung's 8nm Exynos 9820: https://www.anandtech.com/show/16192/the-iphone-12-review/4
Even for a two full node difference the impact isnt as big as the financial cost. And a bigger mobile SoC on 8nm with lower clocks (Exynos 9820 runs at 704Mhz or so) will increase efficiency further.

Who buys a Switch, and why:
I agree with Troyan here. It's very hard to tell why people buy a Switch. But asserting it's "for Nintendo games" as a general rule is almost guaranteed to be false.
I think this stands true for ~15M people (being generous here) which is more or less what the WiiU sold.
The rest... some were looking for a dedicated portable device. Some for the cheapest option to have two childen happy. Some because they needed a new screen during the pandemic. And so on.

The Switch, even being rushed to the market, was presumably the best dedicated handheld gaming device for the majority of its lifespan. This won't be the case this time around. And I can totally see parents going from a Switch to a handheld PC for a child turning 15 yo for example. There is competition this time around, so people in the market for a new device will buy the one that fits their needs best.

The cost of SEC8N vs TSMC N4:
We don't know and no one can assert the reason for going with Samsung was cost (and cost only). And going with SEC8N means Nintendo will have to pay for the inevitable 5LPP/4LPP revisions as there is no way this chip can fit inside a lite version. So if it turns out to be SEC8N, it will be frustrating on many levels for sure. Consumers will have to decide between the competition and this "weaker than what could have been" Switch 2. And Nintendo will have to deal with this choice of them in case the sales are not that good after the first year craze.

 

[Jubei]

Yeah im really not getting any of your arguments. Two things are irefutable

1. PC handhelds are/will be more powerful than Switch 2

2. Switch 2 will outsell any PC handheld 50:1

 

[Shifty Geezer]

I think this stands true for ~15M people (being generous here) which is more or less what the WiiU sold.
The rest... some were looking for a dedicated portable device.

Why ignore the GB/DS/3DS audience? There were many millions who wanted Nintendo games on those handhelds. Brain Training, Nintendogs, Animal Crossing, etc., not to mention the myriad Marios.

The Switch, even being rushed to the market, was presumably the best dedicated handheld gaming device for the majority of its lifespan. This won't be the case this time around. And I can totally see parents going from a Switch to a handheld PC for a child turning 15 yo for example.

Let's revisit this in 5 years. I never make virtual bets as mostly it's guesswork, but in this case I'll happily put my name to a view that this is not going to happen and NSW will outsell PC handhelds by 10 to 1 (illustrative figures, 9.3:1 doesn't count as a loss on my part!).

Even if PC handhelds are comparable and plenty of people are just looking for a way to play GTA or COD or Fortnite portably, Nintendo's branding eclipses anything out there. So just as people were happy to buy inferior consoles because they were the popular ones, people will be happy with a less powerful handheld when it's the brand that's associated with portable gaming, has been for 3.5 decades, offers a good enough experience and at a lower price.

 

[Pressure]

Game	Platform	Copies sold
Mario Kart 8 / Deluxe	Wii U / Nintendo Switch	72,730,000
Animal Crossing: New Horizons	Nintendo Switch	46,450,000
Super Smash Bros. Ultimate	Nintendo Switch	35,140,000
The Legend of Zelda: Breath of the Wild	Wii U / Nintendo Switch	33,990,000
Super Mario Odyssey	Nintendo Switch	28,500,000
New Super Mario Bros. U	Nintendo Switch	26,660,000
Pokémon Sword / Shield	Nintendo Switch	26,440,000
Pokémon Scarlet / Violet	Nintendo Switch	25,690,000
The Legend of Zelda: Tears of the Kingdom	Nintendo Switch	20,280,000

 

[QPlayer]

Yeah im really not getting any of your arguments. Two things are irefutable

1. PC handhelds are/will be more powerful than Switch 2

2. Switch 2 will outsell any PC handheld 50:1

Wii ---> WiiU says hello!

Let's wait and see how it turns out.

 

[Shifty Geezer]

Wii ---> WiiU says hello!

Not in any way comparable as precedent. So many other factors at play between Wii and WiiU.

One can just as readily say "WiiU ---> NSW says hello. Wii U sold 14 million, Switch sold 10x more, 140 million. Ergo NSW2 will sell 10x more also, 1400 million"

 

[QPlayer]

Not in any way comparable as precedent. So many other factors at play between Wii and WiiU.

One can just as readily say "WiiU ---> NSW says hello. Wii U sold 14 million, Switch sold 10x more, 140 million. Ergo NSW2 will sell 10x more also, 1400 million"

Everything can be compared. After the success of the Wii, everyone was convinced that the next N console could only be a success, reminder: it prints the money!!

What is the guarantee that the next N console will be called Switch 2? But if that's the name, what's the guarantee that it will copy the same concept as its predecessor? Reminder: Nintendo has always experimented with new concepts regardless of console success. Let's just say I thought they were crazy for that, but it's still true. Then it either came in or it didn't.

How successful it will be will be revealed only after 1-2 years.

 

[orangpelupa]

https://www.reddit.com/r/NintendoSwitch2/comments/1hu3rte/new_joycon_photos_from_chinese_sns