Switch 2 Speculation

[Pressure]

I just had a random thought that an evolution of the Switch concept would be the inclusion of AR glasses, so the player can have a big screen anywhere they go.

I don't think any casual gamer is asking for that ...

Regardless you can only do so much at a $299 starting price point for the base model.

 

[Frenetic Pony]

I don't think any casual gamer is asking for that ...

Regardless you can only do so much at a $299 starting price point for the base model.

I'm guessing $349 will be the minimum price for 2, wouldn't even be the most shocked at $399.

Also it's Nintendo, AR glasses would hardly be out of line for them if they thought there was a decent chance it would sell. Not that I think good AR glasses are anywhere near a Nintendo price point, or even a reality at all, anytime in the next few years. Maybe not this decade. But a VR addon for Switch 2, that I can see. Sell them Metroid Prime 1 remake and Prime 4 and Skyward Sword and New Pokemon Snap again, this time in VR!

 

[rntongo]

I think the Switch 2 will be $499 at launch

 

[Frenetic Pony]

I think the Switch 2 will be $499 at launch

Oh c'mon, it's Nintendo. That's what Nintendon't

 

[Charlietus]

I think the Switch 2 will be $499 at launch

If we look at the leaked components, it's probably 399. Especially if the chip is Samsung 8nm (please don't be Samsung 8nm...)

 

[rntongo]

Oh c'mon, it's Nintendo. That's what Nintendon't

lmao just hoping they go for a more powerful device than whats been rumoured

If we look at the leaked components, it's probably 399. Especially if the chip is Samsung 8nm (please don't be Samsung 8nm...)

Yes if it remains the same as the leaked components. I just dont see how they would release an 8nm device in 2025. But $399 is the most likely price if we're being serious

 

[Charlietus]

lmao just hoping they go for a more powerful device than whats been rumoured

Yes if it remains the same as the leaked components. I just dont see how they would release an 8nm device in 2025. But $399 is the most likely price if we're being serious

Not only 8nm, but the Samsung variant, that is notoriously inefficient

 

[rntongo]

Not only 8nm, but the Samsung variant, that is notoriously inefficient

Yeah this doesnt make any sense at all in 2025. I think those leaks are of old hardware.

 

[Charlietus]

Yeah this doesnt make any sense at all in 2025. I think those leaks are of old hardware.

I sure hope so. TSMC 4nm would probably make the chip half the size with almost twice the battery life. But who knows with Nintendo, maybe Samsung gave them a great deal.

 

[Tkumpathenurple]

From a quick bit of googling, it seems that the original Switch was 20nm just as the PS4 and XB1 transitioned to 16nm, so it would be pretty in-keeping for the Switch 2 to be a touch behind the likely 5nm PS5Pro.

 

[cheapchips]

From a quick bit of googling, it seems that the original Switch was 20nm just as the PS4 and XB1 transitioned to 16nm, so it would be pretty in-keeping for the Switch 2 to be a touch behind the likely 5nm PS5Pro.

Nintendo's goal with silicon design seem to be go with even older design than the Switch's, despite more modern alternatives being offered by Nvidia. If the process nodes more modern than 8nm we should be shocked and surprised.

 

[Frenetic Pony]

The "goal" of Nintendo is, no doubt, to get to their performance/price ratio then not give a fuck because they're Nintendo.

From the parts shipping manifest for the Switch 2 we can see LPDDR5 7500 listed, assuming a 128bit bus and Nvidia 3XXX we can assume up to 17% more performance than a Steamdeck at highest power draw settings. Which sounds exactly in line with Nintendo not giving a fuck beyond that territory.

 

[Newguy]

The specs definitely say "it was ready to be released in 2022 but the Switch was selling so well we delayed it multiple times". Can't blame them for wanting free money but because of that we'll get a console that's multiple arch gens out of date. For a nintendo handheld it'll be great considering their historic use of very outdated hardware in handhelds (switch excluded) but the enthusiast in me is a little underwhelmed. It could be a lot worse™

 

[Subtlesnake]

The specs definitely say "it was ready to be released in 2022 but the Switch was selling so well we delayed it multiple times". Can't blame them for wanting free money but because of that we'll get a console that's multiple arch gens out of date. For a nintendo handheld it'll be great considering their historic use of very outdated hardware in handhelds (switch excluded) but the enthusiast in me is a little underwhelmed. It could be a lot worse™

1536 ALUs is double what Rog Ally has. On a modern process it would be an extremely ambitious product.

 

[Tkumpathenurple]

The specs definitely say "it was ready to be released in 2022 but the Switch was selling so well we delayed it multiple times". Can't blame them for wanting free money but because of that we'll get a console that's multiple arch gens out of date. For a nintendo handheld it'll be great considering their historic use of very outdated hardware in handhelds (switch excluded) but the enthusiast in me is a little underwhelmed. It could be a lot worse™

Agreed. Although they're a bit of a wildcard, so maybe the leaks aren't quite right and it'll be a touch more modern.

The "goal" of Nintendo is, no doubt, to get to their performance/price ratio then not give a fuck because they're Nintendo.

From the parts shipping manifest for the Switch 2 we can see LPDDR5 7500 listed, assuming a 128bit bus and Nvidia 3XXX we can assume up to 17% more performance than a Steamdeck at highest power draw settings. Which sounds exactly in line with Nintendo not giving a fuck beyond that territory.

Wouldn't that make it just over 100GB/s bandwidth? And isn't RT fairly bandwidth intensive?

Nvidia's definitely the leader in terms of RT but it'll be interesting to see how well they can power through or work around such a limitation.

 

[Newguy]

1536 ALUs is double what Rog Ally has. On a modern process it would be an extremely ambitious product.

Sort of yes but 2x FP throughput is an arch feature of Ampere, it's not the same as 2x SMs/CUs of the ROG Ally. It'll still be 12 SMs vs 12 CUs so they'll have a similar distribution of TMUs/ROPs (48 TMUs, 32 ROPs) although it's probably 1 shader engine for AMD and 2 GPCs for Nvidia but don't quote me on AMD's numbers, I want to say it's 32 ROPs per shader engine for AMD (can be cut down). Low level API + Nvidia support + not using windows will work in its favour more than 2x FP throughput imo, "I don't know how x runs on [console]" has been a constant given enough time and effort and this will be no different

 

[Subtlesnake]

Sort of yes but 2x FP throughput is an arch feature of Ampere, it's not the same as 2x SMs/CUs of the ROG Ally. It'll still be 12 SMs vs 12 CUs so they'll have a similar distribution of TMUs/ROPs (48 TMUs, 32 ROPs) although it's probably 1 shader engine for AMD and 2 GPCs for Nvidia but don't quote me on AMD's numbers, I want to say it's 32 ROPs per shader engine for AMD (can be cut down). Low level API + Nvidia support + not using windows will work in its favour more than 2x FP throughput imo, "I don't know how x runs on [console]" has been a constant given enough time and effort and this will be no different

On the PC we saw a 25-30% speedup for Ampere's doubling of FP32, so it's still a substantial advantage, relative to the Rog Ally, which is already considered an "enthusiast" product.

Wouldn't that make it just over 100GB/s bandwidth? And isn't RT fairly bandwidth intensive?

Nvidia's definitely the leader in terms of RT but it'll be interesting to see how well they can power through or work around such a limitation.

It should be 120 GB/s for a 128-bit bus. But the Rog Ally already scales well to higher power levels with its 102.4 GB/s. So I don't see how 120 GB/s is some terrible bottleneck when we would be lucky to reach 15W docked on a Nintendo handheld.

 

[Newguy]

On the PC we saw a 25-30% speedup for Ampere's doubling of FP32, so it's still a substantial advantage, relative to the Rog Ally, which is already considered an "enthusiast" product.

Also fair but again power budgets, handheld the Switch uses about 7-8W max (online play) total system from my own battery life use, docked 15W give or take for the entire system. The Ally uses about 25W total in the 15W TDP mode, Steam Deck is about 32W average at 15W (40Wh battery, 1.25 hour total time = 32W):

Might have changed as it's a may 2023 video but big total power draw differences are difficult to make up, although at low power the Ally is very underwhelming vs deck so I expect a big advantage for the Switch successor there. Higher TDPs like 25W should still favour the Ally for a couple years but by 2030 I expect the Switch 2 to pull ahead. But if handheld PCs are being made newer versions should have a good lead so whatever floats your boat

 

[Subtlesnake]

Also fair but again power budgets, handheld the Switch uses about 7-8W max (online play) total system from my own battery life use, docked 15W give or take for the entire system. The Ally uses about 25W total in the 15W TDP mode, Steam Deck is about 32W average at 15W (40Wh battery, 1.25 hour total time = 32W):

Might have changed as it's a may 2023 video but big total power draw differences are difficult to make up, although at low power the Ally is very underwhelming vs deck so I expect a big advantage for the Switch successor there. Higher TDPs like 25W should still favour the Ally for a couple years but by 2030 I expect the Switch 2 to pull ahead. But if handheld PCs are being made newer versions should have a good lead so whatever floats your boat

I am not saying the Switch 2 will beat the Rog Ally when the latter is using its higher power modes. I'm saying if Nintendo go with a modern process, then it will be an ambitious product, given the TDP it is targeting. The extra width on the GPU gives them the potential to achieve the same (or greater) performance as the Steam Deck/Rog Ally (15W) at lower clocks. That allows them to release a more portable product, and in handheld mode they will likely get better battery life at the expense of worse performance. (Obviously depending on the size of battery they can fit)

 

[raytracingfan]

The rumored T239 chip is based on the Orin SoC that is manufactured on the 8nm process, correct? Hypothetically speaking, what would be the R&D costs to adapt the design to a different process?