Switch 2 Speculation

Goodtwin

Veteran
Supporter
Although there has been plenty of speculation on there being a Switch Pro; I am of the opinion that it is more likely for Nintendo to release a true successor to the Switch in 2022 or 2023 at the latest. I feel like a 5-6 year generation is long enough, and will allow the OG switch to be sold for a few more years at a very attractive price carried by its extensive catalog of games already released. Essentially giving Nintendo the opportunity to sell a wider variety of hardware SKU's.

The processor is likely to be custom this time, but unlikely to be exotic. Looking at the current line up of ARM CPU cores, the A77 cores seem to support the same instruction set as the A57 cores, making backwards capability simple (feel free to correct me if I am mistaken on this). There is always the possibility that 4 A57 cores could be retained for this as well, seeing as how they arent going to take up much dye space.

GPU cores would most likely be made of of Volta cores. They are rather new at the moment, but this is Nintendo we are talking about, in 2022, do we really expect Nintendo to be using anything more modern? Doubtful. The 7nm process would be pretty inexpensive at this point, and would in theory make 1024 GPU Volta cores at 1-1.5Ghz plausible within the power constraints of a portable system.

Memory bandwidth is hard to speculate because of how rapidly LPDDR memory is advancing. However, the move to a 128 bit memory bus at the very least is almost a sure thing. Being somewhat memory bandwidth limited is likely to remain a burden for portable processors for many years to come.

There have been rumors that Nintendo has partnered with Sharp for a new LCD screen. I do not see much benefit from going beyond a 1080p resolution in a screen that will likely be no larger than 7 inches, but color and contrast, as well as blacks are always a variable with LCD screens.

Nintendo has proven with Switch that their hybrid platform doesn't need, nor will it have parity with competing console manufactures when it comes to third party software support. So while the performance I have laid out in my speculation hardly puts it within a stones throw of what the PS5 will be capable of, it also wont be in all that different of a deficit as the Switch is with the PS4/X1, with the added benefit of being a proven platform that consumers are eager to purchase.
 
There are no details on Tegra Orin, but maybe they'll just shrink Xavier and halve the bus & strap on LPDDR5. Not sure on Denver compatibility though - seems like it shouldn't be an issue, but I don't follow these things. ¯\_(ツ)_/¯
 
A nice upgrade would be a lot more internal storage space imo. I don't own a switch, but that's the first complain I heard, before raw power.
 
A nice upgrade would be a lot more internal storage space imo. I don't own a switch, but that's the first complain I heard, before raw power.

yeah its a huge problem for digital purchases and non-nintendo physical purchase (e.g. MK physical took 10GB+ space of storage). Fortunately microsd cards are dirt cheap if you wait for a discount.

if not, that it'll be like me. Bought 64 GB card with Switch, doh! its not enough. then bought 256 GB card and currently still happy with it.
 
In digital foundry videos, they sometime said that the games on sd cards are slower that internal memory ... ?
 
In digital foundry videos, they sometime said that the games on sd cards are slower that internal memory ... ?

Microsd speed varies.

There's also clock thingy (I don't understand) in switch that affect memory card speed. The dev of hekate keeps updating it in almost all hekate update.

Could be that Nintendo did the safest, slowest, most compatible clock for microsd
 
Memory bandwidth is hard to speculate because of how rapidly LPDDR memory is advancing. However, the move to a 128 bit memory bus at the very least is almost a sure thing.
It is?
I doubt that. They'd be more likely to adopt 64bit LPDDR5 6400 MT/s (effectively twice the effective bandwidth of TX1) than going all the way up to 128bit.


The processor is likely to be custom this time
Not if nvidia asks too much money for it. Which they might, since they never made any semicustom SoC for anyone so it's out of their current core business.
Nintendo could also just adopt a cut-down version of whatever nvidia develops for their next shield TV, assuming they'll make one for 8K60 / HDMI 2.1.


There are no details on Tegra Orin
We know nvidia plans to replace the 500W Pegasus board with two Orin boards.
So it's around 250W worth of 16FF+/12FFN chips. Orin is not what's going into a Switch 2 even if it releases in 2026 using 3nm.

maybe they'll just shrink Xavier and halve the bus & strap on LPDDR5. Not sure on Denver compatibility though - seems like it shouldn't be an issue, but I don't follow these things. ¯\_(ツ)_/¯
Xavier is way too focused on ML and automobile. I'm not sure if Denver is 100% ISA compatible with ARMv8 either.





I think there's a good chance Nintendo might go with a semicustom Samsung/AMD SoC with ARMv8 + RDNA for a Switch 2, even if they have to forego backwards compatibility in the process.

This time it was Nintendo's turn to learn that nvidia isn't a great partner for consoles, and it came in the form of the massive piracy that came true due to terrible security in the platform.
(I thought about linking to reddit's subs for piracy on the Switch but it might be against the rules and anyone can search for it by themselves.)
 
That is a bit of an exaggeration. Wii and DS were easily hacked and I think it was the same for 3ds and wuu? I didn't own those so never really looked into it.

The Switch exploit only affects units sold in the first 15 months or so so overall impact will be limited. A quick look at the procedure doesn't make it look like something the average user will do either. Certainly it's not as easy as it was on let's say the DS where you only had to copy a rom to a ad card. No hacking on the DS itself required.
 
That is a bit of an exaggeration. Wii and DS were easily hacked and I think it was the same for 3ds and wuu? I didn't own those so never really looked into it.
I don't know about Wuu piracy (did anyone even care?).
The others were easily hacked, but you couldn't access entire online stores where you download games for free, and install Android in there to use it for other purposes.
That's like hackers finding a way to install Windows and run pirated PC games on the PS4.


The Switch exploit only affects units sold in the first 15 months or so so overall impact will be limited.
AFAIK it affects all units with a TX1, even the ones being sold right now.


A quick look at the procedure doesn't make it look like something the average user will do either. Certainly it's not as easy as it was on let's say the DS where you only had to copy a rom to a sd card. No hacking on the DS itself required.
It sure looks like installing SXOS is exactly that.
 
As far as I know it only works on switches made before June 2018. It even says so on the sxos page. Which by the way costs money. Does that load off a game card or SD without any user interaction like on the DS?

Sure you can install Android. But apart from "because we can", how many people are actively going to buy a Switch just for that?

I just don't think it's a major issue. Certainly not to the point where Nintendo would have felt burnt by Nvidia. If any it seems Nintendo is benefiting a lot from their relationship with Nvidia.
 
Not if nvidia asks too much money for it. Which they might, since they never made any semicustom SoC for anyone so it's out of their current core business.
Nintendo could also just adopt a cut-down version of whatever nvidia develops for their next shield TV, assuming they'll make one for 8K60 / HDMI 2.1.
Is there any confidence in that? Seems more trouble than its worth seeing as Shield isn't selling in large numbers and there's loads of competition and all nVidia's efforts could be focussed elsewhere in more lucrative markets (where they're starting to face stiffer competition). Do they have a mobile roadmap? If not, a future Switch2 is unlikely to use nVidia IMO. Nintendo have been happy to jump architecture for the mobile devices before so I don't think they'd sweat another transition and they can pick a more reliable partner next time if they start to value hardware compatibility for future devices.
 
maybe they'll shrink switch's tegra even more and duct tape them together. like what they did with GC cpu in Wii, Wii U

Seeing as how it is Nintendo, I have considered this. Even though I expect the processor to be "custom", I doubt it will be anything exotic. Seeing as how Nvidia already has the license for the A57 CPU cores, I could see them using 8 of them, and running at 2 Ghz thanks to the newer process. Same goes with the GPU cores, double them up, and clock much higher. They already used a 128 bit memory bus on the Tegra X2, so perhaps that chip would be the starting point. Nvidia has shifted the Tegra line towards automotive, so there wont be an off the shelf part that is going to fit the bill.
 
If I were betting on a Switch 2, I'd guess it would be a max of 2X perf over the current Switch. A72 x 4 + double the GPU performance on a cheap 7nm process around ~2022. It would be modest boost in performance and in line with what they usually deliver.
 
As far as I know it only works on switches made before June 2018.
Upon further reading, you're correct.
There are other (free) methods to hack the switch, but it seems they're still working on a solution for both the Mariko models and the ones with the 2018 patch.


Is there any confidence in that?
Nope, there's not much confidence in nvidia developing a semicustom chip just for Nintendo despite a higher predicted volume this time.
Maybe if Nintendo's 1st party devs are successfully baited into using nvidia's SDK for long enough to put pressure on the higher ups to pay top dollar for a semicustom made by nvidia..?

Otherwise, there's zero mention of mobile SoCs on nvidia's roadmaps. Xavier is a 30W SoC and Orin seems to be around 100W if it's 7nm.


If not, a future Switch2 is unlikely to use nVidia IMO. Nintendo have been happy to jump architecture for the mobile devices before so I don't think they'd sweat another transition and they can pick a more reliable partner next time if they start to value hardware compatibility for future devices.
Agreed.
Which is why I mentioned Samsung who will have GPUs from AMD potentially with SDKs made by them too, plus they use their own fabs so they might offer a more competitive price at a given node.
As for other SoC makers, I don't see Nintendo going with chinese companies and Qualcomm might not be interested in sharing too many details on Adreno, so maybe AMLogic?
Though at this point without someone who takes over the SDK development maybe they'd prefer to make a custom SoC themselves using Cortex A7x cores and Mali graphics.
 
Otherwise, there's zero mention of mobile SoCs on nvidia's roadmaps. Xavier is a 30W SoC and Orin seems to be around 100W if it's 7nm.
I think this is the main point against nVidia as a supplier of the "Switch 2" SoC. They have simply left the mobile scene. Would it make sense for them to make a one-off design on 7nm or 5nm? Would it make sense for Nintendo to pay for it?

Which is why I mentioned Samsung who will have GPUs from AMD potentially with SDKs made by them too, plus they use their own fabs so they might offer a more competitive price at a given node.
A possibility.
As for other SoC makers, I don't see Nintendo going with chinese companies and Qualcomm might not be interested in sharing too many details on Adreno, so maybe AMLogic?
Though at this point without someone who takes over the SDK development maybe they'd prefer to make a custom SoC themselves using Cortex A7x cores and Mali graphics.
I think Qualcomm would be a real possibility. They are likely to have finished mobile designs on relevant processes, either for straight adoption or for relatively simple modification. Contracts of Switch 2 size bring bargaining power, and cutting out some cell phone functions while extending capabilities as Nintendo sees fit would seem relatively straight forward and doable from a cost perspective.
 
Though at this point without someone who takes over the SDK development maybe they'd prefer to make a custom SoC themselves using Cortex A7x cores and Mali graphics.
That's a very good point. As I understand it, development on Switch is very easy and that'll be because of nVidia. Prior to that, Nintendo's not had a great reputation for ease of development, and I'd worry for devs' sanity if there's not a strong development partner.
 
That's using the TX1 refresh, and we already have the Switch update with that one. AFAIK that's the end of the road for nVidia's mobile architecture, and presumably a last Shield TV design? What will nVidia have in 2022 that can go into a Switch or nVidia Shield?
 
That's a very good point. As I understand it, development on Switch is very easy and that'll be because of nVidia. Prior to that, Nintendo's not had a great reputation for ease of development, and I'd worry for devs' sanity if there's not a strong development partner.

Yes, that's where Samsung's partnership or licensing of tech from AMD becomes potentially interesting. If the GPU tech remains similar to what it is in the PC space, then it should make things relatively easier than a purely mobile part or some custom GPU part. Probably nothing like the support Nintendo gets from NVidia (whatever that is), but worlds better than what existed for developers with the DS series, 3DS series, Wii, Wii-U, etc.

IIRC - the licensing agreement only prohibits products including AMD IP that would directly compete in markets that AMD is currently in. That basically opens up the mobile/portable gaming market avenue for Samsung using AMD GPU IP.

Regards,
SB
 
Back
Top