• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Can people please understand that Switch 2 COULD be more powerful than Steam Deck?!

Impotaku

Member
Amazing all these years have gone by and they still don't understand, you increase a handhelds graphical power the battery life goes to shit. Unless they make it huge & stuff a massive battery in there you are going to only get a tiny battery life similar to valves deck. If a portable has to be close to a plug socket to keep power it's not much of a portable at all. So unless battery tehnology improves there's not much point in making a graphical powerhouse. Not once in the entire handheld timeline has a powerful handheld won.
 
Last edited:

RoadHazard

Gold Member
The Steam Deck uses an SoC in N7 process node, with RDNA2 and Zen2.
By the time the Switch 2 comes out, in a couple of years, it can use the N3 process node. Or even the N3 Finflex, if Nintendo wants the vey best.
This means much higher transistor density, bigger chip, more units, more TFLOPs, etc.
And then it can use much more advanced CPUs and GPUs, than what the Steam Deck has today. Regardless of Nintendo sticking with nVidia or going AMD.

The idea that the Switch 2 can't be more powerful than the Steam Deck is complete non-sense from someone that doesn't understand tech.

The problem with that is that Nintendo doesn't want the very best. They want the cheapest acceptable components they can get their hands on, so they can make as much money as possible from it. Nintendo doesn't do loss leaders.
 

winjer

Gold Member
The problem with that is that Nintendo doesn't want the very best. They want the cheapest acceptable components they can get their hands on, so they can make as much money as possible from it. Nintendo doesn't do loss leaders.

But in 2024-2025, something like RDNA3 or Ada Lovelace, won't be the best. And still it will be much better than the Steam Deck.

Are people here so oblivious to the constant improvement of tech?
 

RoadHazard

Gold Member
But in 2024-2025, something like RDNA3 or Ada Lovelace, won't be the best. And still it will be much better than the Steam Deck.

Are people here so oblivious to the constant improvement of tech?

Well, the Tegra hasn't been evolving in the same way as desktop GPUs.
 

rushgore

Member
LOL. Comparing floating point operations/sec across different architectures (let alone different brands altogether).
 

JCK75

Member
I'm sure it will be, but then Steam Deck 2 will be more powerful thant Switch 2.. and SD2 may actually launch first.
 

winjer

Gold Member
Well, the Tegra hasn't been evolving in the same way as desktop GPUs.

nVidia Orin is releasing this year, with Ampere arch. And it has 2048 cuda cores and 64 tensor units.
It's a 4 TFLOP SoC, that can also do DLSS.
Nvidia has already announced Atlan, in 2021, as it's next SoC.

By 2024-2025, any of these can be used by Nintendo, and still be more powerful than the Steam Deck.
 

Danjin44

The nicest person on this forum
nI7mrAk.gif
 

RoadHazard

Gold Member
nVidia Orin is releasing this year, with Ampere arch. And it has 2048 cuda cores and 64 tensor units.
It's a 4 TFLOP SoC, that can also do DLSS.
Nvidia has already announced Atlan, in 2021, as it's next SoC.

By 2024-2025, any of these can be used by Nintendo, and still be more powerful than the Steam Deck.

Can they be used in a device like the Switch with similar battery life without making the battery and thus the device much larger (battery tech has not really gone anywhere in a long time)?
 

winjer

Gold Member
Can they be used in a device like the Switch with similar battery life without making the battery and thus the device much larger (battery tech has not really gone anywhere in a long time)?

Of course. Remember that the Steam Deck uses the N7 process node from TSMC. The N3 process starts mass production in the second half of 2022.
By 2024-2025 the N3 process will no longer be leading edge.

Now consider that N5 offers 40% power reduction over N7.
And N3 offers another 30% power reduction over N5.
 
It could, steam deck is not even that powerful.

But this is Nintendo, if they had a 2 Tflop part they would downgrade it at least 30% making it functionally behave worse than Steam Deck who's a 1.6 Tflop part.


Also bare in mind that Steam Deck is "powerful" for a 720p screen, but the switch also plugs onto a TV, so at the same flop output they would already be at a disadvantage at 1080p but they'll have pressure to go higher. I guess that's where the tensor cores will come in.
Of course. Remember that the Steam Deck uses the N7 process node from TSMC. The N3 process starts mass production in the second half of 2022.
Nintendo always goes for price-quality on the node side of things.

If they can be 7 nm's for the Switch 2, when everything else is 3nm's... they will. In truth it all depends on Nvidia, but they'll have to face a lot of pressure if they choose a node that makes the chip pricier.
 
Last edited:

winjer

Gold Member
It could, steam deck is not even that powerful.

But this is Nintendo, if they had a 2 Tflop part they would downgrade it at least 30% making it functionally behave worse than Steam Deck who's a 1.6 Tflop part.


Also bare in mind that Steam Deck is "powerful" for a 720p screen, but the switch also plugs onto a TV, so at the same flop output they would already be at a disadvantage at 1080p but they'll have pressure to go higher. I guess that's where the tensor cores will come in.

Nintendo is historically never trying to be ahead on the node war, they always go for price-quality whatever that is.

If they can be 7 nm's for the Switch 2, when everything else is 3nm's... they will

Nintendo choose to use a Maxwell SoC, at a time when nVidia already had Pascall out.
If they use the same strategy, they´ll use Ada Lovelace in 2024-2025. Ada is made in N5, in 2022. But it could be made in N4 or N3 in 2024-2025. Neither of these will be leading edge at that time.
And it's more powerful than what is inside the Steam Deck.
 
Last edited:
Nintendo choose to use a Maxwell SoC, at a time when nVidia already had Pascall out.
If they use the same strategy, they´ll use Ada Lovelace in 2024-2025. Ada is made in N5, in 2022. But it could be made in N4 or N3 in 2024-2025. Neither of these will be leading edge at that time.
And it's more powerful than what is inside the Steam Deck.
Yeah, but powerful on paper is different from the spec Nintendo finds best for balancing battery, heat, and RMA rate.

As I said, unless heat and power consumption is not a concern at all, you can always expect a 30% power output decrease.
Now consider that N5 offers 40% power reduction over N7.
And N3 offers another 30% power reduction over N5.
Also, despite those node advantages, power consumption is still rising. As I'm sure you know, the upcoming RTX 4090 (5nm) is rumoured to be a 800W part, against RTX3090 (8nm) who was a 500W part. But it's not just on GPU's that this is happening we're seeing it everywhere with a few notable exceptions. Node evolution on paper should mean at this point that everyone's current gen desktop i3 to be a 25W part, but that isn't the case.

What I mean is, against Tegra X1, these designs we are discussing might still mean less battery life or a bulkier design to support it. And Nintendo faced with that might opt to halve the chip design. So the 4 Tflop part what was already touched upon could be requested to be manufactured with a 2 Tflops spec with half the processor logic instead. This design exists with the Jetson Orin NX, a 2 TFlop part based on the same Orin/Ampere design as the 4Tflop part, which I find more likely to be used if Nintendo was about to release something. For the Ada Lovelace generation we don't have a product line announced yet, but if there's a step down model of the chip, I'd put the Switch 2 there.

TLDR: whatever everyone thinks they'll get from Nintendo in 2024-25 might be the previous gen chip (as you said), with the node it was introduced with instead of a core shrink, 30% performance penalty against the tech sheet, and, if that is not enough due to costs, heat and power consumption it's possible that they'll opt for a chip a further half of the performance.

That can still can be more powerful than the current Steam Deck, but don't expect the spec to be impressive/flagship even against top range last gen. Also, we're focusing on GPU, but it's possible that CPU performance will have a bigger discrepancy against the current Steam Deck.
 
Last edited:

winjer

Gold Member
Yeah, but powerful on paper is different from the spec Nintendo finds best for balancing battery, heat, and RMA rate.

As I said, unless heat and power consumption is not a concern at all, you can always expect a 30% power output decrease.

Also, despite those node advantages, power consumption is still rising. As I'm sure you know, the upcoming RTX 4090 (5nm) is rumoured to be a 800W part, against RTX3090 (8nm) who was a 500W part. But it's not just on GPU's that this is happening we're seeing it everywhere with a few notable exceptions. Node evolution on paper should mean at this point that everyone's current gen desktop i3 to be a 25W part, but that isn't the case.

What I mean is, against Tegra X1, these designs we are discussing might still mean less battery life or a bulkier design to support it. And Nintendo faced with that might opt to halve the chip design. So the 4 Tflop part what was already touched upon could be requested to be manufactured with a 2 Tflops spec with half the processor logic instead. This design exists with the Jetson Orin NX, a 2 TFlop part based on the same Orin/Ampere design as the 4Tflop part, which I find more likely to be used if Nintendo was about to release something.


Power consumption is rising in the desktop parts, because nVidia is pushing clock speeds as much as they can.
When using a newer process node, a company can choose to have lower power usage at the same clocks. Or increase clocks at the same power usage. Or even just push clocks and power usage to the maximum.
An part using N5 or N3, at the same clock speed will have a much lower power usage than one using N7.

How can anyone think that an SoC made in 2024-2025, will have worse performance and power usage, than a part made in 2021.
Do you think that tech just stops in 2022 and nothing else happens?
 
This is, very clearly, a direct response to that other thread
I posted this in said thread as a simple comment, but I just was in the mood to make this thread about it.

Thinking that the Switch 2 could be as powerful or even more powerful than the Steam Deck is not at all unrealistic. People look at the Switch and see a system barely more powerful than the Wii U, then just assume that the jump to the new system will not reach Steam Deck levels of performance.

so 2 things about that.

1: The Steam Deck is not as powerful as many seem to think it is. The Steam Deck is less powerful in some respects than the PS4, mainly the GPU can not quite keep up with the PS4. And we are talking base PS4 here btw.
The PS4 has a 1.8TF GPU with consistent clock speeds that never change, while the Steam Deck has a 1.6TF GPU with variable clock speeds that can throttle under load.
It does have a decent CPU and enough Memory to perform very well, and it can reach 60fps in games where the PS4 could not, be it at a way lower resolution/graphical fidelity.

2: The Hardware inside the Switch is not a weak as some seem to think.
First of all it outperforms the Wii U and Xbox 360 quite a bit, either allowing for higher framerates or higher resolutions than on those consoles in similar or the same games.
All of this while also having a more modern feature set that can handle modern engines which would not even be able to run on 360 or PS3.​

But most importantly, we have to look at the actual hardware that Nvidia had available since at least back in 2019... namely the Tegra X1+, codename Mariko, which is the SoC used in all current Switch models since 2019

This Tegra X1+ could have easily been used to create a Nintendo Switch Pro, with a big boost in GPU and CPU performance, but Nintendo simply used it due to cost reduction and to have longer battery life.
How powerful is the Tegra X1+ when actually used to it's fullest? well, look at the stats below

Switch max clock speed:
-CPU: 1.02ghz
-GPU: 0.768ghz / 393 GFLOPs (max speed when docked)

Mariko Tegra X1+ max rated clock speed:
-CPU: 1.90ghz
-GPU: 1.267ghz / 649 GFLOPs

again, this is the same chip that is IN EVERY current Switch model and has been since 2019.


Let us assume Nvidia and Nintendo are designing the new console, let us also assume the WORST CASE SCENARIO possible, and that is that Nintendo uses an SoC that is merely double as performant as the Tegra X1+
Basically we are assuming that since 2019, the best Nvidia has to offer now, or in 1 to 2 years when the Switch 2 will release, is a chip that is 2x as powerful.
That would instantly mean 1.3 TF of GPU power and most likely way more performant CPU cores.

so in this worst case scenario we see that the Switch 2 would "only" reach Xbox One S (yes S not FAT) GPU performance
This is of course if Nintendo would use this Worst Case Scenario Chip at it's full speeds available, but still...

So worst case we will probably see a Switch 2 with the GPU power of an Xbox One S, with more modern features than the One S, and a better CPU than the One S.
This would still be below the Steam Deck tho of course.
but remember, this is the worst case possible imo... any wiggle-room above that and that 1.6TF Steam Deck will be very close indeed. If we get a 2.5x increase over the X1+ it would be above the Deck at 1.62TF, if the new tegra is 3x as powerful as the X1+ we are at 1.94TF and have at that point passed the PS4 in raw power.


IMO the real question will be if Nvidia and Nintendo will make use of Nvidia's newer technology with Tensor Cores and RT Cores. The inclusion of both would mean that even at a lower raster performance the Switch 2 could still keep up or even eclipse the Steam Deck in visual quality and performance.
DLSS and better RT hardware could mean games could run or look better on Switch 2 than on Steam Deck.

all of this is of course speculation, and I am in no way saying that any of this will happen, but what I am saying is that it is ridiculous to assume that there is no way that the Switch 2 would possibly reach Steam Deck levels of performance. Especially considering that Valve can not push their profit margins as low as Nintendo with it's higher production capacity and with the amount of first party games they sell every week. Mario Kart 8 hasn't left the top 20 charts in what feels like a decade...

Remember, all Nvidia needs is a 3x jump in GPU power and a reasonable jump in CPU performance to easily outpace the PS4 with their SoC for the Switch 2.
as a comparison the jump from Wii U to Switch is 176GF to 393GF, a 2.2x increase in GPU performance, and that was going from a home console to a tablet sized hybrid that doesn't even use the full power of its hardware
You are correct and ppl need to stop acting like the Steam Deck is a powerhouse. I am a happy steam deck owner but this is a console of compromises (they needed the price to be low). If I had the option of spending $200 more for 50 percent power power, I would. If Nvidia builds in some form of DLSS in switch 2/pro, it would smoke the built in, useless FSR that steam deck has
 

Thanati

Member
A newly released piece of hardware, in the future, is going to be more powerful than hardware available right now?

Really?

You don’t say.

Did you know the PS4 was going to be more powerful than the 360 as well?
 
How can anyone think that an SoC made in 2024-2025, will have worse performance and power usage, than a part made in 2021.
Do you think that tech just stops in 2022 and nothing else happens?
But I don't think it'll have worse performance per clock...

I just think that to conform to Nintendo list of priorities, it will never be as powerful as it could and that has to be taken into account.


Steam deck lasts from 1 hour and 20 minutes to 2 hours when running "AAA games" with a 44 Wh battery. I can tell you here and now that Nintendo would never use that spec on a shipping console. Heat and power consumption are too high for their usual goals and remember they don't want to spend extra cooling it either.

Switch Lite does 3-7 Hours with a 13,6 Wh battery, I don't think they'll go lower than that running time on a shipping spec unless their hands are tied and they also won't want to increase the battery capacity a lot. Normal Dockable Switch uses a 16 Wh battery, I don't think they would accept the added cost of anything over 20Wh.

In short, I believe one of the conditions for Nintendo is that it will have to do at least 3 hours with a battery half as big as the Steam Deck.


If a console matches the current Steam Deck with those limitations in place it'll be powerful per Watt and per cycle, but it won't be powerful on paper.
 
Last edited:

winjer

Gold Member
But I don't think it'll have worse performance per clock...

I just think that to conform to Nintendo list of priorities, it might not be as powerful as it could/should.


Steam deck lasts from 1 hour and 20 minutes to 2 hours when running "AAA games" with a 44 Wh battery. I can tell you here and now that Nintendo would never use that spec on a shipping console. Switch Lite does 3-7 Hours with a 13,6 Wh battery, I don't think they'll go lower than that and they also won't want to increase the battery capacity a lot. And normal Switch uses a 16 Wh battery.

In short, I believe one of the conditions for Nintendo is that it will have to do at least 3 hours with a battery only as big as 20Wh.


If a console matches the current Steam Deck with those limitations in place it'll be powerful per Watt and cycle.

But that would be right if using the same arch.
RDNA3, Ada Lovelace and Zen4, will have better IPC. And they can use faster memory. And more cache.
And with a process node like N5, N4 or N3, they can go with a wider chip, with similar area, while maintaining lower clocks and power usage.
 
The Steam Deck has the following specs (max):
1.638 TFLOPS
51.20 GTexel/s
12.80 GPixel/s
88 GB/s

Switch has the following specs (docked):
393 GFLOPS
12.29 GTexel/s
12.29 GPixel/s
25.6 GB/s

The console needs 4x raw power increase to beat Steam Deck.


PS4 has the following specs:
1.843 TFLOPS
57.60 GTexel/s
25.60 GPixel/s
176 GB/s
 
Last edited:

EverydayBeast

thinks Halo Infinite is a new graphical benchmark
Look at portable gaming the laptops the game boys the deck is steam’s big opportunity.
 
I dont know fuck all about hardware so not gonna contest you there OP. But it seems like your arguement is entirely based on "if nintendo use the newest and most cutting edge tech the result will be X".

But that would mean your whole argument immediately falls apart because that is not Nintendos model and hasnt been for 4 generations. Nintendo categorically have no interest in power, so not sure why this makes any sense or why suddenly Switch 2 is when they finally start to compete on power?

Also not sure if you take features into account when you consider "power", but the Steamdeck destroys the Switch in terms of features, versatility and accessibility. Steamdeck is a super open customizable system whereas Switch is the exact opposite; super locked down ecosystem and really poor features.

Edit: really good technical breakdown though OP, really appreciate it, learned a lot.
 
Last edited:

Caffeine

Member
its nintendo it wont be crazy powerful. count on this. I wouldnt be shocked if they moved away from switch with just a fully portable device. essentially leaving the home console market. something like 3050 gpu in power and a 900p screen. and release in like 2027
 
Last edited:

Gamer79

Predicts the worst decade for Sony starting 2022
So you think that a baseline cheap mobile chip set will automatically beat Steam Deck at a price that is family friendly for a mega portable machine?

I'm sorry but I just don't see it. I think they will purposely avoid using anything that could do that.
The tegra chip they use in the switch is a modified 2015 chip. I could see an improvement if they design a successor.
 

Astral Dog

Member
If i had to guess, in terms of GPU next generation Switch could be around Xbox ONE S level, to account for battery life /price/undocked mode.

But doesn't mean this will be the same as an old Xbox ONE, CPU should be quite more powerful by now and that alone makes a big difference. RAM used for games would also be in larger capacity even if were 8GB total, no idea how they are going to deal with streaming data/storage space 🤷🏻‍♂️ i assume they want faster loading times, and some form of DLSS.

1080p screen

This all speculation of course i have no idea what they are planning( could be worse, could be better) but that's around what i expect, a modern, more efficient Xbox ONE S in tablet
 
Last edited:

Xellos

Member
Switch 2 will probably have an edge, but I also think Nintendo will shoot for lower power consumption, a smaller & lighter chassis, and a smaller battery. Using technology not just for performance but for battery life/thermals/noise. Could wind up being something like Docked Switch 2 > Steam Deck >= Portable Switch 2, at least in terms of specs. Switch 2 will also have more software optimization than Steam Deck, so real world performance (portable) will probably be close.
 
I was under the impression Switch was considered to be a powerful mobile device when it came out in 2017.

A 2023 Switch would surely be comparable if not more powerful than Steam Deck.
 

64bitmodels

Reverse groomer.
Amazing all these years have gone by and they still don't understand, you increase a handhelds graphical power the battery life goes to shit. Unless they make it huge & stuff a massive battery in there you are going to only get a tiny battery life similar to valves deck.
I mean.... solid state batteries exist
 

CuNi

Member
Came in to say just this. Anything 'could' happen. I see the words 'let's assume' and 'speculation' in there too.

But so could be the other thread laid out.
"The Switch 2 could be less powerful".
No one ITT has any specific specs that are confirmed. We don't even know if it will be a Switch 2.
It all is speculation at this point.
 

RhyDin

Member
Modern Nintendo thinks it's okay to release games with such abysmal framerate as Link's Awakening remake. It's clear that there is no seal of quality anymore, so I dunno if it really matters to them if the next system is powerful. They seem content with releasing sub-par quality games, knowing full well that those games will be sell because it bears their intellectual property.
 

Kataploom

Gold Member
Modern Nintendo thinks it's okay to release games with such abysmal framerate as Link's Awakening remake. It's clear that there is no seal of quality anymore, so I dunno if it really matters to them if the next system is powerful. They seem content with releasing sub-par quality games, knowing full well that those games will be sell because it bears their intellectual property.
I don't remember a single sub par Nintendo game selling very well, most of those sports games sell less than one or two millions and the rest are mostly average or total masterpieces for most of their audience... In that case, also graphics are good enough to great, specially with the art styles they go for.

Also their main games are 60fps except for Zelda and Xenoblade and some minor stuff like modern Kirby and Yoshi games.

Idk about framerate in Link's awakening though...
 

Romulus

Member
Pretty sad that the Switch 2 is probably 2 years away and MIGHT be more powerful than a handheld that's already out, and according to the OP isn't even that powerful to begin with lol
 
Top Bottom