By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Nintendo Discussion - How will be Switch 2 performance wise?

 

Your expectations on performance...

Ridiculously below this g... 1 1.79%
 
Way below this gen: Some ... 17 30.36%
 
Slightly below this gen: ... 25 44.64%
 
On pair with this gen: AA... 13 23.21%
 
Total:56
sc94597 said:
bonzobanana said:

I'm thinking well under 1000 Cuda Cores maybe 700-800 I really don't think the Switch 2 will just have a stock T239 and doesn't even that have variable Cuda cores maybe related to yields. Why do you think its very unlikely it will have less cores? It's a small chip and its going to need extra space for Switch compatibility, encryption routines, uncompressing and compressing functionality and other logic Nintendo wants in that chip. Maybe development kits had full T239 chipsets but I suspect the final retail model will be very cut down in comparison. However I'll admit I'm only guessing based on Nintendo's past behaviour. Also that glimpse of Mariokart has been analysed and its not that impressive visually. Yes I know you can't base anything on that glimpse of the game but presumably Nintendo believes that is close to the product that will be shipped. I would expect much more from a 3 Teraflop games console. However I don't want to be Mr Negative in this thread and I still think it will be an amazing console but yes I'm pegging the performance right at the lowest end of expectations.

The T239 is a custom chip designed specifically for the Switch 2. I very much doubt they're going to have dev kits with 12SM (and design a chip around having that many cores) and reduce the real core-count by half. That would make it very hard for developers to optimize their games given that the real chips only have half the performance. It would also be a waste of R&D on Nvidia and Nintendo's part. 

Why do I think the 1,536 core-count (12SM) is accurate?

  • The T239 is a custom design for the Switch 2. It is not designed for anything else. 

  • It's been confirmed through multiple leaks, including the Nvidia leak, but multiple times afterwards. To the point that the likes of Digital Foundry/Eurogamer treat it as all but confirmed. 
  • The GPU clock rates make sense, given the core-count/SMs (they match peak efficiency for handheld mode and aren't too inefficient for docked mode.) 

  • A 1.7 TFLOP Handheld/3.1 TFLOP Docked chip isn't that powerful. It is roughly the mid-end of current gaming PC-handhelds and would be towards the lower-end when Switch 2 releases (we'll have Z2E chips in a hypothetical Rog Ally 2 by then.) That makes sense given that the original Switch fit that same performance slot, or even exceeded it when it released. 

I wouldn't be surprised if Mario Kart 9 (or whatever it is called) is cross-gen so I wouldn't base much off of it, but even if it weren't Mario Kart games haven't been top-of-the-line in terms of graphics for a system since the Gamecube or NDS. 

The original Switch had 5 different development kit specs I think and I believe before it was launched their were leaks of those boards which had more more memory and more performance. I think there were development kits with both 6GB and 8GB of memory. I'm not convinced at all that the Switch 2 will launch at the height of people's expectations. I think Nintendo are benefitting from people confusing development kit spec with final retail spec which will be much weaker. Why go for a cheap 8Nm fabrication process if you are going to push for such a high spec that will burn through battery life quickly? It doesn't really matter what the t239 is capable of at maximum performance, like Intel and other chip makers they simply reduce active functionality on the silicon to have extremely high yields. If they want half the active Cuda cores for battery life, cost and reduced heat that is a choice they can make. Development kits normally have a feature to test at retail spec even if the development kits are much stronger in performance. 

I remember when I had a wii u almost from launch and it was something like 5 years after the 360 and PS3 that happened. Many games were much weaker on wii u than ps3 or 360. Sonic transformed was something like 640p compared to higher resolutions on the other consoles. The sound wasn't as good and loading was longer. The frame rate wasn't brilliant either it suffered from some dips which the ps3 also suffered with but the 360 provided a much more consistent performance. There were many third party games that suffered like that compared to the other 2 much older consoles. The PS3 was offering amazing surround sound in 7.1, but often the wii u felt like it was purely 2 channel. Relying completely on optical disc meant much slower loading. People claimed it had a much better GPU that wasn't utilised but in the end it was confirmed it had a 176 Gflop gpu much weaker than those consoles in raw power and there was never any debate that the wii u had lower CPU resources. The only thing it really benefitted from was its 1GB of memory for the gaming side. Thanks to Nintendo's great optimisation of weak hardware though it still had a good version of Zelda Breath of the Wild. I still think Zelda Windwaker on the Gamecube is a visually stunning game and I think that GPU was only around 8 Gflops. Nintendo knows how people's expectations work and it is in their interests to allow them to think one of their systems is more powerful than it actually is. 



Around the Network
bonzobanana said:

The original Switch had 5 different development kit specs I think and I believe before it was launched their were leaks of those boards which had more more memory and more performance. I think there were development kits with both 6GB and 8GB of memory. I'm not convinced at all that the Switch 2 will launch at the height of people's expectations. I think Nintendo are benefitting from people confusing development kit spec with final retail spec which will be much weaker. Why go for a cheap 8Nm fabrication process if you are going to push for such a high spec that will burn through battery life quickly? It doesn't really matter what the t239 is capable of at maximum performance, like Intel and other chip makers they simply reduce active functionality on the silicon to have extremely high yields. If they want half the active Cuda cores for battery life, cost and reduced heat that is a choice they can make. Development kits normally have a feature to test at retail spec even if the development kits are much stronger in performance. 

I remember when I had a wii u almost from launch and it was something like 5 years after the 360 and PS3 that happened. Many games were much weaker on wii u than ps3 or 360. Sonic transformed was something like 640p compared to higher resolutions on the other consoles. The sound wasn't as good and loading was longer. The frame rate wasn't brilliant either it suffered from some dips which the ps3 also suffered with but the 360 provided a much more consistent performance. There were many third party games that suffered like that compared to the other 2 much older consoles. The PS3 was offering amazing surround sound in 7.1, but often the wii u felt like it was purely 2 channel. Relying completely on optical disc meant much slower loading. People claimed it had a much better GPU that wasn't utilised but in the end it was confirmed it had a 176 Gflop gpu much weaker than those consoles in raw power and there was never any debate that the wii u had lower CPU resources. The only thing it really benefitted from was its 1GB of memory for the gaming side. Thanks to Nintendo's great optimisation of weak hardware though it still had a good version of Zelda Breath of the Wild. I still think Zelda Windwaker on the Gamecube is a visually stunning game and I think that GPU was only around 8 Gflops. Nintendo knows how people's expectations work and it is in their interests to allow them to think one of their systems is more powerful than it actually is. 

The original Switch didn't have a fully custom chip. It inherited its chipset from the Nvidia Shield. Disabling some memory and reducing core-clocks is very different from halving the core-count when you could've designed to have a 6SM chip in the first place because it is totally new hardware. 

We don't know if the chip is 8nm. I am still a bit partial toward Samsung 5nm given what we know about handheld mode's clock rates. But from the specs we know the CPU is heavily underclocked to make up for a high GPU clock. Not impossible on 8nm, given that the docked mode is actually below expectations. 

And again the Wii U situation is very different from the Switch 2's. It inherited a lot of its CPU architecture from the Gamecube/Wii. Regardless though, there are some games on the Wii U (mostly exclusives) that were noticeable improvements over their PS360 counter-parts. Bayonetta 2 probably would have to have compromises to run on the PS360, as an example. 

You can't directly compare Gflops across architectures. A game is more than just floating-point compute. About a third of all compute are integer calculations (according to an Nvidia estimate a few years ago) and even if all compute in games involved floating-points, theoretical maximums can be cut down by bottlenecks. 

But yeah, the Wii U really doesn't have anything to do with the Switch 2's specs. They are very different systems with very different goals. 



sc94597 said:
bonzobanana said:

The original Switch had 5 different development kit specs I think and I believe before it was launched their were leaks of those boards which had more more memory and more performance. I think there were development kits with both 6GB and 8GB of memory. I'm not convinced at all that the Switch 2 will launch at the height of people's expectations. I think Nintendo are benefitting from people confusing development kit spec with final retail spec which will be much weaker. Why go for a cheap 8Nm fabrication process if you are going to push for such a high spec that will burn through battery life quickly? It doesn't really matter what the t239 is capable of at maximum performance, like Intel and other chip makers they simply reduce active functionality on the silicon to have extremely high yields. If they want half the active Cuda cores for battery life, cost and reduced heat that is a choice they can make. Development kits normally have a feature to test at retail spec even if the development kits are much stronger in performance. 

I remember when I had a wii u almost from launch and it was something like 5 years after the 360 and PS3 that happened. Many games were much weaker on wii u than ps3 or 360. Sonic transformed was something like 640p compared to higher resolutions on the other consoles. The sound wasn't as good and loading was longer. The frame rate wasn't brilliant either it suffered from some dips which the ps3 also suffered with but the 360 provided a much more consistent performance. There were many third party games that suffered like that compared to the other 2 much older consoles. The PS3 was offering amazing surround sound in 7.1, but often the wii u felt like it was purely 2 channel. Relying completely on optical disc meant much slower loading. People claimed it had a much better GPU that wasn't utilised but in the end it was confirmed it had a 176 Gflop gpu much weaker than those consoles in raw power and there was never any debate that the wii u had lower CPU resources. The only thing it really benefitted from was its 1GB of memory for the gaming side. Thanks to Nintendo's great optimisation of weak hardware though it still had a good version of Zelda Breath of the Wild. I still think Zelda Windwaker on the Gamecube is a visually stunning game and I think that GPU was only around 8 Gflops. Nintendo knows how people's expectations work and it is in their interests to allow them to think one of their systems is more powerful than it actually is. 

The original Switch didn't have a fully custom chip. It inherited its chipset from the Nvidia Shield. Disabling some memory and reducing core-clocks is very different from halving the core-count when you could've designed to have a 6SM chip in the first place because it is totally new hardware. 

We don't know if the chip is 8nm. I am still a bit partial toward Samsung 5nm given what we know about handheld mode's clock rates. But from the specs we know the CPU is heavily underclocked to make up for a high GPU clock. Not impossible on 8nm, given that the docked mode is actually below expectations. 

And again the Wii U situation is very different from the Switch 2's. It inherited a lot of its CPU architecture from the Gamecube/Wii. Regardless though, there are some games on the Wii U (mostly exclusives) that were noticeable improvements over their PS360 counter-parts. Bayonetta 2 probably would have to have compromises to run on the PS360, as an example. 

You can't directly compare Gflops across architectures. A game is more than just floating-point compute. About a third of all compute are integer calculations (according to an Nvidia estimate a few years ago) and even if all compute in games involved floating-points, theoretical maximums can be cut down by bottlenecks. 

But yeah, the Wii U really doesn't have anything to do with the Switch 2's specs. They are very different systems with very different goals. 

For some reason I thought it was near enough confirmed as 8Nm but not sure why I think that to be honest. I'd thought I'd seen it somewhere. I only quote Wii U because of how Nintendo goes for cheap designs with limited uncompetitive performance as a commercial decision. Gflops is a rough guide to graphics power I think its useful. I'm not a fan of Bayonetta 2 and while very well programmed I don't think it does anything the 360 and PS3 couldn't do. Bayonetta 2 isn't realistic graphics, its fairly basic texturing and its controlled progression in the game. I just don't see it as something other consoles couldn't do. I've seen a few games on ps3 which I thought couldn't be done on wii u or 360 because those games have made as much use of the cell processors as possible so there is a huge amount of game logic and stuff going on, on screen like the Resistance games. Where there were a huge amount of weather effects, explosions and more realistic texturing, full surround sound and lots of decent enemy AI. I've never seen anything like that on wii u or Switch to be honest. There is just such a huge amount of stuff going on, on screen from signs blowing in the wind, flags, dust, light, craft above. I was truly amazed when I first played the game and how much was going on. I don't think I ever had a moment with the wii u where I was wowed by the game engine being impressive. I loved the art style of breath of the wild but they were fairly simple cell shaded type graphics a bit like upgraded windwaker. Don't get me wrong a beautiful game but I never thought that game couldn't be achieved on 360 or PS3 and in fact in someways they probably could have improved it.

https://www.youtube.com/watch?v=rPsOuZGQYa8



Well, so far looking good



 

 

We reap what we sow

Yep, now that we've seen current generation games on the platform (Elden Ring, Hogwarts Legacy, etc) I think everything fit our expectations. Games looked slightly better than on my Rog Ally. 

Metroid Prime Beyond isn't too demanding of a title, but running it at 1080p 120fps definitely alleviates some of the CPU-bottleneck concerns, regardless. Also the 4k 60fps mode has to be with DLSS. Game looked crisp regardless. 

Edit: 

Fast Fusion is looking good at 4k 60fps! 

https://fast.shinen.com/fusion/

Last edited by sc94597 - 5 days ago

Around the Network
160rmf said:

Well, so far looking good

It looks like it will be roughly above an Xbox One X, better in some ways than Xbox Series S and worse in many others.

In portable, it might be around a PS4 Pro or better with a higher hz (120hz) supported as the max. 



Lifetime Sales Predictions 

Switch: 161 million (was 73 million, then 96 million, then 113 million, then 125 million, then 144 million, then 151 million, then 156 million)

PS5: 115 million (was 105 million) Xbox Series S/X: 40 million (was 60 million, then 67 million, then 57 million. then 48 million)

PS4: 120 mil (was 100 then 130 million, then 122 million) Xbox One: 51 mil (was 50 then 55 mil)

3DS: 75.5 mil (was 73, then 77 million)

"Let go your earthly tether, enter the void, empty and become wind." - Guru Laghima

Some Donkey Kong screenshots. 

These games are going to look crisp on my 4k OLED 16 inch portable monitor, which is how I'll likely play most of them! 



I was expecting 1440p 60 but 4K 60 is great.

The 120 was a surprise.



Yeah they hit all of the image-quality targets for me. 4k 60fps/1080p 120fps for Switch level titles, 1080p 30fps - 60fps for current-gen AAA titles (Elden Ring definitely is sub-60 and probably 1080p or 1080p-like resolution), in between those two for Switch 2 exclusives, VRR display (YES, YES, YES) - 40fps-50fps is very playable on the Rog Ally with its VRR display, 1080p target on the internal display, etc.

This definitely feels like a Rog Ally/Series S tier device in terms of graphics-settings, and falling in between the two in terms of image-quality (Series S > Switch 2 > Rog Ally.) 

That's exactly what we expected, and where Nintendo needs to be for third party support.

Edit: Docked 1440p support too! 

Last edited by sc94597 - 5 days ago

It's a pretty capable handheld device thankfully. It should age better than the Switch has and considering that the PS5 could still be getting some notable games by 2034-2035 it'll be capable enough for the clear majority of developers for about a decade.

Last edited by Norion - 5 days ago