StarTopic Future Nintendo Hardware & Technology Speculation & Discussion |ST| (Read the staff posts before commenting!)

[Arrtyfact]

I’ll just say this:


You have freelance developers who do this for fun and made an emulator on PC for the wider audience to use if they want


And you have two multi-billion dollar companies, one who made the original soc of the device and has documentation for it, and the other who documents literally everything that happens on their platforms.



I’ll leave for you to decide if this backwards compatibility is impossible/not happening.

I agree that backwards compatibility on the Super Switch is definitely possible.

I highly doubt Nintendo/Nvidia will go the route of full software emulation though. It's just too complicated to have perfect compatibility. I understand that some people make switch emulators on their free time, but it is far from perfect. If you want to emulate games from recent systems, prepare yourself to often (depends on the game) face issues, either graphical glitches, audio distorted/not working, crashes, etc.

If you look at the Dolphin emulator (which emulates GameCube and Wii), the official game compatibility list shows that over 60% of tested games have minor graphical/audio glitches (they can still be fully completed, though), and Dolphin has been in development for almost 2 decades. Emulating a Switch is significantly more difficult than GameCube and Wii.

And I think 2 other major problems here are expectation and cost: Most emulators are free and people should understand that if something doesn't work as expected, it's okay. They maybe wasted a bit of time, but that's pretty much it. On the other hand, if someone buys an expensive Super Switch, tries to play a Base Switch game, and finds some glitches and/or the game crashes from time to time, even though that same game was working perfectly on their cheaper Base Switch, they will probably complain. People have big expectations when they buy expensive hardware, especially from a multi-billion dollar company.

We can just look at the backlash Nintendo received when some emulated Nintendo 64 games on Nintendo Switch Online had graphical issues.

Testing would also become a major issue. Since we want perfect emulation, every single possible scenario in every single Base Switch game ever released or releasing in the future would need to be tested and confirmed to be fully functional. That same testing would also need to be done every time the emulator receives an update to avoid regressions.

Okay, I'm probably exaggerating here, but still.

Those are the same reasons I would be very surprised if Sony actually released a PS3 emulator for PS Plus, instead of the current streaming solution.

 

[ILikeFeet]

I agree that backwards compatibility on the Super Switch is definitely possible.

I highly doubt Nintendo/Nvidia will go the route of full software emulation though. It's just too complicated to have perfect compatibility. I understand that some people make switch emulators on their free time, but it is far from perfect. If you want to emulate games from recent systems, prepare yourself to often (depends on the game) face issues, either graphical glitches, audio distorted/not working, crashes, etc.

If you look at the Dolphin emulator (which emulates GameCube and Wii), the official game compatibility list shows that over 60% of tested games have minor graphical/audio glitches (they can still be fully completed, though), and Dolphin has been in development for almost 2 decades. Emulating a Switch is significantly more difficult than GameCube and Wii.

And I think 2 other major problems here are expectation and cost: Most emulators are free and people should understand that if something doesn't work as expected, it's okay. They maybe wasted a bit of time, but that's pretty much it. On the other hand, if someone buys an expensive Super Switch, tries to play a Base Switch game, and finds some glitches and/or the game crashes from time to time, even though that same game was working perfectly on their cheaper Base Switch, they will probably complain. People have big expectations when they buy expensive hardware, especially from a multi-billion dollar company.

We can just look at the backlash Nintendo received when some emulated Nintendo 64 games on Nintendo Switch Online had graphical issues.

Testing would also become a major issue. Since we want perfect emulation, every single possible scenario in every single Base Switch game ever released or releasing in the future would need to be tested and confirmed to be fully functional. That same testing would also need to be done every time the emulator receives an update to avoid regressions.

Okay, I'm probably exaggerating here, but still.

Those are the same reasons I would be very surprised if Sony actually released a PS3 emulator for PS Plus, instead of the current streaming solution.

the biggest problem with your comparison is that comparing fan made emulators to what Nintendo/Nvidia has isn't a good one. Nintendo/Nvidia has none of the problems with Dolphin because they have the source to the games and to the APIs used to make it, as well as direct access to the hardware. when you have all that, you can actually make a better translation layer. think of it as talking to the writer of the book directly and being handed the notes the writer made vs translating a book after it was publish and that was a translation of the source, and no access to the writer's intentions.

also, this isn't equivalent to a PS3 emulator. it's more like Xbox One > Series or PS4 > PS5

 

[Arrtyfact]

the biggest problem with your comparison is that comparing fan made emulators to what Nintendo/Nvidia has isn't a good one. Nintendo/Nvidia has none of the problems with Dolphin because they have the source to the games and to the APIs used to make it, as well as direct access to the hardware. when you have all that, you can actually make a better translation layer. think of it as talking to the writer of the book directly and being handed the notes the writer made vs translating a book after it was publish and that was a translation of the source, and no access to the writer's intentions.

also, this isn't equivalent to a PS3 emulator. it's more like Xbox One > Series or PS4 > PS5

It would be much easier for Nintendo and Nvidia, I agree with that, but having perfect compatibility (and testing it) for all games (which I think is what people would expect) is extremely difficult and costly, both in time and money. It could potentially almost be less risky and expensive (I'm guessing here, I have no idea really) to have an overclocked X1 next to Drake, or a "Maxwell Mode" (as was mentioned before), to take care of running Base Switch games.

 

[Blue Monty]

It would be much easier for Nintendo and Nvidia, I agree with that, but having perfect compatibility (and testing it) for all games (which I think is what people would expect) is extremely difficult and costly, both in time and money. It could potentially almost be less risky and expensive (I'm guessing here, I have no idea really) to have an overclocked X1 next to Drake, or a "Maxwell Mode" (as was mentioned before), to take care of running Base Switch games.

You don’t need perfect compatibility 99% is enough as PS5/XSX have shown

 

[Arrtyfact]

You don’t need perfect compatibility 99% is enough as PS5/XSX have shown

Hmm, you're right. But I guess that 99% would still need to work perfectly somehow.

Also just in case, I'm not saying the next switch will not be backward compatible. I definitely think it will be, just that in my opinion a hardware solution would be easier and less risky

 

[kvetcha]

Hmm, you're right. But I guess that 99% would still need to work perfectly somehow.

Also just in case, I'm not saying the next switch will not be backward compatible. I definitely think it will be, just that in my opinion a hardware solution would be easier and less risky

where could you possibly fit a hardware solution?

 

[ILikeFeet]

Hmm, you're right. But I guess that 99% would still need to work perfectly somehow.

Also just in case, I'm not saying the next switch will not be backward compatible. I definitely think it will be, just that in my opinion a hardware solution would be easier and less risky

easier, but more costly and ultimately pointless. the 99% compatible case is still the best one. any errors can be fixed in future OS patches and/or updates to the games themselves

 

[12cheeper]

where could you possibly fit a hardware solution?

I'm guessing he's referring to getting the games to run natively as opposed to simple emulation.

 

[Deleted member 1324]

I am not saying 60fps BotW2 is impossible. I'm saying that physics simulations are highly framerate sensitive, and that I would place money against Nintendo taking their physics driven game and launching a 30fps and 60fps version simultaneously. Not a lot of money, mind you

I’ll buy you the game physical or digitally if it’s not 60fps on Drake.

 

[undefined]

BotW's physics engine is Havok, isn't it? Do we know if framerate changes causes issues with it? I saw some places mentioning that going over 60fps breaks it, seems like some games using it are capped to 60 frames on PCs

 

[Arrtyfact]

easier, but more costly and ultimately pointless. the 99% compatible case is still the best one. any errors can be fixed in future OS patches and/or updates to the games themselves

I believe having (and maintaining) 99% compatibility exclusively via software emulation and for a modern system like the Switch would be an overwhelming task, even for 2 multi-billion dollar companies. We're talking literally thousands of games here, not just a small curated selection. It would also be a major maintainability hassle. Running the games natively would be a much simpler, basically bulletproof solution.

Sure, you would not get all of the performance benefits of Drake when running Base Switch games, but at least they would run, while waiting for the Drake-enhanced patches (hopefully).

I'm guessing he's referring to getting the games to run natively as opposed to simple emulation.

Yep

 

[Dekuman]

It would be much easier for Nintendo and Nvidia, I agree with that, but having perfect compatibility (and testing it) for all games (which I think is what people would expect) is extremely difficult and costly, both in time and money. It could potentially almost be less risky and expensive (I'm guessing here, I have no idea really) to have an overclocked X1 next to Drake, or a "Maxwell Mode" (as was mentioned before), to take care of running Base Switch games.

probably game specific patches if emulation is not perfect, they just need to support 99.99% of the games to start and have a list of unsupported games

 

[TheCongressman]

BotW's physics engine is Havok, isn't it? Do we know if framerate changes causes issues with it? I saw some places mentioning that going over 60fps breaks it, seems like some games using it are capped to 60 frames on PCs

Yes, there were some issues with high fps over 60. Idk if it's the game, or emulation issues. But regardless of why, the fact is most of those issues have been fixed for a while. It's not like Nintendo can't patch the game to correct it on Drake.

 

[ILikeFeet]

I believe having (and maintaining) 99% compatibility exclusively via software emulation and for a modern system like the Switch would be an overwhelming task, even for 2 multi-billion dollar companies. We're talking literally thousands of games here, not just a small curated selection. It would also be a major maintainability hassle. Running the games natively would be a much simpler, basically bulletproof solution.

Sure, you would not get all of the performance benefits of Drake when running Base Switch games, but at least they would run, while waiting for the Drake-enhanced patches (hopefully).

or maybe it doesn't. as I said, Sony and MS doing doing the same thing and there are minimal (not zero though) issues.

and translation layers can offer performance benefits. unpatched games would run at their peak resolution and framerate. again, like Series and PS5

 

[Deleted member 1324]

BotW's physics engine is Havok, isn't it? Do we know if framerate changes causes issues with it? I saw some places mentioning that going over 60fps breaks it, seems like some games using it are capped to 60 frames on PCs

Yeah. There were different emulators using different versions of the game with different hardware configurations with some people using unlocked framerates over 60fps in the first year or two. There were certainly some bugs and issues but by early 2019 the game was exactly the same as the Wii U version but ran at higher resolutions and at 60fps on a powerful for the time gaming PC.

There are videos of BotW running on a modded launch Switch at 60fps in this very thread in the past couple of pages by lowering the resolution and removing the 30fps framerate cap. The game is absolutely fine.

Drake will run it at 60fps without breaking a sweat if the console is indeed the rumoured 6x CPU, 2x GPU, 2x RAM, 2x Bandwidth with DLSS. In fact I would bet it would run at 100+ fps. Of course BotW 2 will push better visuals but 60fps will still be well within reach imo even if it offer a ‘prefer framerate’ mode to get there.

 

[mariodk18]

How much less power do they need to only emulate the Switch GPU?

 

[Serif]

Left: mClassic off, 1080p on 1440p monitor
Right: mClassic on, 1440p on 1440p monitor

Since someone notified me in the Digital Foundry thread that my media embeds with the mClassic comparisons were broken, reuploaded. Left off, right on.

Yeah. There were different emulators using different versions of the game with different hardware configurations with some people using unlocked framerates over 60fps in the first year or two. There were certainly some bugs and issues but by early 2019 the game was exactly the same as the Wii U version but ran at higher resolutions and at 60fps on a powerful for the time gaming PC.

Tantalus also got Skyward Sword running at 60, with partial emulation and source code access, when emulator devs and modders couldn't do it. I imagine it must be more difficult to do this conversion on an older game engine, than for a game currently in development, that could theoretically be targeting multiple platforms already if the BotW2 team had access to Drake devkits in 2020.

 

[Pokemaniac]

How much less power do they need to only emulate the Switch GPU?

Depends a lot on the exact approach and how hardware assisted it is. If, for example they go with a translator with a cache approach (bonus points for using shared caches so online connected systems don't have to start with nothing), then the overall emulation overhead could become fairly negligible once the cache is sufficiently populated.

 

[KeijiKG]

Since someone notified me in the Digital Foundry thread that my media embeds with the mClassic comparisons were broken, reuploaded. Left off, right on.

Tantalus also got Skyward Sword running at 60, with partial emulation and source code access, when emulator devs and modders couldn't do it. I imagine it must be more difficult to do this conversion on an older game engine, than for a game currently in development, that could theoretically be targeting multiple platforms already if the BotW2 team had access to Drake devkits in 2020.

Personally speaking, taking a pic of your TV all looks the same to me regardless of resolution settings.
I'm more likely to notice a difference from an Elgato or other direct capture device to be honest.

 

[ShadowFox08]

I’ve been thinking and I think the wait is going to hurt Nintendo. I think there is so much hope and expectations around their next console no way can it live up to it. Only saving grace would be amazing must have software. It’s going to be interesting to see what happens. I personally don’t have any care about it now. It’s just been so much of a rollercoaster of rumors and misinformation. Software is more important than hardware at this point for me.

Software is the most important for sure, but I have no doubt Nintendo will deliver on that. But I can't lie when I say that I'm tired of playing 720p docked games... Even first party. If Drake isn't at least ps4 level without DLSS (as well as native 4k switch games), I'm gonna be dissapointed.

Yeah. There were different emulators using different versions of the game with different hardware configurations with some people using unlocked framerates over 60fps in the first year or two. There were certainly some bugs and issues but by early 2019 the game was exactly the same as the Wii U version but ran at higher resolutions and at 60fps on a powerful for the time gaming PC.

There are videos of BotW running on a modded launch Switch at 60fps in this very thread in the past couple of pages by lowering the resolution and removing the 30fps framerate cap. The game is absolutely fine.

Drake will run it at 60fps without breaking a sweat if the console is indeed the rumoured 6x CPU, 2x GPU, 2x RAM, 2x Bandwidth with DLSS. In fact I would bet it would run at 100+ fps. Of course BotW 2 will push better visuals but 60fps will still be well within reach imo even if it offer a ‘prefer framerate’ mode to get there.

6x CPU, 2x GPU, 2x Ram and 2X bandwidth.. Especially the latter 3. Where did you get this?

Man those specs are awful.

 

[Hermii]

Software is the most important for sure, but I have no doubt Nintendo will deliver on that. But I can't lie when I say that I'm tired of playing 720p docked games... Even first party. If Drake isn't at least ps4 level without DLSS (as well as native 4k switch games), I'm gonna be dissapointed.

6x CPU, 2x GPU, 2x Ram and 2X bandwidth.. Especially the latter 3. Where did you get this?

Man those specs are awful.

The only thing we actually have a solid source on, is the size and feature set of the gpu.

And the cpu architecture (a78) seems very likely.

 

[Thugstas]

The only thing we actually have a solid source on, is the size and feature set of the gpu.

And the cpu architecture (a78) seems very likely.

Re…. True …

 

[Deleted member 1324]

Software is the most important for sure, but I have no doubt Nintendo will deliver on that. But I can't lie when I say that I'm tired of playing 720p docked games... Even first party. If Drake isn't at least ps4 level without DLSS (as well as native 4k switch games), I'm gonna be dissapointed.

6x CPU, 2x GPU, 2x Ram and 2X bandwidth.. Especially the latter 3. Where did you get this?

Man those specs are awful.

It’s a mid range prediction based on what we know of the chip and the likely memory set up they will go with again mid range prediction.

If those specs are roughly what it turns out to be can I ask why the specs are awful? A 6x leap in CPU compute will get every Switch game to 60fps. The GPU and memory bandwidth will allow most games to run at 1080p native (which will all DLSS up to 4k). It will have much, much better image quality than a Steam Deck and better performing ports because it’s a console and not a pc.

What exactly are you expecting for a probable $400 hybrid console either late this year or in Q1 next year?

 

[ILikeFeet]

It’s a mid range prediction based on what we know of the chip and the likely memory set up they will go with again mid range prediction.

If those specs are roughly what it turns out to be can I ask why the specs are awful? A 6x leap in CPU compute will get every Switch game to 60fps. The GPU and memory bandwidth will allow most games to run at 1080p native (which will all DLSS up to 4k). It will have much, much better image quality than a Steam Deck and better performing ports because it’s a console and not a pc.

What exactly are you expecting for a probable $400 hybrid console either late this year or in Q1 next year?

Going from 256 cores to 1536 cores is a 6x increase itself. Even without linear scaling, you're still over 2x. And then there's the architectural changes on top of that. You're vastly underselling the gpu

 

[Serif]

Personally speaking, taking a pic of your TV all looks the same to me regardless of resolution settings.
I'm more likely to notice a difference from an Elgato or other direct capture device to be honest.

If I had a capture device I'd use it. I'm unsure if it could capture the scaling from 1440p to 4k that the CX is doing, though, as I assume I'd get a raw 1440p image. That being said, there are plenty of direct-feed shots of the mClassic already. I think the edge enhancement is still visible in those shots.

 

[Angel Whispers]

Wholeheartedly agree with @ShadowFox08 - He is correct in saying those specs are awful. If the bar was set in Hades, these proposed specs would be the equivalent of finding a way to limbo under it. That's how terrible they are, and I'm not even being hyperbolic there. There's pessimism, then there's trying to "Because Nintendo" one's way into a state of perpetual Doom Narrative, and I don't think enough people really understand what the consequences would be for Nintendo if such expectations were the reality. It's why I've been fiercely critical of some of the discussion points.

For all in this thread, but specifically those who stand on the side of doom and disappointment, please ask yourselves the following questions:

1) We know of the 1536 CUDA Core GPU from the Nvidia leak. If you haven't started there, get familiar, and make sure you understand what that number means - Even at the 2017 Switch's clocks, this would be a 1.179TF system in portable mode, and a 2.356TF home console. But there comes a point where you save nothing on battery life, cooling, or power consumption, and the 2017 Switch portable clocks would be well below that point. So, how do you expect this thing to be clocked, given that even at a point below where there would be no savings, the portable mode would still be at least SIX times better than the 2017 one BEFORE considering the evolution in architecture, improved lithography process, natural gains in efficiency and performance, RT & DLSS?

2) What kind of games do you expect Nintendo to get from their partners on such a successor, PS5/XS ports, or more Very LTTP XB1/PS4 ports, which are already possible on the 2017 Switch?

3) Considering the improvements in 3rdP support on the 2017 Switch, Is the latter in the second question really helpful, considering that 70% of Switch owners already own one of XB1/PS4, and more if you count a phone from 2018 at the earliest, or gaming laptops and PCs?

4) How successful do you expect this system to be with the circumstances of these gutterball expectations?

5) If you believe that a successor can be just as successful with those circumstances, how will they achieve that?

6) If you believe that this successor can be even more successful with those circumstances, what growth strategy do you think they'll adopt to achieve that, considering that they would no longer have "THE definitive portable experience" as a selling point, as phones and tablets (actual ARM-derivative devices) will leave them behind in no time at all?

7) If you still believe that Nintendo are that unambitious, then what do you believe Nvidia's aspirations are in this partnership?

I feel that one has to answer all of these questions, then look at the material from the horse's mouth (1536 CUDA Core GPU, RT & DLSS, A78 Class license making octa-core A78C the most probable CPU for the SoC, UE5 support confirmed, Overwatch 2 confirmed). When you've done that, and only then, please tell us why your expectations are "realistic".

 

[Ojoloco]

Wholeheartedly agree with @ShadowFox08 - He is correct in saying those specs are awful. If the bar was set in Hades, these proposed specs would be the equivalent of finding a way to limbo under it. That's how terrible they are, and I'm not even being hyperbolic there. There's pessimism, then there's trying to "Because Nintendo" one's way into a state of perpetual Doom Narrative, and I don't think enough people really understand what the consequences would be for Nintendo if such expectations were the reality. It's why I've been fiercely critical of some of the discussion points.

For all in this thread, but specifically those who stand on the side of doom and disappointment, please ask yourselves the following questions:

1) We know of the 1536 CUDA Core GPU from the Nvidia leak. If you haven't started there, get familiar, and make sure you understand what that number means - Even at the 2017 Switch's clocks, this would be a 1.179TF system in portable mode, and a 2.356TF home console. But there comes a point where you save nothing on battery life, cooling, or power consumption, and the 2017 Switch portable clocks would be well below that point. So, how do you expect this thing to be clocked, given that even at a point below where there would be no savings, the portable mode would still be at least SIX times better than the 2017 one BEFORE considering the evolution in architecture, improved lithorgraphy process, natural gains in efficiency and performance, RT & DLSS?

2) What kind of games do you expect Nintendo to get from their partners on such a successor, PS5/XS ports, or more Very LTTP XB1/PS4 ports, which are already possible on the 2017 Switch?

3) Considering the improvements in 3rdP support on the 2017 Switch, Is the latter in the second question really helpful, considering that 70% of Switch owners already own one of XB1/PS4, and more if you count a phone from 2018 at the earliest, or gaming laptops and PCs?

4) How successful do you expect this system to be with the circumstances of these gutterball expectations?

5) If you believe that a successor can be just as successful with those circumstances, how will they achieve that?

6) If you believe that this successor can be even more successful with those circumstances, what growth strategy do you think they'll adopt to achieve that, considering that they would no longer have "THE definitive portable experience" as a selling point, as phones and tablets (actual ARM-derivative devices) will leave them behind in no time at all?

7) If you still believe that Nintendo are that unambitious, then what do you believe Nvidia's aspirations are in this partnership?

I feel that one has to answer all of these questions, then look at the material from the horse's mouth (1536 CUDA Core GPU, RT & DLSS, A78 Class license making octa-core A78C the most probable CPU for the SoC, UE5 support confirmed, Overwatch 2 confirmed). When you've done that, and only then, please tell us why your expectations are "realistic".

What is the relationship between Overwatch 2 and Switch drake?

 

[Blue Monty]

The 70% Switch owners having a PS4/XboxOne is not only outdated but also was just in NA. In Japan maximum it is 33% ish and that is only if all PS4 owners have a Switch.

 

[Deleted member 1324]

Going from 256 cores to 1536 cores is a 6x increase itself. Even without linear scaling, you're still over 2x. And then there's the architectural changes on top of that. You're vastly underselling the gpu

Of course I’m merely allowing for the potential for Nintendo to be extremely conservative with clocks for heat and battery life issues. I’d guess it would be more like a 3x leap in GPU performance over the current Switch GPU but who knows it could be more.

 

[Skittzo]

Of course I’m merely allowing for the potential for Nintendo to be extremely conservative with clocks for heat and battery life issues. I’d guess it would be more like a 3x leap in GPU performance over the current Switch GPU but who knows it could be more.

They're not going to lower the clocks by more than 2x. There's a point where lowering clocks actually reduces power efficiency, and the Switch's current clocks are already pretty low on that curve.

The only information we really know for sure is that the GPU has 6x the cores as the base Switch. Adding in architecture efficiency gains we're looking at probably an effective 7x GPU boost minimum, assuming they don't raise the GPU clocks at all.

 

[mjayer]

Of course I’m merely allowing for the potential for Nintendo to be extremely conservative with clocks for heat and battery life issues. I’d guess it would be more like a 3x leap in GPU performance over the current Switch GPU but who knows it could be more.

Don’t you have a source? Was your guy kinda bummed about the juice in the machine or pumped? I think he would come off a little bummed if it was only a 2-3x increase to an already relatively under powered console?

 

[Arrtyfact]

or maybe it doesn't. as I said, Sony and MS doing doing the same thing and there are minimal (not zero though) issues.

and translation layers can offer performance benefits. unpatched games would run at their peak resolution and framerate. again, like Series and PS5

As far as I know, PS5/Xbox Series run PS4/Xbox One games natively, not via emulation. Xbox Series uses emulation for OG Xbox and 360, and even then it's far from supporting every OG and 360 titles, which in this case I believe is totally acceptable because it's multiple generations back, and not just the previous one. I'm not sure how people would feel if Nintendo announced a brand new, more powerful and expensive Switch and then said "You can play as much as 20% of the games you already own on this new system!"

As for translation layers, I was talking about backwards compatibility being handled 100% via software emulation. Like I said before:

I believe having (and maintaining) 99% compatibility exclusively via software emulation and for a modern system like the Switch would be an overwhelming task, even for 2 multi-billion dollar companies.

In the sense that they would simulate the entire Base Switch system via software, and not translate things like API calls.

I think making sure Drake has a Maxwell backwards compatibility mode would be a simpler solution, but of course I'm not Nintendo/Nvidia/a hardware designer so I'm just speculating here

 

[KMStwo]

Of course I’m merely allowing for the potential for Nintendo to be extremely conservative with clocks for heat and battery life issues. I’d guess it would be more like a 3x leap in GPU performance over the current Switch GPU but who knows it could be more.

3× leap is still something like ~1.2 tflops, way below the PS4 before DLSS you talked about a few weeks ago.

 

[Deleted member 1324]

3× leap is still something like ~1.2 tflops, way below the PS4 before DLSS you talked about a few weeks ago.

And the Switch was 1/4 PS4 and it still ran The Witcher 3 and the DOOM games.

The new Switch will not have 6x the GPU performance of the current model. That’s just setting people up for disappointment. The DLSS cores will do the heavy lifting and is actually a strong case for as low as a 2-3x GPU leap.

 

[Skittzo]

And the Switch was 1/4 PS4 and it still ran The Witcher 3 and the DOOM games.

The new Switch will not have 6x the GPU performance of the current model. That’s just setting people up for disappointment. The DLSS cores will do the heavy lifting and is actually a strong case for as low as a 2-3x GPU leap.

Considering the source of the information we have this feels a lot like a "because Nintendo" argument. 6x the GPU cores is one of the few bits of information we have that is extremely ironclad. That's much more certain than even it being a hybrid.

Switch's GPU currently runs at 307MHz at the lowest setting. You're arguing that they'll likely be lowering that to around 153MHz?

 

[Stinky Horse]

They're not going to lower the clocks by more than 2x. There's a point where lowering clocks actually reduces power efficiency, and the Switch's current clocks are already pretty low on that curve.

The only information we really know for sure is that the GPU has 6x the cores as the base Switch. Adding in architecture efficiency gains we're looking at probably an effective 7x GPU boost minimum, assuming they don't raise the GPU clocks at all.

These 2 cards have same number of fp32 cores and similar clocks:

UserBenchmark: Nvidia GTX 1080 vs RTX 3050

gpu.userbenchmark.com

 

[Raccoon]

Shuntaro Furukawa banged his fist on his desk. "Make the electricity slower!" he exclaimed. "It can't even be a little bit good, our fans demand it be shit!"

 

[Hermii]

These 2 cards have same number of fp32 cores and similar clocks:

UserBenchmark: Nvidia GTX 1080 vs RTX 3050

gpu.userbenchmark.com

Am I reading this right? The GTX1080 is kicking RTX3050s ass?

 

[Deleted member 1324]

Considering the source of the information we have this feels a lot like a "because Nintendo" argument. 6x the GPU cores is one of the few bits of information we have that is extremely ironclad. That's much more certain than even it being a hybrid.

Switch's GPU currently runs at 307MHz at the lowest setting. You're arguing that they'll likely be lowering that to around 153MHz?

I’m suggesting it could be as low as Switch GPU 2-4x on paper. I was told PS4 with PS4 Pro like IQ. PS4 Pro needed double it’s GPU cores for checkerboard rendering to get the games close to 4K iq. Drake won’t need to do it that way as it has specialised silicon for image reconstruction (DLSS). Going this way will save on heat and battery life rather than pure GPU brute force.

 

[Serif]

I'm expecting a big leap because Nintendo will probably wring out every last drop of Drake and its subsequent die shrinks into the mid 2030s. Spend a lot on a good chip and API development with future-proofing measures like DLSS now, and it'll last for a good while.

If you want to 'because Nintendo' this, then uh, I guess they'll stick with Drake for so long that it'll become underpowered again? lol

 

[Stinky Horse]

Am I reading this right? The GTX1080 is kicking RTX3050s ass?

Well there's no raytracing benchmarks on there
If you switch to the 3070 you can see game benchmarks around ~33% better than the 1080, at 230% of the fp32 cores.

 

[KMStwo]

And the Switch was 1/4 PS4 and it still ran The Witcher 3 and the DOOM games.

The new Switch will not have 6x the GPU performance of the current model. That’s just setting people up for disappointment. The DLSS cores will do the heavy lifting and is actually a strong case for as low as a 2-3x GPU leap.

How does the Switch running those games(with a far worse result) could explain that a 2 or 3 times increase could give results of a 1.8 Tflops console? I don't think "PS4 before DLSS" is 1.2 Tflops in practical usage even considering Nvidia architecture.

As I have read above, the GPU has to be clocked very low to only have a 2 to 3 times increase.

I just wanted to point out it is weird you talk about a 2-3 times GPU increase when you said you heard PS4 before DLSS which is still atleast a decent 4 times increase.

 

[Skittzo]

I’m suggesting it could be as low as Switch GPU 2-4x on paper. I was told PS4 with PS4 Pro like IQ. PS4 Pro needed double it’s GPU cores for checkerboard rendering to get the games close to 4K iq. Drake won’t need to do it that way as it has specialised silicon for image reconstruction (DLSS). Going this way will save on heat and battery life rather than pure GPU brute force.

There's this odd assumption I keep seeing that DLSS will somehow be a system level thing or most games will utilize it. When both of these things will not be true. Personally I'm pretty confident that no more than 5-10% of all games will use DLSS, it is not a general purpose solution for all games and all engines, and not all devs will be familiar with it.

Plus there is the fact that to even use DLSS in the first place the GPU needs to have a minimum performance level. 6x the previous GPU is not a very high bar for a GPU with 6x the cores and like 60% higher power efficiency to pass.

I honestly can't see any reasoning to believe as low as 2x beyond "because Nintendo", there is no technical justification presented.

 

[Deleted member 1324]

How does the Switch running those games(with a far worse result) could explain that a 2 or 3 times increase could give results of a 1.8 Tflops console? I don't think "PS4 before DLSS" is 1.2 Tflops in practical usage even considering Nvidia architecture.

As I have read above, the GPU has to be clocked very low to only have a 2 to 3 times increase.

I just wanted to point out it is weird you talk about a 2-3 times GPU increase when you said you heard PS4 before DLSS which is still atleast a decent 4 times increase.

Guess it depends on what you’re talking about when using x. There’s core count, clock speed, tflops, other bottlenecks (something is only as fast as it’s slowest part in many cases) and actual real world performance.

My point is Sony doubled their GPU silicon for Pro to make PS4 games run at 4k like iq. Drake doesn’t have to do that since it already has specialised silicon for that so they don’t need as big a GPU increase as some are expecting.

People expecting a near 3 tflop GPU (when docked) are going to be severely disappointed imo.

 

[Deleted member 1324]

There's this odd assumption I keep seeing that DLSS will somehow be a system level thing or most games will utilize it. When both of these things will not be true. Personally I'm pretty confident that no more than 5-10% of all games will use DLSS, it is not a general purpose solution for all games and all engines, and not all devs will be familiar with it.

Plus there is the fact that to even use DLSS in the first place the GPU needs to have a minimum performance level. 6x the previous GPU is not a very high bar for a GPU with 6x the cores and like 60% higher power efficiency to pass.

I honestly can't see any reasoning to believe as low as 2x beyond "because Nintendo", there is no technical justification presented.

DLSS is now a native part of the updated SDK / API. All Drake games going forward from day one will support it at a base level.

Older games supporting it will be up to individual publishers much like native “PS5 apps” separate from their PS4 version. Xenoblade 3 is already primed for it.

 

[ILikeFeet]

As far as I know, PS5/Xbox Series run PS4/Xbox One games natively, not via emulation. Xbox Series uses emulation for OG Xbox and 360, and even then it's far from supporting every OG and 360 titles, which in this case I believe is totally acceptable because it's multiple generations back, and not just the previous one. I'm not sure how people would feel if Nintendo announced a brand new, more powerful and expensive Switch and then said "You can play as much as 20% of the games you already own on this new system!"

As for translation layers, I was talking about backwards compatibility being handled 100% via software emulation. Like I said before:

In the sense that they would simulate the entire Base Switch system via software, and not translate things like API calls.

I think making sure Drake has a Maxwell backwards compatibility mode would be a simpler solution, but of course I'm not Nintendo/Nvidia/a hardware designer so I'm just speculating here

no one here expects software emulation because it's not expected to be needed. there is still some amount of compatibility to take advantage of. it's a more pessimistic view of BC but it doesn't seem realistic

Am I reading this right? The GTX1080 is kicking RTX3050s ass?

Userbenchmark is incredibly unrealiable. they should never be cited as a source

 

[Skittzo]

Guess it depends on what you’re talking about when using x. There’s core count, clock speed, tflops, other bottlenecks (something is only as fast as it’s slowest part in many cases) and actual real world performance.

My point is Sony doubled their GPU silicon for Pro to make PS4 games run at 4k like iq. Drake doesn’t have to do that since it already has specialised silicon for that so they don’t need as big a GPU increase as some are expecting.

People expecting a near 3 tflop GPU (when docked) are going to be severely disappointed imo.

Switch is currently a 393GFlops docked machine. 6x that is 2.3TF. Not sure where 3TF is coming from.

DLSS is now a native part of the updated SDK / API. All Drake games going forward from day one will support it at a base level.

Older games supporting it will be up to individual publishers much like native “PS5 apps”.

This is not how DLSS works. Your game needs to have a TAA solution that can gather pixel motion vector information, and if your game or engine doesn't have that you need to add it manually before you can use DLSS. DLSS is also nearly useless for 2D games as I understand it.


Here's a thought experiment. What do you think Mariko would've been capable of if they clocked it higher? We know for a fact homebrew folks can run docked mode clocks in handheld mode, and that alone is a 2.25x boost in GPU clocks. Maybe someone familiar with homebrew can tell us the battery life in that configuration?

If a 12nm Mariko from 2019 with 256 cores can do 2.25x the GPU performance of the base Switch without overheating or destroying battery life, do you really think a 8nm-5nm Drake from 2022-2023 with 1536 cores can only do 2x before it gets too hot or the battery life is too short?

 

[Skittzo]

Here's a better question: why would they design a 12SM chip if they could get the same results they want from a 2SM chip, which is far, far smaller and therefore cheaper per unit to manufacture?

 

[Mr Swine]

Here's a better question: why would they design a 12SM chip if they could get the same results they want from a 2SM chip, which is far, far smaller and therefore cheaper per unit to manufacture?

Because Nintendo is insane!!!

Homer Simpson crazy

 

[Skittzo]

As far as I can tell (with logic, not any inside info) this product will be used to bridge the gap to the PS5 gen (really the XSS) sorta like the base Switch did with the PS4 gen. DLSS will primarily be used by AAA publishers to get their game to 4k, or 1440p, or even in some cases maybe 1080p. However the game will need to run without DLSS first, which will require a fair bit of grunt for current gen games.