• Hey everyone, staff have documented a list of banned content and subject matter that we feel are not consistent with site values, and don't make sense to host discussion of on Famiboards. This list (and the relevant reasoning per item) is viewable here.

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

Even if our guesses end up being pretty far away from what the succ will eventually be, I would recommend to adhere to some assumptions which are:

  • that the unit should have same form factor as the current Switch
  • same power draw as the current Switch
  • same screen resolution

I know, some will argue. But I think assuming those things will ground the discussion in reality more effectively than if we ignore them. A Switch would be too heavy to be part of a VR headset anyway and a stationary console would remove one feature that makes the Switch unique.
 
Even if our guesses end up being pretty far away from what the succ will eventually be, I would recommend to adhere to some assumptions which are:

  • that the unit should have same form factor as the current Switch
  • same power draw as the current Switch
  • same screen resolution

I know, some will argue. But I think assuming those things will ground the discussion in reality more effectively than if we ignore them. A Switch would be too heavy to be part of a VR headset anyway and a stationary console would remove one feature that makes the Switch unique.
Definitely agree with the first point. While I'm not in total agreement with the third one, I think it's a good idea to keep discussion invariant about the screen resolution when discussing power, since the big DLSS questions will probably only be relevant to producing the 4K output on the big screen.

Regarding the second point, I agree that it's a good idea for producing a basic idea of the specs, but we shouldn't discount the possibility. Switch has a pretty low power draw as a stationary device, and barely uses its docked configuration in any smart way. Discussing the options to make the dock situation more capable is definitely a worthwhile discussion, of course under the assumption that we can by no means assume anything will change regarding that. An increase in docked power output would be especially relevant as the target resolution in docked mode (4K, which obviously won't be reached by every game by any means) becomes very far removed from that of the handheld mode (720p or 1080p). So I would warn against leaving the discussion by the wayside, while staying realistic about the notion that it is very much a possibility (indeed a likelihood) that the dock will be cross-compatible and thus non-upgraded.
 
How I'm I doing that? Making the cooler larger doesn't has to increase the size of the device that much. 30W is not a crazy high amount to dissipate. Also, the dock itself could provide extra cooling.
if it wasn’t crazy high amount, do you think Nintendo would have opted to clock the GPU of the switch high enough to accommodate for such a power draw when docked?
 
if it wasn’t crazy high amount, do you think Nintendo would have opted to clock the GPU of the switch high enough to accommodate for such a power draw when docked?
Switch clocks are the maximum that the TX1 silicon can allow reliably. The exact same clocks you get from a rooted Shield Tv under load, and the Shield Tv has no battery. Orin doesn't has that issue as far as we know, and Drake looks to be a large chip, so it should have a higher power ceiling when cooling is adequate.
 
Last edited:
Switch clocks are the maximum that the TX1 silicon can allow reliably. The exact same clocks you get from a rooted Shield Tv under load, and the Shield Tv has no battery. Orin doesn't has that issue as far as we know, and Drake looks to be a large chip, so it should have a higher power ceiling with when cooling is adequate.
The shield TV clocks it’s CPU higher than what the switch does.

And the GPU is nearly twice what the switch does when docked.
 
Dakhil mentioned that the Switch is too heavy to be used as a VR headset so there is that.

Current Switch is around 300gr heavy, while PS VR is around 600gr heavy, so in what exactly sense is/would be too heavy?

I counting in case of VR, Nintendo would provide some kind of Switch head holder or something similar, not type of LABO VR system of using.
 
you know, with all the jokes about Monolith making a 360p game on drake, I think we're looking past someone who might push the system:

Creatures Inc

they've expressed interest in using ray tracing and Unreal Engine in recent interviews
 
Can't assume the same conditions as the 2017 Switch. Various parts have become lighter and more powerful and more efficient at the same time.
We have better cooling and lithography processes, too, while chip design has evolved, too. So, some elements can work closer to one another, then we have other intangibles such as unannounced tech, which could still be a possibility for the Switch's successor, even now. This also means that the form factor could be smaller, lighter, AND more powerful, or other form factors might be born. We also can't rule out the possibility of a change in dimensions, even if slight. A backwards compatible successor could surely play all existing Switch games in both modes with docked performance only? With all of that in mind, I wouldn't rule out the possibility of 1080p or even 1440p screens in this future device, should VR become a thing on it at any point. Both have been around on phones for quite a while.
 
you know, with all the jokes about Monolith making a 360p game on drake, I think we're looking past someone who might push the system:

Creatures Inc

they've expressed interest in using ray tracing and Unreal Engine in recent interviews
Detective Pikachu 2 gonna look like the movie.
 
0
I could see 1080p screen if Nintendo wants to start pushing VR furthere.

Man, I hope Nintendo doesn't think a tablet with a 1080P screen would be suitable for a VR setup in 2022 and beyond. That would be a dreadful experience. I love VR and am obsessed with the idea of Nintendo going in on VR in a real way - I've been day one for PSVR and both Quests - but I doubt I'd even bother with something like that. Aside from the subpar display, they'd need to provide some sort of solution for tracking on whatever accessory they create to hold the tablet which would likely mean that accessory wouldn't be cheap. Of course, they could also just ignore tracking and pretend that gyro would be acceptable, which would be... lol...

I really don't see any reason to bump up the tablet screen beyond 720p for now. If they want to do VR, and i think they absolutely should, it's a standalone headset SKU powered by Drake that can be used in TV mode as a regular Switch or bust. Labo VR 2 would be a waste of everyone's time.

Also, as far as the weight situation goes, I'm pretty sure the Switch weighs slightly more than half as much as the Quest 2. Weight shouldn't be a deal-breaker imo.
 
I think it's obvious the power draw will go up in docked mode for a few reasons.
First, when the Switch launched their last home console was extremely unsuccessful while the 3DS was keeping them alive. I'm sure this caused them to more heavily prioritize the handheld performance than they otherwise would. There is no longer a fear that they will have to pivot to a dedicated handheld, docked is an important feature for most users and thus the performance of it will be a more important selling point for them this time. Nintendo knows their customers are hungry for 4k, and a higher docked performance also helps enormously with getting key third party games. Even if these games don't end up running well on handheld, at least having them functional in docked mode is a big selling point for the console.

Second, the Orin chip is designed for higher power draw than the Maxwell was. There's a lot more value they can get out of increased power draw here, with key features like DLSS and raytracing. Letting Maxwell use more power probably wouldn't have benefitted them that much so the tradeoffs (devaluing handheld performance, potentially needing a bigger fan, risking damaging consoles over time) were not worth it. But in this case I think it's obvious that getting raytraced reflections and 4k via DLSS is a very high priority that they would be stupid not to use.

Third, and most importantly, there has to be a bigger gap in performance between handheld and docked this time, because of resolution differences. There is very little difference in the resolution of many games in handheld and docked on the switch (900p vs 720p). But with the next console, the handheld resolution probably wont increase that much. Even if they have a 1080p display, I would see a lot of games opting for 720p or 900p anyway, simply because it's such a negligible benefit on the small handheld screen that they would rather prioritize performance in other areas. But with docked, they are definitely going to want a better resolution than 900p. 2k is possible, though 4k is more likely, but either way that by itself necessitates a larger gap between the two. So increasing power draw in docked, does not mean power draw in handheld has to go up. And I see no reason to think they can't squeeze docked power draw up to 25 watts without melting the console.
 
You are right, games won’t fully utilize the hardware in the PS5, and the XSS exists.

But it is already evident that publishes aren’t targeting XSS level hardware, they aren’t scaling up from that.

They’re scaling down from a higher platform, and since PS4 was the market leader during its time, pubs will target for the PS5 as it (and PS4) is still the larger portion of many third party sales.

XBSX is second fiddle with the PS5 being the first thing they look to optimize for. And the Series S is third fiddle.

Anything ported to Drake would be by a separate team and possibly a later port. Better situation than what happened with the Switch and XB1/PS4 at the very least. Especially at the start.

Drake just greatly eases and “simplifies” the process of this happening. But it is still different enough that it being day and date wouldn’t be for all titles that are multiplat and do not have some timed exclusivity attached to it. Just not so far removed like switch games are which can be like a year or several months later of a port.
Unless it's made by teams that work for Activision, then PS5 will play third fiddle. But would they keep XSs first in mind?
 
I think it's obvious the power draw will go up in docked mode for a few reasons.
First, when the Switch launched their last home console was extremely unsuccessful while the 3DS was keeping them alive. I'm sure this caused them to more heavily prioritize the handheld performance than they otherwise would. There is no longer a fear that they will have to pivot to a dedicated handheld, docked is an important feature for most users and thus the performance of it will be a more important selling point for them this time. Nintendo knows their customers are hungry for 4k, and a higher docked performance also helps enormously with getting key third party games. Even if these games don't end up running well on handheld, at least having them functional in docked mode is a big selling point for the console.

Second, the Orin chip is designed for higher power draw than the Maxwell was. There's a lot more value they can get out of increased power draw here, with key features like DLSS and raytracing. Letting Maxwell use more power probably wouldn't have benefitted them that much so the tradeoffs (devaluing handheld performance, potentially needing a bigger fan, risking damaging consoles over time) were not worth it. But in this case I think it's obvious that getting raytraced reflections and 4k via DLSS is a very high priority that they would be stupid not to use.

Third, and most importantly, there has to be a bigger gap in performance between handheld and docked this time, because of resolution differences. There is very little difference in the resolution of many games in handheld and docked on the switch (900p vs 720p). But with the next console, the handheld resolution probably wont increase that much. Even if they have a 1080p display, I would see a lot of games opting for 720p or 900p anyway, simply because it's such a negligible benefit on the small handheld screen that they would rather prioritize performance in other areas. But with docked, they are definitely going to want a better resolution than 900p. 2k is possible, though 4k is more likely, but either way that by itself necessitates a larger gap between the two. So increasing power draw in docked, does not mean power draw in handheld has to go up. And I see no reason to think they can't squeeze docked power draw up to 25 watts without melting the console.
I believe the reason why we saw a lot of 720p to 900p in handheld and docked games was because the handheld mode was using a higher clock handheld profile than normal or we had a bottleneck like bandwidth.

Switch level fidelity games at 4k shouldn't be a problem. Hell if we do get 12 SMs, native 4k should be doable from some current 720p docked games.

PS4 level games at 4k with DLSS? I dunno.

Maybe the highest is xbone fidelity at 4k with DLSS.

edit: wanted to combine my last two posts. opps
 
Last edited:
I believe the reason why we saw a lot of 720p to 900p in handheld and docked games was because the handheld mode was using a higher clock handheld profile than normal or we had a bottleneck like bandwidth.

Switch level fidelity games at 4k shouldn't be a problem. Hell if we do get 12 SMs, native 4k should be doable from some current 720p docked games.

PS4 level games at 4k with DLSS? I dunno.

Maybe the highest is xbone fidelity at 4k with DLSS. But that's xbone x territory..

edit: wanted to combine my last two posts. opps
Again, may I reiterate
12SMs of Drake (Between Ampere and Lovleace uArch wise) likely would match up to the PS4 Pro when at 1Ghz which is well in excess of the OG PS4, and that is before DLSS.

So if the PS4 Pro can checkerboard to 4K a decent amount of the time and that's only 1440-1620p internally for a 4K output, then Drake can assuredly do the same with DLSS that is more flexible by taking 720p up to 4K (or 720p to 1440p with NIS to take it to 4K, but you get the idea)
 
Last edited:
Quoted by: LiC
1
I'm not sure if it would be possible or even necessary, but I hope the next Switch supports UHS-II microSD cards. Speed more than 100MB/s would be great, even though the CPU would also help a lot.
 
0
I believe the reason why we saw a lot of 720p to 900p in handheld and docked games was because the handheld mode was using a higher clock handheld profile than normal or we had a bottleneck like bandwidth.

Switch level fidelity games at 4k shouldn't be a problem. Hell if we do get 12 SMs, native 4k should be doable from some current 720p docked games.

PS4 level games at 4k with DLSS? I dunno.

Maybe the highest is xbone fidelity at 4k with DLSS. But that's xbone x territory..

edit: wanted to combine my last two posts. opps
On a base level it might not quite be PS4 level in 4k, however with raytracing and Nintendo's excellent visual design with their first party games I'm sure their games will still look like they're at that level even if they technically aren't.

EDIT: I should also say that I am taking a very cautious line with DLSS. I'm assuming that the games will run at 1440p and DLSS up to 4k. If the drake tensor cores are enough to go from 1080p or even 720p in good quality then it would definitely be better than the PS4 in 4k.
 
And Erista limits Nintendos clock profiles. Not Mariko.
The switch with its cpu clocked at like 20% higher wouldn’t cause so much of an issue. And that 20% would be enough for a decent amount to hold a 30FPS better or a 60FPS better.

Mind you, I don’t disagree with a higher TDP, I just don’t see why it needs to be that high. Steam deck is far larger of a device and is apparently very noisy with its fans.

Unless it's made by teams that work for Activision, then PS5 will play third fiddle. But would they keep XSs first in mind?
Considering that at the moment they are third party, the PS5 would still be the premiere platform. After the acquisition who knows how MS will play the game. Fortunate for Sony, it’s hardware isn’t far from what the Series consoles have, even if it’s a weird amalgamation with some custom elements that won’t be found elsewhere.


I completely disagree but we're all entitled to our opinions. 1080p I feel is the sweet spot. With 720p I can still easily see the jaggies.
More often than not, that’s not because of the resolution lol


It’s because it’s subnative

Take Genshin Impact for example, on my phone with TAA on, it looks better than the raw image when I select the “High” option for resolution. But if I go for Low with TAA on it doesn’t look to aggressively bad. Just softer.

Genshin Impact on High is like 734p or 784p and almost no one complains about this. If switch OLED is close to my phone in screen size, which is 6.7”, having a TAA would help a lot or simply being native would have a lot.

Hell, having both would help.
 
Last edited:
0
When your base console targets close to 4K, there 's a lot of scaling down that you can do. Much more so than when your base console targets (and often misses) 1080p. In that sense XSS and Drake should be in a reasonably good place.

On a base level it might not quite be PS4 level in 4k, however with raytracing and Nintendo's excellent visual design with their first party games I'm sure their games will still look like they're at that level even if they technically aren't.

EDIT: I should also say that I am taking a very cautious line with DLSS. I'm assuming that the games will run at 1440p and DLSS up to 4k. If the drake tensor cores are enough to go from 1080p or even 720p in good quality then it would definitely be better than the PS4 in 4k.
DLSS costs don't vary a lot based on input resolution, they mostly vary based on output resolution.
 
Again, may I reiterate
12SMs of Drake (Between Ampere and Lovleace uArch wise) likely would match up to the PS4 Pro when at 1Ghz which is well in excess of the OG PS4, and that is before DLSS.

So if the PS4 Pro can checkerboard to 4K a decent amount of the time and that's only 1440-1620p internally for a 4K output, then Drake can assuredly do the same with DLSS that is more flexible by taking 720p up to 4K (or 720p to 1440p with NIS to take it to 4K, but you get the idea)
The assertions that Drake is "between Ampere and Lovelace" reminds me a lot of when people were freaking out over the original Switch using Maxwell and insisting it had to either be Pascal or have so many changes it would be close to it. Turns out the Switch uses a stock Maxwell TX1 and yet, it was and is just fine. T239 is a customized Orin SoC, and it has an Ampere GPU. It's the furthest revised one, sure, and it seems like it may borrow one or two features from Lovelace, but saying it's not Ampere or in between the two is still a stretch. And again the new Switch will be just fine even if's not on the cutting edge.
 
I don't know how feasible it is but I'd be content with a 720p screen that can simply support 30, 40 and 60Hz. Save the 120Hz VRR for the docked performance.
 
0
Has anyone tried BotW and SMO VR modes? What's it like?
Labo VR is meant to be held up to your face in portable mode with the joy-cons attached, so there's no sense of how heavy the headset feels.

I have tried it with a Pro controller while lying down, letting the headset rest on my face. It didn't feel too heavy... it was the cardboard part that was more annoying.

There are some headsets on Amazon with headstraps, apparently with decent reviews. I see no complaints about weight. The Oculus quest 2 is ~500 grams, Switch tablet is ~300 grams + this headset and it'd be ~600 grams, so heavier but not unreasonably so, I'd have to try it for a fairer assessment.

I genuinely enjoyed BotW VR. Obviously there's the screendoor effect and it is noticeably low res, so I can only play around an hour before getting nauseous. But it was really cool, like the next evolution of Ocarina/Majora 3D's diorama effect. I'm currently doing a playthrough entirely in VR.
 
In general, I think the strongest argument for increasing the screen resolution is maintaining a reasonable power gap between docked and handheld mode. The current Switch has trouble reliably maintaining a 2.25x resolution gap between the modes, going all the way to 9x is unlikely to improve the situation.

That said, I can definitely pick out individual pixels on the Switch display. Do I sometimes hold the device closer to my face than is typical? Probably (I use it lying in bed a bunch), but that doesn't invalidate the use case.
 
The assertions that Drake is "between Ampere and Lovelace" reminds me a lot of when people were freaking out over the original Switch using Maxwell and insisting it had to either be Pascal or have so many changes it would be close to it. Turns out the Switch uses a stock Maxwell TX1 and yet, it was and is just fine. T239 is a customized Orin SoC, and it has an Ampere GPU. It's the furthest revised one, sure, and it seems like it may borrow one or two features from Lovelace, but saying it's not Ampere or in between the two is still a stretch. And again the new Switch will be just fine even if's not on the cutting edge.
??????????
You kinda missed the point?

The main point here is that it seems Lovelace is a very minor tweaked version of Ampere (Cache and RT Cores per GPC being the only major differences outside of node seemingly).

And Drake inherits one of those at least (cache), and the Jury is out on if the indicated RT core improvement for Lovelace and Orin (Lovelace itself seems to halve the RT core count over Ampere like Orin, with Drake being still 1RT/SM) has also passed onto Drake making it "Lovelace on a different node with double the RT cores"

Or if it's
"Ampere with Lovelace's Cache"

The main point is that Lovelace and Ampere are very similar and that likely is intentional, but the main takeaway is that Drake does have elements of Lovelace in its structure which is the main reason why I say it's "Between" Ampere and Lovelace which I say is a fair assessment to take.
 
Last edited:
The current Switch has trouble reliably maintaining a 2.25x resolution gap between the modes, going all the way to 9x is unlikely to improve the situation.
what do you mean by this? The impression that I've gathered across the several months of speculation is that the switch 2 would still maintain that 2.25x resolution gap internally, and have DLSS to supersample the image from 1080p to 1440p-2160p, not to actually render 720p in handheld and then render at 2160p before DLSS.

The assertions that Drake is "between Ampere and Lovelace" reminds me a lot of when people were freaking out over the original Switch using Maxwell and insisting it had to either be Pascal or have so many changes it would be close to it. Turns out the Switch uses a stock Maxwell TX1 and yet, it was and is just fine. T239 is a customized Orin SoC, and it has an Ampere GPU. It's the furthest revised one, sure, and it seems like it may borrow one or two features from Lovelace, but saying it's not Ampere or in between the two is still a stretch. And again the new Switch will be just fine even if's not on the cutting edge.
It's not exactly unheard of for a console to borrow certain features not available on the consumer graphics cards at the time but is present in a future architecture. PS4 Pro is one of the most famous examples of it, switch being another with the FP16, and I believe that the 360 was very competitive in its day and had features that became present later in future uArch's, or a future GPU uArch implements a feature that is similar to an older console equivalent (like the PS2 "Mesh Shaders"), or even the PS5 with its weird RDNA1+2 amalgamation + some other features that are as similar to the DirectX12 compliant feature-set as possible. And it isn't like Ampere and Lovelace are, from what I can tell based on information right now, that drastically different. Now, calling it Lovelace I do agree is wrong, but an in between of Lovelace and Ampere or calling it "Ampere+" could be more fitting for this Drake thing.
 
Me waiting for Drake leaks

52a2c.jpg
 
??????????
You kinda missed the point?

The main point here is that it seems Lovelace is a very minor tweaked version of Ampere (Cache and RT Cores per GPC being the only major differences outside of node seemingly).

And Drake inherits one of those at least (cache), and the Jury is out on if the indicated RT core improvement for Lovelace and Orin (Lovelace itself seems to halve the RT core count over Ampere like Orin, with Drake being still 1RT/SM) has also passed onto Drake making it "Lovelace on a different node with double the RT cores"

Or if it's
"Ampere with Lovelace's Cache"

The main point is that Lovelace and Ampere are very similar and that likely is intentional, but the main takeaway is that Drake does have elements of Lovelace in its structure which is the main reason why I say it's "Between" Ampere and Lovelace which I say is a fair assessment to take.
How does Drake inherit cache from Lovelace? Drake has either the same amount of cache as Orin or less, and in the same configuration. This is what I mean, the reaching to connect Drake to later architecture -- however similar to Ampere Lovelace may be -- when in reality it's exactly what you'd expect from a customized version of Orin. Largely, components are being cut down or removed from Orin to Drake. Even some improvements Orin had over desktop Ampere aren't present anymore. Maybe a couple new features are being added (nothing concrete on that front yet), but nothing all that important, and nothing that changes what the architecture really is.

It's not exactly unheard of for a console to borrow certain features not available on the consumer graphics cards at the time but is present in a future architecture. PS4 Pro is one of the most famous examples of it, switch being another with the FP16, and I believe that the 360 was very competitive in its day and had features that became present later in future uArch's, or a future GPU uArch implements a feature that is similar to an older console equivalent (like the PS2 "Mesh Shaders"), or even the PS5 with its weird RDNA1+2 amalgamation + some other features that are as similar to the DirectX12 compliant feature-set as possible. And it isn't like Ampere and Lovelace are, from what I can tell based on information right now, that drastically different. Now, calling it Lovelace I do agree is wrong, but an in between of Lovelace and Ampere or calling it "Ampere+" could be more fitting for this Drake thing.
Maybe I'm just splitting hairs, but being the most recent revision to Ampere, and therefore the closest to Lovelace, doesn't make it in between or Ampere+ to me. I think you need something that sets it apart from Ampere, or actually makes it like Lovelace, to say that.
 
How does Drake inherit cache from Lovelace? Drake has either the same amount of cache as Orin or less, and in the same configuration. This is what I mean, the reaching to connect Drake to later architecture -- however similar to Ampere Lovelace may be -- when in reality it's exactly what you'd expect from a customized version of Orin. Largely, components are being cut down or removed from Orin to Drake. Even some improvements Orin had over desktop Ampere aren't present anymore. Maybe a couple new features are being added (nothing concrete on that front yet), but nothing all that important, and nothing that changes what the architecture really is.


Maybe I'm just splitting hairs, but being the most recent revision to Ampere, and therefore the closest to Lovelace, doesn't make it in between or Ampere+ to me. I think you need something that sets it apart from Ampere, or actually makes it like Lovelace, to say that.
But that's the thing, what makes Lovelace Lovelace outside of "5nm + more Cache + better but half the RT Cores"
 
Labo VR is meant to be held up to your face in portable mode with the joy-cons attached, so there's no sense of how heavy the headset feels.

I have tried it with a Pro controller while lying down, letting the headset rest on my face. It didn't feel too heavy... it was the cardboard part that was more annoying.

There are some headsets on Amazon with headstraps, apparently with decent reviews. I see no complaints about weight. The Oculus quest 2 is ~500 grams, Switch tablet is ~300 grams + this headset and it'd be ~600 grams, so heavier but not unreasonably so, I'd have to try it for a fairer assessment.

I genuinely enjoyed BotW VR. Obviously there's the screendoor effect and it is noticeably low res, so I can only play around an hour before getting nauseous. But it was really cool, like the next evolution of Ocarina/Majora 3D's diorama effect. I'm currently doing a playthrough entirely in VR.
Nice. I’ve been tempted to try it if I ever see it on sale.
 
0
I genuinely enjoyed BotW VR. Obviously there's the screendoor effect and it is noticeably low res, so I can only play around an hour before getting nauseous. But it was really cool, like the next evolution of Ocarina/Majora 3D's diorama effect. I'm currently doing a playthrough entirely in VR.

This is where I really think Nintendo's contribution to VR could separate itself from the competition. If they've got a VR device with this Drake chip in it on market while they're still targeting OG Switch for development of their flat games, they should have plenty of headroom to do that 3DS stereoscopic diorama effect thing with basically any game they put out. Sure, it's not the kind of experience most think of when they think VR, but switching between playing a Zelda, Mario, Splatoon, etc on TV then jumping into the headset for an extra level of immersion, even if the core gameplay doesn't change, would be a nice value add for the platform.

I caught myself thinking about this a few times this weekend while playing Triangle Strategy. Having the gameplay area suspended in the void like a little diorama I can grab and rotate manually while UI elements are offloaded from the gameplay view into VR space would be so cool.
 
I was watching this video, and it looked like the 12.74 TF RTX3060 held almost firm against the Radeon RX 6800, which has 16.17 TF. Is the AMD flop-for-flop comparison really so skewed that AMD is still behind after NVIDIA doubled its CUDA count per SM?

If so, then that comparison with the Radeon 680M could be a good point of reference for what Drake could do in docked mode, perhaps somewhere in between that and the GTX 1650?
Just a thought off the top of my head, but since RDNA2 doesn't have dedicated hardware like the RT cores, I assume that whenever there's stuff that RT cores would handle on Nvidia cards, instead get handled by shaders on the counterpart RDNA2 cards, right? Ergo, it'd introduce variance to how much of the on paper FLOPS actually get used for everything else being rendered?
 
I really couldn’t care less about VR, but I’m biased since I have next to no vision on my left eye.

So selfishly I’m hoping Nintendo won’t even consider VR when designing the Drake Switch.
 
Just a thought off the top of my head, but since RDNA2 doesn't have dedicated hardware like the RT cores, I assume that whenever there's stuff that RT cores would handle on Nvidia cards, instead get handled by shaders on the counterpart RDNA2 cards, right? Ergo, it'd introduce variance to how much of the on paper FLOPS actually get used for everything else being rendered?
no, RDNA2 does have hardware RT acceleration. it just does less than Nvidia's RT cores, which tests intersections of rays and triangles, and bvh traversal
 
Maybe I'm just splitting hairs, but being the most recent revision to Ampere, and therefore the closest to Lovelace, doesn't make it in between or Ampere+ to me. I think you need something that sets it apart from Ampere, or actually makes it like Lovelace, to say that.
I mean, what separates Maxwell from Pascal that needed it to be such a different classification. Or Volta and Turing that was enough to differentiate it from each other?

Or what about Maxwell and Maxwell 2.0?
Just a thought off the top of my head, but since RDNA2 doesn't have dedicated hardware like the RT cores, I assume that whenever there's stuff that RT cores would handle on Nvidia cards, instead get handled by shaders on the counterpart RDNA2 cards, right? Ergo, it'd introduce variance to how much of the on paper FLOPS actually get used for everything else being rendered?
RDNA2 does have dedicated RT hardware called “Ray Accelerators”, but they are based on the TMUs being modified for this purpose.
 
Last edited:
what do you mean by this? The impression that I've gathered across the several months of speculation is that the switch 2 would still maintain that 2.25x resolution gap internally, and have DLSS to supersample the image from 1080p to 1440p-2160p, not to actually render 720p in handheld and then render at 2160p before DLSS.
Not everything can or will go through DLSS, and it wouldn't necessarily be limited to docked mode.
 
0
Please read this staff post before posting.

Furthermore, according to this follow-up post, all off-topic chat will be moderated.
Last edited:


Back
Top Bottom