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

[Kise Ryota]

Not so long ago I believed Drake was too big for sec 8nm (for a product like the Switch (form factor and battery life)). This has not changed for me.

I just can't believe in a GPU clock of 660MHz handheld or 1.1GHz+ docked using that node. Actually I don't even believe those clocks are related to a nintendo product, but if they are, it would have to be in another node IMO.

I'll stick with that Orin power estimation tool; I think it's safer to expect the lowest clocks possible while still having a not so great battery life (if using sec 8nm). It'll still be better than what I expected from switch 2 (if it's coming this year). But I do believe they could be leaving something like 40% or 50% more performance on the table by going with this node (while still having a just OK battery life). The idea was that to have a SoC of this size with very constrained clocks (because of the node), it would be better to just make a smaller SoC with higher clocks. I still believe on that.

PS: by the orin tool, just the GPU would have a power consumption of 19.9W at 1032.75MHz.

 

[lucapfstr]

I wrote about this before. Mochizuki's Bloomberg report was the only source including that "middle of its lifecycle" quote in Feb. 2022. Any other articles saying the same were simply regurgitating his story.

That Furukawa quote was not in Nintendo's Feb. 3rd official transcript. No other reputable media—not even Bloomberg Japan's report written by another journalist—contained that sentence. AFAIK, the last time Furukawa declared the Switch being "at the mid-point of its lifecycle" was on Nov. 5, 2021.

Unless Mochizuki got an exclusive interview, which not even the Bloomberg Japan reporter attended, it doesn't seem an accurate recollection. This wasn't the only time that I questioned certain details of his articles/tweets in this thread, and I'd encourage people to compare his stories with other major media reports (or at the very least, Bloomberg Japan).

Yeah, iirc in 2020 Furukawa said it’s barely halfway through and ready to break conventions of previous consoles. Or maybe he said it later. Clearly he was alluding to the oled as Nintendo was essentially bringing their handheld revision strategy to their new console.

Then in 2021 it was in the mid-point. Extrapolating this, the new switch should come in 2025 by the latest.

So the question becomes this: do we take the quote literally - we see another revision this year and a successor in 2025 - or do we assume a successor this year with around 2 more years of switch support?

 

[LiC]

I wish I still shared the hope some of you do. Regarding TotK OLED, if Drake is a simultaneous launch with that game, I find it hard to believe they’d sell a SE OLED for the game they’re also using to promote and drive sales of their new hardware.

If I decouple Drake from TotK, a launch in 2023 seems more plausible, I admit. Although I’m not sure why they wouldn’t have done a TotK launch if it’s coming months later.

But more importantly for a 2023 CY launch, I can’t square it with Nate saying his H1 2023 hardware was shelved. I know there’s a lot of question marks we have about what Nate knows that will be cleared up in his podcast, but even with that one data point we do have from him - H1 2023 device shelved - in the event Drake still releases this year, Nintendo shelved one device for H1 this year and replaced it with another months later?

I don’t know. If there’s some context or faulty logic here that I’m missing or mixing up from the roller coaster that has been the past week, happy to be corrected.

I do still have some hope until we hear Nate’s podcast. I want this device in 2023 still, bad!

Well, for me the problem is I don't believe the "H1 2023 device shelved" data point from any angle. It's hard to even talk about what might have been cancelled when the information is so contradictory. Even if something was cancelled, I just don't think there's any way that everything Nate is saying about it can possibly refer to just one device -- simultaneously a mid-gen refresh small and early enough that Nintendo could cancel it without seriously pissing off dozens of partners and publishers and becoming a major mainstream news event -- and a system with a development time of 3+ years launching in 2023 with enough kits in the hands of third parties for Nate to even know about it, featuring DLSS and ray tracing. How can those possibly be the same device? How was any kind of mid-gen refresh ever going to launch in 2023, six years after the original? Why would a mid-gen refresh break (native) backwards compatibility with a new GPU architecture? And it goes on and on.

I wrote about this before. Mochizuki's Bloomberg report was the only source including that "middle of its lifecycle" quote in Feb. 2022. Any other articles saying the same were simply regurgitating his story.

That Furukawa quote was not in Nintendo's Feb. 3rd official transcript. No other reputable media—not even Bloomberg Japan's report written by another journalist—contained that sentence. AFAIK, the last time Furukawa declared the Switch being "at the mid-point of its lifecycle" was on Nov. 5, 2021.

Unless Mochizuki got an exclusive interview, which not even the Bloomberg Japan reporter attended, it doesn't seem an accurate recollection. This wasn't the only time that I questioned certain details of his articles/tweets in this thread, and I'd encourage people to compare his stories with other major media reports (or at the very least, Bloomberg Japan).

The placement of the quote may be suspect but those are articles were only three months apart. If the Switch was in the middle of its lifecycle in November 2021 then it wasn't meaningfully past that characterization in February 2022.

 

[bloopland33]

Everything you say is true, but the reality is we don’t know the volume of special edition OLEDS produced. For all we know, the special editions are a great way to sell off more Marikos, clearly the target audience is collectors and particular franchise-lovers

The chances of new hardware with TOTK is slim but not impossible, and the chances of some new hardware this year at all (not necessarily drake) I’d say are pretty high because new hardware HAS released every other year since 2017.

Your points about SEs aren’t bad at all.

For the record, I am on holiday 2023. As the target.

Hey, that'd be great. I’ll take holiday at this point! So how do you personally read or resolve Nate’s comments about H1 hardware being shelved? If he’s correct, it’s odd that they had one thing planned for H1 but shelved that device in favor of something different for H2 of the same year. And by odd I mean that sounds so bizarre, doesn’t it?

 

[ITSMILNER]

Switch 2, holiday 2023 with Mario Odyssey 2 is the only other way I see it coming this year.

 

[ReddDreadtheLead]

Except engines like Unity already had literal Switch exporters built in...? This isn't the 2000s, ARM architectures are widely supported and will be well documented to Drake developers like the TX1 already was. There's a reason NVN2 exists, and why Switch already had rather anormal 3rd party support as it is, some of the concerns mentioned here make me worry about the thread, lol.

where did you get ARM from my comment? I’m talking about the GPU. Ampere is different from RDNA1-2.

 

[Foltzie]

I am unshakeable in this belief but people (I talk with anyways) never seem to take it seriously because they only see support for Switch dying the moment Switch 2 is announced.

Apple didn't kill the first iPhone the moment the iPhone 2 launched

Sure and its true the cross-gen lifecycle between PS4/5 and XB1/XS is longer than before, it still seems somewhat harder to believe that there will be an extended cross-gen cycle for Switch while Drake launches. For example, I'm curious to see how gracefully UE5 titles can scale down to Switch.

Still, unless someone has a ballpark BOM that shows that Drake Switch could slot profitably into a $299/$199 price point, I expect Nintendo to maintain the Mariko based Switch line for as long as it sells.

I also agree that for Indie titles and Nintendo titles it is much more reasonable to expect support, especially assuming Nintendo offers a graceful way to package and ship titles targeting NVN and NVN2 profiles.

 

[Lancelot]

where did you get ARM from my comment? I’m talking about the GPU. Ampere is different from RDNA1-2.

Nobody is coding bare metal for this, dude... Software APIs like NVN, Vulkan, DirectX, etc exist for a reason, so you can code for the thing as long as it's powerful enough. Drake is, even if they'll have to carry DLSS as a crutch sometime.

 

[bmfrosty]

i also think by ten year (or longer) lifespan that includes any system launched under the Switch moniker. as well as being indicative they intend to support the base model with software for the whole duration.

This is what I've been thinking. I think Nintendo (and most third parties) will try and support Switch and Switch NG for as long as they are, but I think we've hit the end of impossible ports. If a game is only viable on Switch NG, then if it shows up Switch OG at all, it will be a cloud version.

 

[bloopland33]

Well, for me the problem is I don't believe the "H1 2023 device shelved" data point from any angle. It's hard to even talk about what might have been cancelled when the information is so contradictory. Even if something was cancelled, I just don't think there's any way that everything Nate is saying about it can possibly refer to just one device -- simultaneously a mid-gen refresh small and early enough that Nintendo could cancel it without seriously pissing off dozens of partners and publishers and becoming a major mainstream news event -- and a system with a development time of 3+ years launching in 2023 with enough kits in the hands of third parties for Nate to even know about it, featuring DLSS and ray tracing. How can those possibly be the same device? How was any kind of mid-gen refresh ever going to launch in 2023, six years after the original? Why would a mid-gen refresh break (native) backwards compatibility with a new GPU architecture? And it goes on and on.


The placement of the quote may be suspect but those are articles were only three months apart. If the Switch was in the middle of its lifecycle in November 2021 then it wasn't meaningfully past that characterization in February 2022.

Interesting thoughts to consider, thanks. I really hope you’re right! Obviously Nate’s view on his own info isn’t as limited as our view, but I wonder if he grappled with the same confusion at any point. Suppose we’ll know soon.

 

[C_Ali]

I am unshakeable in this belief but people (I talk with anyways) never seem to take it seriously because they only see support for Switch dying the moment Switch 2 is announced.

Apple didn't kill the first iPhone the moment the iPhone 2 launched

What?

As someone who worked at Apple we never called the 3G “iPhone 2” but anyway, Apple did kill manufacture of the original iPhone with the 3G. Tech support still remained but we no longer sold the device. Now, if you bought it from a third party that had nothing to do with Apple. But for all intents and purpose the 3G had full backing of the companies resources moving forward after its launch.

 

[Skittzo]

Huh? I’m confused.

I was on the other side of this debate 2-ish months ago, and thought the conclusion that docked Drake could be 2.5tf-4tf before DLSS?

If we treat DLSS as a 2x multiplier then the effective tf would be 5.0tf-8.0tf.

Series S is 4 tf while Switch is 0.6tf and One S was 1.2.

So Switch is 0.5x a One S but could match the Series S at the high end, on the GPU side, before DLSS.

Meanwhile, I thought going from 3 A57s to 7 A78s was a bigger cpu jump that what we saw between One S and Series S as well.

Where am. I mistaken?

Short answer: we won't know until we have clocks.

Long answer:

Flops are a misleading metric for GPUs, especially compared across vendors. Generally Nvidia flops are more efficient than AMD flops, and an Ampere flop is different than a Maxwell flop, etc. etc. So the flops will not tell the full story. On the GPU side, without knowing clocks, we can likely assume that Drake will be able to hit close to 2TF, maybe higher or maybe lower (depending on clocks again) but this doesn't mean it will be literally half of a Series S. They can't be compared in that way. Still, it won't be up there in raw power. I've been saying it's likely to be in the PS4-PS4Pro range in raw power before DLSS is applied, and I think that's still applicable.

CPU wise 3A57s to 7A78s is indeed a huge jump, but so was the jump from XB1/PS4 to XBS/PS5, I'm not sure how the two jumps compare.

RAM will be an issue for Drake, specifically bandwidth, that will make it hard to compare more favorably to one of the stationary consoles.

Personally I think it'll be in a better position to receive ports from XBS than Switch was to XB1 mainly because of publishers and middleware evolving to better support Nintendo but I'm not sure if the actual real world specs will be any closer.

 

[LiC]

 

[Lancelot]

Short answer: we won't know until we have clocks.

Long answer:

Flops are a misleading metric for GPUs, especially compared across vendors. Generally Nvidia flops are more efficient than AMD flops, and an Ampere flop is different than a Maxwell flop, etc. etc. So the flops will not tell the full story. On the GPU side, without knowing clocks, we can likely assume that Drake will be able to hit close to 2TF, maybe higher or maybe lower (depending on clocks again) but this doesn't mean it will be literally half of a Series S. They can't be compared in that way. Still, it won't be up there in raw power. I've been saying it's likely to be in the PS4-PS4Pro range in raw power before DLSS is applied, and I think that's still applicable.

CPU wise 3A57s to 7A78s is indeed a huge jump, but so was the jump from XB1/PS4 to XBS/PS5, I'm not sure how the two jumps compare.

RAM will be an issue for Drake, specifically bandwidth, that will make it hard to compare more favorably to one of the stationary consoles.

Personally I think it'll be in a better position to receive ports from XBS than Switch was to XB1 mainly because of publishers and middleware evolving to better support Nintendo but I'm not sure if the actual real world specs will be any closer.

Well, tons of people in this thread have already mentioned how the Orin in Drake is almost 1/1 compared to desktop Ampere... The 3000 Series has already been shown to be more efficient than both RDNA1 and 2 in almost all workloads including raytracing. 2 TFLOPS in a handheld Drake is indeed half of Series S in that case, maybe even 1.8x depending on the exact efficiency overhead Ampere has over RDNA2.

 

[Kingpin Rogers]

Guessing we don't have any info on it yet but is it safe to assume Drake will support HDR? Not necessarily in the handheld part but in docked mode?

 

[ReddDreadtheLead]

Nobody is coding bare metal for this, dude... Software APIs like NVN exist for a reason, so you can code for the thing as long as it's powerful enough. Drake is, even if they'll have to carry DLSS as a crutch sometime.

What? That doesn’t mean that this is all done for free. Drake is fundamentally going to be a different GPU architecture than the other consoles which are very similar. Even if games aren’t coded to the metal anymore(mostly), that’s not a reason to believe that there won’t be degrees of differences to account for due to the different architecture.

Like you can’t just approach RDNA2-based console with respect to RT the same way as he would approach say an ampere-based console. Some things Ampere does that RDNA2 doesn’t, like proper BVH traversal.

Regardless if you believe there will be or won’t be RT, the architectures are different and how it’s approached is also different.

Hey, that'd be great. I’ll take holiday at this point! So how do you personally read or resolve Nate’s comments about H1 hardware being shelved? If he’s correct, it’s odd that they had one thing planned for H1 but shelved that device in favor of something different for H2 of the same year. And by odd I mean that sounds so bizarre, doesn’t it?

My take is that I don’t believe it at all. I don’t need an insider’s comment on this when nvidia first hand has given enough information to work with, without saying a single thing.


I won’t knock him for his direct knowledge though. He does know things there. But that’s about it.

 

[Lancelot]

What? That doesn’t mean that this is all done for free. Drake is fundamentally going to be a different GPU architecture than the other consoles which are very similar. Even if games aren’t coded to the metal anymore(mostly), that’s not a reason to believe that there won’t be degrees of differences to account for due to the different architecture.

Like you can’t just approach RDNA2-based console with respect to RT the same way as he would approach say an ampere-based console. Some things Ampere does that RDNA2 doesn’t, like proper BVH traversal.

Regardless if you believe there will be or won’t be RT, the architectures are different and how it’s approached is also different.




My take is that I don’t believe it at all. I don’t need an insider’s comment on this when nvidia first hand has given enough information to work with, without saying a single thing.


I won’t knock him for his direct knowledge though. He does know things there. But that’s about it.

Except you can? Like, with that logic how do you run a single game on PC for both RDNA2 and Ampere without major graphic bugs (even with DirectX raytracing carrying the former)? APIs have always been translation layers for radically different architectures, and Nvidia is definitely making sure NVN2 is as efficient and low-level as the original version for the Switch already was. You're talking like some amateurs developed this API and the drivers for Switch/Drake, if the hardware can indeed handle these workloads as it has the compute required, it will.

 

[Pokemaniac]

Yep also probably a pain to port x86 software to ARM, plus CPU strains.

Modern software is typically not especially tightly bound to a particular CPU ISA. It's not entirely a non-factor, but it's probably not the primary issue going between game consoles by a long shot.

 

[Skittzo]

Well, tons of people in this thread have already mentioned how the Orin in Drake is almost 1/1 compared to desktop Ampere... The 3000 Series has already been shown to be more efficient than both RDNA1 and 2 in almost all workloads including raytracing. 2 TFLOPS in a handheld Drake is indeed half of Series S in that case, maybe even 1.8x depending on the exact efficiency overhead Ampere has over RDNA2.

Oh maybe I was unclear, I meant it wouldn't be exactly half since generally ampere flops are more efficient. So if it's exactly 2tf on paper it would effectively be more than half.

 

[HankyPanky]

This is what I've been thinking. I think Nintendo (and most third parties) will try and support Switch and Switch NG for as long as they are, but I think we've hit the end of impossible ports. If a game is only viable on Switch NG, then if it shows up Switch OG at all, it will be a cloud version.

even if there's only a handful of releases years down the line that would still count as support. i can imagine any Pokemon games getting base Switch versions the whole way through, even if they look rubbish and run at 20fps (like the current one lol)

 

[LiC]

Guessing we don't have any info on it yet but is it safe to assume Drake will support HDR? Not necessarily in the handheld part but in docked mode?

I don't think there any limitations for it in the pipeline, and NVN2 supports it (on Windows) where NVN1 didn't at all. Not sure it's a slam dunk for a feature Nintendo would want to expose if it would be unavailable in handheld, but they could choose to do that.

 

[Vash]

Do you know what thread you're in

Yes, Future Nintendo Software Technology Speculation thread

 

[Brofield]

Much more relevant example than the iPhone: almost every major ps5 and series game since those launched had last gen versions.

True, but I only use the Apple comparison as the company Nintendo most closely emulates haha

What?

As someone who worked at Apple we never called the 3G “iPhone 2” but anyway, Apple did kill manufacture of the original iPhone with the 3G. Tech support still remained but we no longer sold the device. Now, if you bought it from a third party that had nothing to do with Apple. But for all intents and purpose the 3G had full backing of the companies resources moving forward after its launch.

My post essentially simplified my meaning, in that tech support/software was still maintained, updated and created for the old iPhone as well as the next generation.

I honestly didn't remember what iPhone 2/3G was originally called; I've only ever followed the Android ecosystem haha

 

[Foltzie]

What?

As someone who worked at Apple we never called the 3G “iPhone 2” but anyway, Apple did kill manufacture of the original iPhone with the 3G. Tech support still remained but we no longer sold the device. Now, if you bought it from a third party that had nothing to do with Apple. But for all intents and purpose the 3G had full backing of the companies resources moving forward after its launch.

With respect, while you are 100% factually correct, the general intent of the comment seems to have been that Apple continues to support older models of phones with software/OS updates. Further, even normally sells some prior models at a discount (even creating the SE line to get more sales from older production lines). So while the OG iPhone was retired aggressively from retail, it at least continued to get software updates for several years.

Now does Nintendo do the same thing? Hopefully.

 

[NateDrake]

@NateDrake

Is your podcast coming this week, any idea yet?

Still working on scheduling a recording time.

 

[lucapfstr]

Your points about SEs aren’t bad at all.

Hey, that'd be great. I’ll take holiday at this point! So how do you personally read or resolve Nate’s comments about H1 hardware being shelved? If he’s correct, it’s odd that they had one thing planned for H1 but shelved that device in favor of something different for H2 of the same year. And by odd I mean that sounds so bizarre, doesn’t it?

I’ll wait for Nate to clarify, but at the moment it doesn’t make any sense. Nintendo wouldn’t outright cancel something unless there’s a clear replacement. So two things that would make it “make sense” is the successor isn’t called a switch or the successor is coming this year and is replacing that revision.

Regardless of what anyone says, I’m pretty confident we get new hardware this year. It could be a small revision or a completely new thing. History and patterns are stronger than nonsensically saying 2024. Ngl if it’s not this year is 2025 IMO

 

[fwd-bwd]

The placement of the quote may be suspect but those are articles were only three months apart. If the Switch was in the middle of its lifecycle in November 2021 then it wasn't meaningfully past that characterization in February 2022.

It wasn't only a 3 month gap. In 2020 and 2021 Furukawa was comfortable with identifying the Switch as being in mid lifecycle, but did not repeat that talking point once through the whole 2022. It seems that he could no longer speak of it that way in an investor meeting or official interview.

 

[Pokemaniac]

I don't think there any limitations for it in the pipeline, and NVN2 supports it (on Windows) where NVN1 didn't at all. Not sure it's a slam dunk for a feature Nintendo would want to expose if it would be unavailable in handheld, but they could choose to do that.

Depending on the screen technology, enabling it in handheld mode might not be a problem.

 

[HankyPanky]

the Uncle is just grabbing a coffee before he gets to work assembling the Drake Switch. believe!

 

[Lancelot]

Oh maybe I was unclear, I meant it wouldn't be exactly half since generally ampere flops are more efficient. So if it's exactly 2tf on paper it would effectively be more than half.

Ah okay, that's what I mean. In any scenario Drake flops are still more efficient and at worst equally efficient, therefore... It has a very high chance of getting closer to Series S in docked than the TFLOP numbers actually say. Pair it up with DLSS and the RT cores, and I have no reasons to doubt it'll beat it in real world scenarios (including raytracing, which dictate image complexity for a decent number of people).

 

[ReddDreadtheLead]

Except you can? Like, with that logic how do you run a single game on PC for both RDNA2 and Ampere without major graphic bugs (even with DirectX raytracing carrying the former)? APIs have always been translation layers for radically different architectures, and Nvidia is definitely making sure NVN2 is as efficient and low-level as the original version for the Switch already was. You're talking like some amateurs developed this API and the drivers for Switch/Drake, if the hardware can indeed handle these workloads as it has the compute required, it will.

No one said NVN2 was inefficient? And NVN/NVN2 is only for Nintendo Tegra based console. That does not mean this is a simple job that can be done in an afternoon.

And I don’t really think DirectX12 is really a comparable thing in this context as that is not as low level as GNM or NVN or even the Xbox version of the DX12U API. The PC landscape isn’t 1:1 with the console landscape. DX12U isn’t for AMD or NVidia hardware or even INTEL. While like I mentioned NVN2 is only for the Nintendo Tegra based console, it’s not meant to take RDNA, or GCN, or anything else from AMD when developing for it even. It favors NVidia hardware of Turing or later.

And I’m not claiming that they’re amateurs that are making these APIs and drivers, you’re putting words in my mouth. I’m saying Drake being a different architecture does not mean it’s going to be as simple as press a button when developing for it compared to the other consoles, which are a lot more similar than Drake is.


For example: InfinityWard uses a lot of tricks that only work on AMD based hardware because the consoles are AMD-based hardware. Ij this instance, they use the MSAA hardware for VRS and compute culling, which increases the performance on AMD hardware. This carries over to the PC side which also does this. This isn’t done at all on Nvidia hardware, and it is ported from an external studio for MW.

and why you see such a disparity in the performance between 4090 and 7900XTX in benchmarks. Like the latter being 40% better. That’s extreme.


Trust me, I do understand where you’re coming from, but that is not really what I’m getting at.

 

[Skittzo]

I feel like all of the "halfway through the life cycle" talk is extremely pointless because it can really mean whatever you want it to mean. It's hard to glean too much from that IMO.

 

[Lancelot]

No one said NVN2 was inefficient? And NVN/NVN2 is only for Nintendo Tegra based console. That does not mean this is a simple job that can be done in an afternoon.

And I don’t really think DirectX12 is really a comparable thing in this context as that is not as low level as GNM or NVN or even the Xbox version of the DX12U API. The PC landscape isn’t 1:1 with the console landscape. DX12U isn’t for AMD or NVidia hardware or even INTEL. While like I mentioned NVN2 is only for the Nintendo Tegra based console, it’s not meant to take RDNA, or GCN, or anything else from AMD when developing for it even. It favors NVidia hardware of Turing or later.

And I’m not claiming that they’re amateurs that are making these APIs and drivers, you’re putting words in my mouth. I’m saying Drake being a different architecture does not mean it’s going to be as simple as press a button when developing for it compared to the other consoles, which are a lot more similar than Drake is.


For example: InfinityWard uses a lot of tricks that only work on AMD based hardware because the consoles are AMD-based hardware. Ij this instance, they use the MSAA hardware for VRS and compute culling, which increases the performance on AMD hardware. This carries over to the PC side which also does this. This isn’t done at all on Nvidia hardware, and it is ported from an external studio for MW.

and why you see such a disparity in the performance between 4090 and 7900XTX in benchmarks. Like the latter being 40% better. That’s extreme.

You still have no reasons to doubt it'll be able to show the compute that we can see right now. NVN2 is infinitely lower level and will be Drake's first representative when porting software to it, i'm not putting words in your mouth. You're just being stupidly pessimistic about imaginary real-world scenarios for Drake compute to be lesser in comparison to the others. That's not going to happen, no matter how much you doubt about its efficiency.

 

[hologram]

I feel like all of the "halfway through the life cycle" talk is extremely pointless because it can really mean whatever you want it to mean. It's hard to glean too much from that IMO.

Right. In 2020, it made perfect sense; in 2021, it seemed to imply a slightly longer-than-usual gap between consoles (since 2025 didn't seem out of the realm of possibility for a successor); in 2022, it became obviously meaningless. There's almost no chance that Nintendo doesn't release upgraded hardware before 2027

EDIT: Although this is of course with a fairly literal definition of "halfway through its life cycle". But if there's wiggle room anyway, it becomes a not terribly meaningful statement

 

[Concernt]

I feel like all of the "halfway through the life cycle" talk is extremely pointless because it can really mean whatever you want it to mean. It's hard to glean too much from that IMO.

3DS was in the middle of its life cycle a year before Switch launched, after all. Which is where we were with Switch in like 2021.

I'm confident this device is soon, as I've said before. Nintendo's IR meeting comments are quite concerting!

 

[ILikeFeet]

life cycles is such a meaningless metric. Switch lasting 10 years isn't that unreasonable if Nintendo wants to keep the submission process open for developer

 

[bloopland33]

My take is that I don’t believe it at all. I don’t need an insider’s comment on this when nvidia first hand has given enough information to work with, without saying a single thing.


I won’t knock him for his direct knowledge though. He does know things there. But that’s about it.

Appreciate you sharing your thoughts, thanks.

 

[RennanNT]

Huh? I’m confused.

I was on the other side of this debate 2-ish months ago, and thought the conclusion that docked Drake could be 2.5tf-4tf before DLSS?

Yeah, I think @Dakhil misunderstood the question with Series X or something like that. Even accounting TF comparisons between different architectures being flawed and going with a conservative "2TF docked", Drake GPU should be closer to the Series S than Switch was to the One S without DLSS.

Series S is 4 tf while Switch is 0.6tf and One S was 1.2.

Actually, Switch is 0.4TF, OG XB1 is 1.3TF and One S had a small bump to 1.4TF.

Meanwhile, I thought going from 3 A57s to 7 A78s was a bigger cpu jump that what we saw between One S and Series S as well.

Yes, it is a bigger jump. But it's worth to mention that Series S more than doubled the CPU clocks, so the gap could remain roughly the same if Drake has the same 1GHz clock as the Switch.

 

[JoshuaJSlone]

I think trying to read too much into executive quotes about lifespans is foolhardy. I just take it as: everyone with a decently successful console likes to say their current baby will be around forever. PlayStation 4 will have a 10 year life span, but so will PS3. Xbox One will last 10 years--conservatively!. But nobody should be surprised about that because the Xbox 360 also was to have a 10 year lifespan. I'm sure I read similar quotes about the PS1 in gaming magazines last century. But the market always decides, and the platform makers always have some new hardware ready many years before a decade is over. PS4 as a heavy software player does seem destined to last 10 years, even if its hardware is long since a non-factor.

 

[Kevin]

If the new chip really is basically a PS4 Pro, combining that with Nintendo's art and technical teams will yield some truly impressive results considering a base PS5 and PS4 Pro gave us titles like; Last of Us Part II, God of War & God of War: Ragnarok, Horizon: Zero Dawn & Horizon: Forbidden West, Sackboy, etc. All of these looked incredible on those systems and theoretically the Switch 2 chip could be even more capable with a more modern CPU and other architecture features not found in the PS4 or PS4 Pro.

Getting this in a portable hybrid system like Switch will be extremely cool. Can't wait to see what Nintendo's own development teams are able to achieve in the next 5+ years when this new system finally releases. I think we are all going to be in for a real treat because this upgrade is going to be a massive one.

 

[Lancelot]

Well, well... Seems like Bandai Namco has a collab with Nintendo for something, is there a chance this is a Drake game?

 

[Concernt]

Well, well... Seems like Bandai Namco has a collab with Nintendo for something, is there a chance this is a Drake game?

Yes.

 

[C_Ali]

True, but I only use the Apple comparison as the company Nintendo most closely emulates haha


My post essentially simplified my meaning, in that tech support/software was still maintained, updated and created for the old iPhone as well as the next generation.

I honestly didn't remember what iPhone 2/3G was originally called; I've only ever followed the Android ecosystem haha

I get it, just a caution in how some things are phrased vs. perceived.

With respect, while you are 100% factually correct, the general intent of the comment seems to have been that Apple continues to support older models of phones with software/OS updates. Further, even normally sells some prior models at a discount (even creating the SE line to get more sales from older production lines). So while the OG iPhone was retired aggressively from retail, it at least continued to get software updates for several years.

Now does Nintendo do the same thing? Hopefully.

The software aspect is obvious, the distinction in that comment (to me) were the models of the phone (SE didn’t start until after said generations). To kill a product can mean many things to a company. To Apple development of the iOS and maintenance of it were what kept back-compatibility feasible between generations(models). As someone on the inside I can tell you the continued support of the original iPhone was a consequence of essentially using the same kernel, just disable any features not applicable to a specific model. But you’re still developing internally for the latest generation and manufacturing the latest, at least Apple was, until software couldn’t be supported on older hardware thus support eventually dropped.

Ultimately, what I’m trying to say is “wasn’t this why Nintendo chose to work with Nvidia initially?” If neither company can pull off what is essentially cross generational development/support, why even have a deal? Absolutely nobody should expect the OG switch to just die off, but from Nintendo’s perspective if done right it should be essentially dead internally within Nintendo. A company always has to weigh the risks of supporting/dropping hardware and it’s exacerbated in Nintendo’s case as they have to sell THEIR own games. What resources can be allocated are heavily considered. This is why many companies have issues completely banishing generations, at some point you have to “kill” almost all support for a former generation internally, with very minimal resources, if any ,keeping it afloat.

I would say Nintendo will do what’s in their best interests but I don’t see the purpose of the Nvidia deal if it wasn’t to make a “family” of devices, with Switch 2 boasting some feature/s not available for OG but allowing Nintendo’s development to cover all their “Switch” devices while essentially focusing on the latest device.

 

[Lancelot]

If the new chip really is basically a PS4 Pro, combining that with Nintendo's art and technical teams will yield some truly impressive results considering a base PS5 and PS4 Pro gave us titles like; Last of Us Part II, God of War & God of War: Ragnarok, Horizon: Zero Dawn & Horizon: Forbidden West, Sackboy, etc. All of these looked incredible on those systems and theoretically the Switch 2 chip could be even more capable with a more modern CPU and other architecture features not found in the PS4 or PS4 Pro.

Getting this in a portable hybrid system like Switch will be extremely cool. Can't wait to see what Nintendo's own development teams are able to achieve in the next 5+ years when this new system finally releases. I think we are all going to be in for a real treat because this upgrade is going to be a massive one.

Note, it's a PS4 Pro with a superpowered CPU linearly when docked without any tricks and RT cores involved. When putting DLSS, these and modern techniques on the mix, it's easily beating Series S in real world scenarios. Nintendo is going to make truly groundbreaking things for this.

 

[Deleted member 887]

I think I've mentioned before how my expectations are pretty modest, but let's not pretend that those test scores can be for "literally anything else". They're for a an Orin chip for Nintendo. Probably Drake.

I'm not trying to rag on you or knock your hope or anything like that. Just want to keep some (subtle!) facts straight.

These tests are not for an Orin chip for Nintendo. dlssDonutTest is a generic DLSS test, and has been copied into the NVN2 tree from an external location. dlssDonutTest can be used to test multiple graphics API integrations for DLSS - NVN2, but also DX11, DX12, and Vulkan.

The test scripts posted have been altered over time, and at one point also tested a Vulkan integration of DLSS, with absolutely zero relationship to Nintendo, it's API, or its hardware, and used the same test cases and wattage labels.

I could write a long post about the other things that these numbers could refer to, but that's been done to death. I just want to say that there is no hard evidence that links these tests directly to Drake, and it entirely possible that the tests actually predate NVN2 and are, in fact, references to something that has meaning to the DLSS team and not the NVN2 team at all.

 

[Kevin]

Note, it's a PS4 Pro linearly when docked without any tricks and RT cores involved. When putting DLSS, these and modern techniques on the mix, it's easily beating Series S in real world scenarios. Nintendo is going to make truly groundbreaking things for this.

If it really is this good then that only amplifies my expectations and how great I think we'll have it.

 

[ReddDreadtheLead]

You still have no reasons to doubt it'll be able to show the compute that we can see right now. NVN2 is infinitely lower level and will be Drake's first representative when porting software to it, i'm not putting words in your mouth. You're just being stupidly pessimistic about imaginary real-world scenarios for Drake compute to be lesser in comparison to the others. That's not going to happen, no matter how much you doubt about its efficiency.

Can you please show me where I said I doubt that Drake will actually be able to handle the work load? Because now you’re just putting assumptions I never put out and trying to argue as if I put these assumptions out.

 

[Lancelot]

If it really is this good then that only amplifies my expectations and how great I think we'll have it.

It's almost like it's too good for Nintendo studios, right? Haha, jokes aside, i'm wondering how long it'll take for the EPD teams to make a Horizon in a regular basis, as the absolute bare minimum for every 1st party release.

 

[Lancelot]

Can you please show me where I said I doubt that Drake will actually be able to handle the work load? Because now you’re just putting assumptions I never put out and trying to argue as if I put these assumptions out.

Sigh... You've been literally implying all this time Drake (Ampere) is a different architecture to the RDNA2 in PS5 and Series X, therefore it can't be directly compared to them. Sure, but all your arguments fall apart when Orin is almost 1/1 Ampere, and it's been shown to be more efficient flop-per-flop than the architecture in those consoles. The only way to not get this performance is if Nvidia fucked up the NVN2 implementation, which as i'm telling you, that's not going to happen.

Note, you changed the ballparks from originally being an ARM architecture like the Switch, therefore = bad and now you moved to the GPU being different in comparison... Again, sure, but do you know AMD has historically struggled to hit the efficiency levels that Nvidia architectures do? You used AMD so much times in your old message i'm even wondering if you know who's the manufacturer behind this, the architecture in Drake is superior to the next gen consoles even if it doesn't have the same level of compute those have. Therefore, this extra performance must appear on the hardware as long as we have a properly made NVN2 for it.

 

[Concernt]

I'm not trying to rag on you or knock your hope or anything like that. Just want to keep some (subtle!) facts straight.

These tests are not for an Orin chip for Nintendo. dlssDonutTest is a generic DLSS test, and has been copied into the NVN2 tree from an external location. dlssDonutTest can be used to test multiple graphics API integrations for DLSS - NVN2, but also DX11, DX12, and Vulkan.

The test scripts posted have been altered over time, and at one point also tested a Vulkan integration of DLSS, with absolutely zero relationship to Nintendo, it's API, or its hardware, and used the same test cases and wattage labels.

I could write a long post about the other things that these numbers could refer to, but that's been done to death. I just want to say that there is no hard evidence that links these tests directly to Drake, and it entirely possible that the tests actually predate NVN2 and are, in fact, references to something that has meaning to the DLSS team and not the NVN2 team at all.

Yeah, that doesn't explain why they're specifically testing the power consumption of Switch. That's... A pretty direct link to a Switch. And we know the Drake is for a Switch.

It tested how Vulkan performed in given power constraints, which happened to be those of Nintendo Switch, and was right there in the NVN2 tree. Which makes it pretty obvious that these are performance goals for a certain piece of silicon NVN2 is meant to run on it. If they had this performance goals in mind last February, I am sure they had the limitations of the node in mind. Will we see the device reach those speeds? Maybe not. I don't personally think we will. But we do have clear evidence they were used as developer targets, so it's also not likely we'll see a Drake Switch land drastically under those scores. A bit, maybe 40% less? But 60, 70, 80% less? No.


There's no device I can think of that needs to fit in the Switch's power constraints and would make use of DLSS. It's just extremely unlikely such tests were for anything BUT the next Switch processor.

That said I think, maybe if it is not developer targets, probably the DLSS team showing the NVN2 team the power savings DLSS allows. Possibly. Either way, more likely Switch than not Switch.