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

[Deleted member 5277]

I dunno man, you might be right about developers looking at it differently but I keep pulling up these videos of 4K BOTW for years now and always being disappointed with how same it looks to me. Side-by-side huge difference, ya, and maybe on a HUGE screen the difference will be super clear as well... but on a medium size screen it feels very meh to me. The 60 fps is something that makes me desire it but that doesn't change the look. So I'm just wondering what was this really showing off?

Draw distances? Meh. Could they have implemented RTX in it? Has it been a testbed for all the wizbang new features of drake and they got something going that's pretty damn impressive?

Again its unlikely the demo was intended to wow people into buying the Switch 2. They likely had other demos to do that, we know they had the Matrix demo. It likely served a very basic purpose, to show something very specific to the developers who were in theory supposed to be the sole audience to ever see it, as we've speculated it could be something like

"Look, we have a large enough power jump that at a bare minimum we can go from 900p/30 for BoTW to (much higher resolution)/60 on Drake" which if it is 4k, that isn't nothing, assuming they used DLSS to reach that which seems like a given.
Or it was, "Look at how easy it was for us to get old code running on our new API, it took (comparatively small amount of time) to get this".
Or it could even be a sell on how easy it is to use DLSS to get your game up to those resolutions on their system, ignoring even the pure computational jump.

They're trying to sell devs on the idea of working on their hardware, not on how Nintendo games will look. Sadly without knowing what exactly was said and shown its hard to say what the sell was.

 

[mjayer]

Using a previous generation game is not meant to show off how pretty the potential graphics are; that's what the Matrix demo was for (which had the audience so hyped they started making irresponsible PS5 comparisons). If BotW was running at 4K res and 60 fps, that's 5.76 times more pixels than the original (dynamic) 900p while doubling the framerate, so more than 11 times as much GPU work done per second, all for the low cost of switching NVN to NVN2. Anything else would be a bonus.

Ya I guess so... just seems like something you could say and wouldn't need to show, which is my point... it doesn't wow. (IMO)

 

[LiC]

Ya I guess so... just seems like something you could say and wouldn't need to show, which is my point... it doesn't wow. (IMO)

It's not a Nintendo Direct. You were never supposed to be "wowed" by it because you were never supposed to hear about it. I and tgtCG have already explained what its purpose was. And it's strange that your criticism of this demo is that Nintendo should have told devs "trust us bro" instead of showing them practical examples to prove what they were saying. Why show the UE5 demo, then? They could have just told everyone, "Remember the Matrix demo? We can run that at 4K and it looks as good as a PS5. Anyway, thanks for coming, bye!"

 

[LuigiVs]

so how early do they generally do private console demos ahead of release? could it mean we'll get some kind of teaser before the year ends?

We already know they demoed the Switch at E3 before the October reveal. I can't imagine them putting out even small teaser before December. They don't need as much lead up as they did last time, so it depends on when they plan to release and how much time they want to hype it up.

 

[Concernt]

Pocketnow covering Nintendo Switch 2

Well, there we have it. Into the mainstream it goes!

The goose is loose now, Nintendo, the ball is in your court. You're in the same arena as iPhone now!

 

[JoshuaJSlone]

Ya I guess so... just seems like something you could say and wouldn't need to show, which is my point... it doesn't wow. (IMO)

Then it's a good thing this wasn't part of a big presentation to the world.

 

[Serif]

so how early do they generally do private console demos ahead of release? could it mean we'll get some kind of teaser before the year ends?

Some developers were shown Switch 1 demos behind closed doors at E3 2016, so ~4 months before the official reveal. Though different scenario with the Wii U being essentially dead at the time and 'NX' being public knowledge.

 

[Herb Alpert]

Pocketnow covering Nintendo Switch 2

Well, there we have it. Into the mainstream it goes!

The goose is loose now, Nintendo, the ball is in your court. You're in the same arena as iPhone now!

It will be a nightmare to get one on launch day

 

[Discostew]

It will be a nightmare to get one on launch day

And I'm planning on getting two of them.

 

[SoldierDelta]

I dunno man, you might be right about developers looking at it differently but I keep pulling up these videos of 4K BOTW for years now and always being disappointed with how same it looks to me. Side-by-side huge difference, ya, and maybe on a HUGE screen the difference will be super clear as well... but on a medium size screen it feels very meh to me. The 60 fps is something that makes me desire it but that doesn't change the look. So I'm just wondering what was this really showing off?

Draw distances? Meh. Could they have implemented RTX in it? Has it been a testbed for all the wizbang new features of drake and they got something going that's pretty damn impressive?

While that's a reasonable point to make, that's not the purpose of the demo. Even if they just showcase how powerful the system is via resolution and framerate, that's good enough for a lot of developers. It's not a "game" showcase, it's a technical one.

As for RTX or any other new "features", that's best saved for other technical showcases at later dates. Y'know, different demos for different purposes.

 

[Concernt]

It will be a nightmare to get one on launch day

As of the latest "Funcleposting", production is underway in some capacity. So an early-mid 2024 launch should have plenty of stock.

But yes, this thing isn't just anticipated by dedicated nerds, this thing isn't even revealed and it already has widespread mindshare.

 

[LiC]

We already know they demoed the Switch at E3 before the October reveal. I can't imagine them putting out even small teaser before December. They don't need as much lead up as they did last time, so it depends on when they plan to release and how much time they want to hype it up.

Some developers were shown Switch 1 demos behind closed doors at E3 2016, so ~4 months before the official reveal. Though different scenario with the Wii U being essentially dead at the time and 'NX' being public knowledge.

The above has been brought up a lot, but what we know is that Todd Howard saw a Switch demo at E3. That's not necessarily the same as what happened at Gamescom. The story from Gamescom seems like a bunch of people got invited in to sit for a presentation -- enough for it to leak! -- whereas if only the likes of Todd Howard were being invited in 2016, that sounds more like some exclusive previews for decision-makers.

 

[SoldierDelta]

It will be a nightmare to get one on launch day

In the immortal words of Shaggy from the hit masterpiece "Scooby Doo" (2002), "Are you challenging me?"

 

[RedSpring]

Ya I guess so... just seems like something you could say and wouldn't need to show, which is my point... it doesn't wow. (IMO)

You're thinking from the viewpoint of a player not as a dev who would be working on the device. Visual appeal isn't the main thing you would want. If someone come into your work selling a product that they want you to use without a demo or only talking about how pretty their UI instead of the workflow, you would politely but firmly tell them to leave.

 

[YolkFolk]

How do you think games like Xenoblade 2 will look on the Switch 2’s alleged 8” 1080p display?

 

[Concernt]

politely but firmly tell them to leave.

As if they ordered something well done.

 

[Serif]

How do you think games like Xenoblade 2 will look on the Switch 2’s alleged 8” 1080p display?

Bad without a patch.

 

[Paul_Subsonic]

Do y’all think any of the stuff in this video is still relevant?

Obviously, DLSS has been updated and refined since then, plus they didn’t have a clue what ships/technology Nvidia and Nintendo would be using by the year 2024.

I was just curious if any of the technical hurdles and such still add anything to the conversation today based on what we know now.

(As an aside, it’s been a little annoying hearing so many podcast with smart people discussing all the Switch 2 rumors, but they seem to be missing a ton of the technical details leaked from Nvidia that this board seems to have a lot of familiarity with.)

I'd say mostly no.
The video is based on Turing (which performs worse than Ampere for DLSS), and napkin math even shabbier than the Drake DLSS Estimator 6000.
This was an interesting video in 2021 to see if the idea of Dlss on a switch system was possible, but I think in 2023 it's not really relevant anymore.

 

[Deleted member 887]

It essentially is free extra performance as long as you're comparing it to natively rendering the targeted output resolution. If you start at the native input resolution and add DLSS on top, then your performance will drop a bit in exchange for a big IQ boost. It's great tech either way, but you can look at it from different angles.

Skittzo is likely thinking in terms of T239, where we're imagining DLSS running on a RTX card cut down to the bone, and the costs of DLSS are maximized. There has never been an RTX card as small as the leaked GPU in Drake, and there is reason to believe that there will be some edge cases where DLSS is actually more expensive than native rendering.

Do y’all think any of the stuff in this video is still relevant?

No, but a little yes? Alex's analysis is very good here, but his data is actually surprisingly good, but he's got a missing piece that changes the result pretty dramatically.

Alex estimates Switch 2 performance based on Orin, and estimates DLSS 2 cost based on a 2060, and then tries to extrapolate how long 4k DLSS would take on Switch 2. He lands at about 10ms, which is too slow for 60fps, but pretty good for 30fps.

Since then we've gained more information about both Orin and DLSS 2, and the best estimates we have put DLSS 2 taking anywhere from 2-6ms. Prediction is hard, and Drake sits at a place that is especially fuzzy for this kind of estimation, but the range is still solidly in 60fps territory.

I've been the DLSS pessimist over the course of this thread, but by now I'm 75% convinced that the faster number, the 2ms number, is closest to the correct one.

 

[Alovon11]

Skittzo is likely thinking in terms of T239, where we're imagining DLSS running on a RTX card cut down to the bone, and the costs of DLSS are maximized. There has never been an RTX card as small as the leaked GPU in Drake, and there is reason to believe that there will be some edge cases where DLSS is actually more expensive than native rendering.


No, but a little yes? Alex's analysis is very good here, but his data is actually surprisingly good, but he's got a missing piece that changes the result pretty dramatically.

Alex estimates Switch 2 performance based on Orin, and estimates DLSS 2 cost based on a 2060, and then tries to extrapolate how long 4k DLSS would take on Switch 2. He lands at about 10ms, which is too slow for 60fps, but pretty good for 30fps.

Since then we've gained more information about both Orin and DLSS 2, and the best estimates we have put DLSS 2 taking anywhere from 2-6ms. Prediction is hard, and Drake sits at a place that is especially fuzzy for this kind of estimation, but the range is still solidly in 60fps territory.

I've been the DLSS pessimist over the course of this thread, but by now I'm 75% convinced that the faster number, the 2ms number, is closest to the correct one.

Yeah like, take a look at the TOP comparison Alex uses to facilitate those numbers.

T239 at Docked Clocks akin to Thraktor's clocks will very likely be over 20 times faster than the 10 TOPs Alex Estimated for Docked performance based on ADAS. (Orin AGX 32GB having 14SMs at 900MHz resulting in 200 TOPs, so 12SMs at 1.1GHz should be similar if not better)

 

[Pokemaniac]

That's how it worked with DLSS 1, where it was using neural networks trained for each individual game, and the version of DLSS everyone agreed was worse than the current.
DLSS 2 onward uses a neural network that is updated by Nvidia through driver updates for use by all titles. There's definitely some code updates that need to be done to the games themselves over time, but the neural network is not touched by developers directly because it sits within the hardware driver itself. At least, on the PC side of things. As I mentioned, Nintendo's patent seems to involve offering a different deployment method while keeping the underlying principle of how neural networks are generated, trained, tested and used unchanged.

It's sounding like you're treating neural network updates as though it were an untested code update, when that's not the case. New training samples cannot destroy the neural network's functionality, and that is entirely by design/how machine learning works; if new training samples cause failure to replicate the target image in the absolute barrage of tests performed, those results are disposed of, as it self-selects only the methods that produce desirable results across all of its samples. As it only spits out a better result when that better result is ready and only then is distributed to hardware through updates, a new version of that neural network that generates worse results than the previous iteration is so unlikely as to be not worth much consideration.

As I said, this is how DLSS Super Resolution already functions as of v2, so if you find the method lacking in consistency or reliability, you are arguing against its use in totality.

It's frankly not a comment of merit, console games are already "tuned for form factor, heat and [when relevant] battery", even PS5 and XBS when that same demo was built to run on them.

Yeah, that's not actually how it works. As far as I can tell, the models (and yes, there is more than one, with slots reserved for additional models) are shipped as part of the DLSS redistributables and don't change unless the developer explicitly opts in to a (pretty new) OTA update feature, and even that will only situationally update the model. The driver is seemingly not involved in this process.

A new model is the same as untested code in this context. There is not a simple linear progression of more training==more better, which Nvidia acknowledges by providing multiple options in the first place. There's no guarantee that a new model will work just as well with your specific art style or renderer as the one you tested the game with and were satisfied with.

 

[Deleted member 887]

I'd say mostly no.
The video is based on Turing (which performs worse than Ampere for DLSS), and napkin math even shabbier than the Drake DLSS Estimator 6000.
This was an interesting video in 2021 to see if the idea of Dlss on a switch system was possible, but I think in 2023 it's not really relevant anymore.

For those of you following along, Paul did the best work on DLSS performance estimation that convinced me 60fps was possible. I've actually come around to a simpler estimation since then, which gives even better numbers, but might be too simple.

So where Paul was the optimist and me the pessimist for a while, it's sort of Switched, where I consider Paul's numbers to be the worst case figures, and these other numbers the best case.

 

[Multiverse]

so we all thinking Breath of the Wild will look like this? on Switch sucessor

Never liked these mock-ups. The contrast is too high. BOTW has a very harmonious color palate. The big reason is there’s almost no blacks in it. The shadows are hazy in color, with no blacks. It’s like an impressionist painting.

Turning Red actually did something similar:

Edit: watched the video and it’s fine. The thumbnail made it look like one of those people that turn the contrast and saturation to an 11 for no reason.

 

[Deleted member 887]

Yeah like, take a look at the TOP comparison Alex uses to facilitate those numbers.

T239 at Docked Clocks akin to Thraktor's clocks will very likely be over 20 times faster than the 10 TOPs Alex Estimated for Docked performance based on ADAS. (Orin AGX 32GB having 14SMs at 900MHz resulting in 200 TOPs, so 12SMs at 1.1GHz should be similar if not better)

Not to burst your bubble, but not at all. Orin and desktop Ampere don't have the same TOPS numbers. Orin has a second bank of tensor cores (branded the Deep Learning Accelerator) which RTX 30 (and Drake) don't have. Orin also runs some tensor ops at double speed, again, Drake and RTX 30 do not do that.

So Alex is actually using too high a baseline to estimate from, but then he's referring to Orin Nano which is much smaller than Drake. Weirdly, these things nearly cancel out. Drake would have like 23 INT8 TOPS, by your own numbers, not 200. 46 with sparsity.

 

[Serif]

DLSS'ing to 1080p and 1440p and then spatial upscaling to 2160p will still produce great results, as DLSS applies AA to the image, and at 'living room distance' properly upscaled 1080p/1440p are still sharp and detailed.

It's unfortunate that 4K TVs do not scale 1080p well by default. On a LG CX at 0 (default) sharpness, 1080p is blurred to all hell.
Turn it up to at least 20 and it's closer to nearest neighbor, and looks sharp. I tested this myself. Native 1080p games like Mario Kart and Dark Souls look fantastic on my set.

Since the Switch 2 will be sending a 2160p signal to the TV, the work to upscale 1080p will be performed by the console, and it will hopefully not be a blurry upscale. For pixel art games, we could see an immediate '4K upgrade' through a 4X integer scale, without patches.

If 1080p after DLSS ends up being the resolution for the most demanding games, I would be more than satisfied.

 

[Multiverse]

I fully assume the BOTW demo is supposed to show backwards compatibility.

And the Matrix Demo shows what it can do for third-party games with all the engines and the new graphical effects.

 

[Hermii]

Yeah like, take a look at the TOP comparison Alex uses to facilitate those numbers.

T239 at Docked Clocks akin to Thraktor's clocks will very likely be over 20 times faster than the 10 TOPs Alex Estimated for Docked performance based on ADAS. (Orin AGX 32GB having 14SMs at 900MHz resulting in 200 TOPs, so 12SMs at 1.1GHz should be similar if not better)

But Orins tensor cores are twice as fast as desktop/ Drake, did you account for that?

 

[ILikeFeet]

I fully assume the BOTW demo is supposed to show backwards compatibility.

This is definitely an important piece of context that we are missing. Was this native or BC, because the latter would be the important thing we're all waiting on confirmation on. Given that the leakers haven't spoken on whether or not it was Switch code or native code, I assume Nintendo said nothing about that and just wanted to focus on pixel pushing over the Switch

 

[TheSentry]

What if the BotW demo was to show off ray tracing and GI situation running at 4K 60fps, or something like that..

 

[Hermii]

This is definitely an important piece of context that we are missing. Was this native or BC, because the latter would be the important thing we're all waiting on confirmation on. Given that the leakers haven't spoken on whether or not it was Switch code or native code, I assume Nintendo said nothing about that and just wanted to focus on pixel pushing over the Switch

Yea, woudnt make sense to show off BC in a dev demo imo.

 

[Deleted member 887]

from your percpective, did the matrix demo make full of use of the hardware specs as well as DLSS? as well as, was it able to deliver as expected or slightly better?

I don’t think @oldpuck was present at Gamescom.

well based on what we've heard so far at least lol

Haha, yeah, I was not at Gamescom this year - next year though! But I won't be invited to any Nintendo demos regardless, I'm fairly certain.

Let's rephrase - if you had asked me to guess how well the Matrix Awakens might run on Switch 2, I'd say that a really careful port with some custom settings might look better than Series S, at similar frame rates. We don't know details, but that's what it sounds like happened. So I'd say right on the money.

 

[Alovon11]

Not to burst your bubble, but not at all. Orin and desktop Ampere don't have the same TOPS numbers. Orin has a second bank of tensor cores (branded the Deep Learning Accelerator) which RTX 30 (and Drake) don't have. Orin also runs some tensor ops at double speed, again, Drake and RTX 30 do not do that.

So Alex is actually using too high a baseline to estimate from, but then he's referring to Orin Nano which is much smaller than Drake. Weirdly, these things nearly cancel out. Drake would have like 23 INT8 TOPS, by your own numbers, not 200. 46 with sparsity.

If it were half the rate, then it'd be 100TOPs docked, not 23?

The DLA doesn't just double the TOP number, the TOP number on Orin's scale scales somewhat well as corecounts and clocks drop.

 

[Alovon11]

But Orins tensor cores are twice as fast as desktop/ Drake, did you account for that?

Not sure if it's double the rate, but even then it'd be 100TOPs, 10X faster than the 10TOPs Alex Speculated in that video

 

[mjayer]

I dunno I think if you have a demo you are showing devs to get them excited to work on your new platform, and this demo is meant to show what is possible... well I just expect more than "Ya this was ported over easy, and it runs at a higher resolution." and then they sit there and watch 30 seconds of 4k BOTW in awkward silence. "Okay our next demo is something else you've seen before..."

Welp.

I could be wrong and it really is just that but I WANNA SEE IT. That's all I'm saying.

 

[ILikeFeet]

What if the BotW demo was to show off ray tracing and GI situation running at 4K 60fps, or something like that..

that'd be some EXTREMELY important details to leave out. the answer might be simpler in that they were just demonstrating outputs. the Matrix already covers the ray tracing aspect

I dunno I think if you have a demo you are showing devs to get them excited to work on your new platform, and this demo is meant to show what is possible... well I just expect more than "Ya this was ported over easy, and it runs at a higher resolution." and then they sit there and watch 30 seconds of 4k BOTW in awkward silence. "Okay our next demo is something else you've seen before..."

Welp.

I could be wrong and it really is just that but I WANNA SEE IT. That's all I'm saying.

that's what the Matrix demo is for. these demos don't exist in a vacuum

 

[MadMarvainy]

What storage format do we think they will use? I imagine it will be something in between the high-speed NVMes and the flash storage on Switch.

R&C on Steam Deck is noticeably slower than its PS5 counterpart so I wonder how games that target SSDs will fair on Switch 2.

 

[Danny]

I dunno I think if you have a demo you are showing devs to get them excited to work on your new platform, and this demo is meant to show what is possible... well I just expect more than "Ya this was ported over easy, and it runs at a higher resolution." and then they sit there and watch 30 seconds of 4k BOTW in awkward silence. "Okay our next demo is something else you've seen before..."

Welp.

I could be wrong and it really is just that but I WANNA SEE IT. That's all I'm saying.

I mean it makes sense to show how a game made for Wii U/Switch can look when ported over to Switch 2. But you're acting like the BOTW demo was the only thing they showed at the demo. We know that they also showed the Matrix Awakens demo and probably other demos as well besides those two. I don't really get what you're complaining about. This was a demo for devs to get a sense of the new consoles specs, not to gain hype from fans.

 

[TechnicalTip6578]

Haha, yeah, I was not at Gamescom this year - next year though! But I won't be invited to any Nintendo demos regardless, I'm fairly certain.

Let's rephrase - if you had asked me to guess how well the Matrix Awakens might run on Switch 2, I'd say that a really careful port with some custom settings might look better than Series S, at similar frame rates. We don't know details, but that's what it sounds like happened. So I'd say right on the money.

understandable, thank you for the answer & apologies for the ridiculous question lol

 

[Thraktor]

The "souped up" BOTW demo is very curious to me. It's possible that it's just a resolution and/or frame rate bump, and there't nothing wrong with showing something like that to developers. Taking a game that ran at ~900p/30fps on previous gen hardware and showing it running at 4K/60fps, or thereabouts, definitely indicates a significant bump in performance. It's a curious choice for showing that kind of performance bump, though, given its age. TOTK has been out long enough that they surely would have been able to get it running at a decent level on Switch 2 by this point. Not good enough for release, I should clarify, but good enough to use as a demo for a technical audience. Nobody watching these demos is expecting to see something that's 100% polished and ready to release, just something that gives them an idea of the performance level of the hardware.

A theory I have is that BOTW was used as an internal testbed for new technologies and techniques on the new hardware, and they're showing developers the results of that. Let's say you're Nintendo in 2020 or 2021 or whenever, and this new hardware you were planning has got to the point where you can actually start doing some early software tests. You've got a small team of graphics engineers, and you want to get up and running quickly and really put the system through its paces; ray tracing, DLSS, asset streaming, etc. You probably don't want to waste time building something from scratch, so you need to choose an existing game to build upon. Firstly, you want something built with an engine the team is already familiar with, which means an EPD game. Secondly, you want a game that's finished, not something in development where code is constantly changing. Finally, it should be something that is graphically ambitious and gives you room to really push the hardware.

BOTW seems like the obvious choice. Want to test out the streaming tech and FDE? There's a big open world for you to zoom around. Want to test out some probe-based RTGI? BOTW already has a probe-based GI setup, which you can retrofit for RT. Want to test out DLSS? Well, to be honest it's not any better than any other EPD game for that, but no worse, either. Taking BOTW and using it to try to push Switch 2 hardware (or early prototypes, etc.) as hard as they can seems like a pretty good way to get early feedback on the new device. It's also conceivable that it would then end up as a pretty good way to show off the hardware to developers too.

 

[Serif]

A theory I have is that BOTW was used as an internal testbed for new technologies and techniques on the new hardware, and they're showing developers the results of that.

This tracks. At GDC '17 Nintendo shared how they used previous Zelda game assets to test their HD engine for BotW. Their experiments with Wind Waker ended up becoming WWHD, here's TP and SS as well.

 

[BDGAME]

About the Matrix demo.
The leakers say it is running with ray tracing and, I belive in a stable frame rate. We don't now about the native resolution or the final output, but we can make some assumptions.

First if we compare with Steam deck. There is any configuration where that demo runs well in deck, even without the Lumen? if not, this put the hardware better than deck in GPU and in CPU.

The other part is ne final output. If they have all the car traffic and particles that xbox Series S had, plus a output bigger than 1080p, thanks to DLSS, then that small portable system will be another Nvidia's black magic, because it will be capable to run any current gen game in a acceptable level.

What we can't imagine right now is how everything will work on portable. I mean, if that demo is already running really low, like 540p, how they will make run well on portable? Can 360p to 1080p will be enough to compensate the diferences?

 

[chocolate_supra]

Never liked these mock-ups. The contrast is too high. BOTW has a very harmonious color palate. The big reason is there’s almost no blacks in it. The shadows are hazy in color, with no blacks. It’s like an impressionist painting.

Turning Red actually did something similar:

That's the biggest issue I have with a lot of 4K60RTX BotW videos, they take away a lot of the painterly look that gives BotW that humid, outdoorsy feeling. I expect what Nintendo does with that sort of graphical power will be much closer to that screenshot of Turning Red (or if I dare to dream, something like Puss In Boots TLW ) than a lot of the "BotW, but moar grafficks!!" videos that we see on youtube.

Edit: watched the video and it’s fine. The thumbnail made it look like one of those people that turn the contrast and saturation to an 11 for no reason.

Damn clickbait thumbnails

 

[ILikeFeet]

About the Matrix demo.
The leakers say it is running with ray tracing and, I belive in a stable frame rate. We don't now about the native resolution or the final output, but we can make some assumptions.

First if we compare with Steam deck. There is any configuration where that demo runs well in deck, even without the Lumen? if not, this put the hardware better than deck in GPU and in CPU.

The other part is ne final output. If they have all the car traffic and particles that xbox Series S had, plus a output bigger than 1080p, thanks to DLSS, then that small portable system will be another Nvidia's black magic, because it will be capable to run any current gen game in a acceptable level.

What we can't imagine right now is how everything will work on portable. I mean, if that demo is already running really low, like 540p, how they will make run well on portable? Can 360p to 1080p will be enough to compensate the diferences?

the demo running better than Steam Deck is most likely less due to hardware and more due to software optimization. all these SD videos are using the 5.1 version (or whichever) the demo launched with and has no optimization done other than turn down the settings as low as they can go

That's the biggest issue I have with a lot of 4K60RTX BotW videos, they take away a lot of the painterly look that gives BotW that humid, outdoorsy feeling. I expect what Nintendo does with that sort of graphical power will be much closer to that screenshot of Turning Red (or if I dare to dream, something like Puss In Boots TLW ) than a lot of the "BotW, but moar grafficks!!" videos that we see on youtube.

it's something I noticed about a lot of reshade ray tracing videos. their colors are fucked to high hell. RT doesn't increase contrast/gamma like that

 

[ReyVGM]

I will lose my excrement if Elden Ring is announced for the Ultra Switch!

 

[SoldierDelta]

I will lose my excrement if Elden Ring is announced for the Ultra Switch!

I read this as "excitement" first and I was like "Why would you do that?" and then I reread it and just said out loud "Lmao based".

I should probably read slower sometimes...

 

[Concernt]

I read this as "excitement" first and I was like "Why would you do that?" and then I reread it and just said out loud "Lmao based".

I should probably read slower sometimes...

The Nintendo Switch Detritus

 

[OniLgnd]

I could be wrong

You are. Plenty of others have explained it far better than I could, but I'll try to put it another way: You are thinking like a gamer, not like a developer. These meetings are not for gamers, there will be different showcases for us. They are for developers who care about things like how easy it is to get their games running on the new system, and a bunch of other little things that either of us don't even know about because we are not developers.

 

[Hermii]

About the Matrix demo.
The leakers say it is running with ray tracing and, I belive in a stable frame rate. We don't now about the native resolution or the final output, but we can make some assumptions.

First if we compare with Steam deck. There is any configuration where that demo runs well in deck, even without the Lumen? if not, this put the hardware better than deck in GPU and in CPU.

Seconding what Ilikefeet said. This demo is probably optimized by epic and Nvidia themselves using NVN 2. You shouldn't underestimate the difference this makes over the SD running the pc version over proton.

 

[LinkURL]

BotW in 4K@60 with original assets (without compression), plus effects and draw distance at maximum and an RT global illumination would be very impressive.

 

[Deleted member 887]

If it were half the rate, then it'd be 100TOPs docked, not 23?

The DLA doesn't just double the TOP number, the TOP number on Orin's scale scales somewhat well as corecounts and clocks drop.

The DLA is a separate bank of tensor cores, which run at the same clock as the rest of the GPU. There are still tensor cores in the GPU as well. So yes, when core counts come down, so do the TOPS, when the clock goes down, so do the TOPS.

Instead of trying to reverse engineer from Orin's numbers, just use desktop Ampere, since we know they are identical.

256 ops/clock, 4 tensor cores per SM, doubled for sparsity
512 sparse ops * 1GHz clock speed * 48 tensor cores = 24.576 sparse INT8 TOPS