Based on how tech has progressed since the Switch, how powerful could a Switch Pro in early 2021 actually be?
- Thread starter ItWasMeantToBe19
- Start date
- Status
If I have same kind of arguments and my English is not my native language, that doesnt mean that those arguments are bad or wrong.
Also I dont understand this talk about same person, because I post only on this place, I am not posting on other places on internet.
You need HDMI 2.0 or later in the dock for that, and enabling it by update for devs to patch that in, don't think you necessarily need a pro for that, but a better dock for sure.
Hm Ok, more inline with what I was expecting,
Thought that 10 seemed like a lot, but good to know it was clarified.
XB1S is 1400gflops, PS4 is 1843gflops or just over 31% faster, though XB1S has a faster CPU, so games needing to use gpgpu functions would bring the GPUs a bit closer together, as much as 25% gap.
That gap is where I believe the Switch Pro will end up, not a very big range, it's really splitting hairs because Ampere has stuff like VRS that would allow it to out perform PS4's GPU in newer games, again before taking DLSS into account.
I posted years ago on NeoGaf and VGchartz, but that was years ago and I think I had different nick names..
My dude, it's a common name. Let's roll it back a bit and talk about vidja.
Let's all just take it down a peg so Warnings dont get sent.
Probably gonna be 30 regardless of Res.
They seemed to have settled on that FPS rate.
Yeah, I don't think it will get close to PS4's memory bandwidth, highest it can go is 102GB/s, without being unrealistic, still Ampere is a lot better with memory bandwidth than GCN.
Hmm? That's kind of surprising given how insanely well MK8D has been selling.
Where are these rumors coming from?
Sorry, I missed your preceding post. I am out of here and won't distract any further.
the current switch could have had HDR if Nintendo made a better dock
It's cool man don't worry, but I noticed you're new here so I have to warn you that it's very easy to get banned on Era if mods percieve you to be causing issues, etc.
Actually Nintendo really cares about its 3rd party support. You can easily make your game around a certain hardware a deliver the best of the silicon when you are making first party games. The problem is, no 3 re party Dev will make their game around a weak hardware especially if it's not the lead platform of the generation. What nintendo will aim with this new model is to make the lead platform of the PS5 gen consoles. It will not be possible in 2023 whenthere will be 40M PS5 and 0 switch 2 on the market. That's why an 8 cores CPU is plausible. It would maintain parity of CPU count with PS5's APU.
This. Nintendo has (for the first time since SNES) the upper hand on the home console market. They will probably use a more available 8 nm node at a lesser cost than Sony's 7 nm highly clocked APU at their advantage. But for that, their futur SOC have to have all the features of the PS4/XSS features and being compatible with all next gen engines.
Everybody is talking about the absurdity of 1.2~1.8TFLOPs SOC with XOS performances (before DLSS) in 2021. I actually expect this new model to be on par with XOS performances at 720p in handheld mode... Even before DLSS. The Galaxy S10 with the 8 nm Exynos 9820 is now running Fortnite at 60 FPS with better presets than TX1 on 20 nm and probably Xbox one S APU. I expect a more modern 8 nm chip to be even more powerful with the best GPU uArch available.
I know you're talking about your expectations, but isn't the rumour that the next Nintendo Switch model being capable of Xbox One S performance before DLSS only applicable to TV mode? If so, then the performance in handheld mode would be between the Nintendo Switch and the Xbox One before DLSS.
"Just"
Doing that would require a pretty massive upgrade. I don't even think the device we're talking about could do that, though maybe through DLSS.
Monolith will add their own proprietary raytracer to the Xeno engine and grind XB2 down to 540p docked on the Pro.
That's honestly much more likely than it hitting 1080p 60fps.
I'm of the belief that their renderer just needs more work. I don't see 1080p/60 out of the realm of possibility for the pro
A new SOC can probably be on par with Xbox one S in handheld mode (at 720p). A 600 GFLOPs Volta GPU should be able to do the trick. This kind of GPU only consumes 10W on Xavier NX (the whole SOC, not just the GPU). I would expect a 768 Ampere GPU to be 30~50% more performant at the same clocks (800MHz)/transistor budget while consuming less power. If the magnesium alloy rumor is real, the same chip could go beyond 10W (probably 15W ... Just like Orin S). In the worst case scenario, Orin S would have a 1.7 TFLOPs Ampere GPU while running a 1100MHz and consuming 15W if there was absolutely no gain transitioning from TSMC's 12 nm and Samsung's 8 nm. 76% of 1100MHz would give us a 1.3TFLOPs chip running at 844MHz (considering a 76% decrease between TX1 clocks on shield TV and TX1 on switch). It would consume less than 12W ... on TSMC's 12 nm. And it would still not require a magnesium alloy.
Yeah I follow Nintendo rumors pretty closely and I haven't heard any of these software rumors that are being tossed around here anywhere outside of this thread.
If you "follow the software" in regards to console releases the 3D Mario team, the Splatoon team, and the Mario Kart team could all have something this year or next year and then the Switch 2 could launch Spring 2024 (in line with the 7 year lifespan that's previously been mentioned) and have a new 3D Mario, new Mario Kart, and new Splatoon all year one. A 2D Mario can come out alongside the Mario movie if they want to syngergize a game release with it.
A Switch Pro this year would still have Breath of the Wild 2 and the rumored DK and Metroid games.
What does it hit now in docked mode? 720p dynamic and 30fps with drops? I don't see a ~2.8x better GPU and much better CPU completely fixing that. Like you said if they take more time to optimize it maybe they could but I don't think they'll bother.
I could see 1080p 30fps for sure though.
I mean offcourse they would loved to have them (big 3rd party games, because point was mainly about them) on Switch but not at any cost, they are breaking sales records even if they dont most of big 3rd party games that are on XB1/PS4.
Also, I dont think Nintendo cares to be lead platform, they want just great sales, they are certainly not upset why for instance they still dont have GTA or CoD game (IMO thats sucs) and have install base of 80m as 31.december.
They are not competing directly with Sony/MS consoles, thats why most of Switch owners have more platforms and not only Switch.
Personally dont expect more than 4 cores for this revision, but I guess we can only wait and see.
We are comparing two different architectures though in PS4's GCN vs Nvidia's Ampere, which we already have seen over the years that Nvidia's GPU's don't need the same memory bandwidth to achieve PS4 level of performance.
PS4 is only 30-35% stronger than XboxOne S, but I still expect techniques like DLSS and VRS to allow such a Switch revision to punch well above it's weight visually.
Yes in comparison to XboxOneS vs PS4, memory bandwidth was what held the XboxOne back moreso than raw GPU performance.
I think this is what it comes down to in the end and can Nvidia and Nintendo make the most technically balanced system within a specific budget.
with 2.8x gpu power you'll be hitting well over 30fps at 1080. probably within the ballpark of 45-60fps as is. with a better cpu and memory, you're removing a lot of the choke points for the game. and this is before DLSS
Onix555: Would your 4 A78 configuration retain the 4 little cores or drop them?
The fact that the Tx1 has those A53's makes me think that shrinking to 8 nm then going with 6 big/0 little should result in no major change to the area taken up by the CPU (since 4 big/4 little is approximately 5 big in that respect).
-
In general, while I'm not sure, I'm still leaning hard in the direction of the next Switch model going with A78C and more than 4 cores. And the primary reason actually has less to with Nintendo's side and more on ARM's side. That the A78C variant exists and was announced suggests to me that at least one device utilizing it will be launched no later than early 2022. And the particular niche that the A78C's changes would serve is more or less gaming* with a power budget too small for X1's. So it's down to either Nintendo or some competitor out of nowhere.
*but hey, if anybody is aware of other workloads on a low power device that would noticably benefit from having more than 4 big cores within one cluster instead of linking up two clusters, I'd love to hear of them.
The fact that the Tx1 has those A53's makes me think that shrinking to 8 nm then going with 6 big/0 little should result in no major change to the area taken up by the CPU (since 4 big/4 little is approximately 5 big in that respect).
-
In general, while I'm not sure, I'm still leaning hard in the direction of the next Switch model going with A78C and more than 4 cores. And the primary reason actually has less to with Nintendo's side and more on ARM's side. That the A78C variant exists and was announced suggests to me that at least one device utilizing it will be launched no later than early 2022. And the particular niche that the A78C's changes would serve is more or less gaming* with a power budget too small for X1's. So it's down to either Nintendo or some competitor out of nowhere.
*but hey, if anybody is aware of other workloads on a low power device that would noticably benefit from having more than 4 big cores within one cluster instead of linking up two clusters, I'd love to hear of them.
I'm not sure that dlss would work in that case as there is a floor in terms of power, but now knowledgeable people will be able to answer that.
I believe that the handheld will pack whatever the minimum specs are for dlss to be functional. Anything more than that would be gravy.
If this year's model actually does have the hardware necessary for DLSS, then I don't think calling it merely a revision would really be accurate. For all intents and purposes, that would probably be a full, next gen Switch that would probably continue to receive games well after support is dropped for the current model.
If the screen actually ends up being mini-LED, then it seems pretty likely.
If Nintendo can successfully position themselves as the lead platform without significantly compromising their vision, there's no reason for them not to do it, though. More third party games selling on the system is generally going to be a win for them.
For open world games I've found that framerate doesn't really scale well with GPU and does not scale linearly with CPU either. It takes a lot of grunt to get an open world game to 60fps.
then the bound isn't on the GPU. they're most likely hitting memory hard since I think the Xenoblade games have a lot of alpha effects, which can crater frame rate
I'm honestly not sure if Thugstas's post is referring to only the performance in TV mode.
I'm more of a marketing guy than a tech guy, so obviously better informed people can show me how I'm wrong.
This is my guess. I was always thought (as I a posted before) this revision would be VR related and, even if I look like fool, I'm still believing it, sort of...
Before I thought that a Switch Pro could be an "TV only" device with a base dock and a VR headset...now I have a different idea.
Maybe if this new upgrade is powerful enough the new Switch could still be used as a VR headset but in "handheld mode", similarly to how the VR Labo kit works.
While entertaining this idea I thought what Nintendo may need in terms of graphics capability to run current Switch games in VR. My guess is about 3 times the performance of a "base" Switch in docked mode...the aim should be to double performance and give a small increase of base resolution (while the bigger boost would be obtained with the aid of DLSS).
If this new SoC has 12SMs (as some suggested previously), taking base Tegra X1 (4SMs) as reference I would expect 1,53 TF of performance @ 1,00Ghz. Now, Z0m3le previously reported a hidden 1,267 Ghz GPU clock speed setup in the last Switch firmware. Assuming this would be the docked clock of the new device, I would expect the handheld clock to drop around 60% - 760 Mhz (which by the away is in the same ballpark as Ampere Max-Q mobile GPUs). If my rough math is correct 12SMs @760Mhz would be around a 1,18 TF performance, so around 2.5 to 3 times the performance of the current Switch while docked, without factoring DLSS.
The reason why I still think about it, is simply because it's just a too good opportunity for Nintendo to pass. Right now they are in the best position to launch mass market VR games: they have the right IPs (looking at their Mario Kart VR cabinet made me realize that) as opposed to Oculus/Facebook; they have the "customer expectations" advantage, in a sense their user base is usually not expecting hi-end graphics and realistic looking games (which we know are taxing to render in VR); lastly this is the right timing since MS/Sony have just launched their new consoles, it will take some time before they could convince their user base to buy another expensive add-on.
Plus, it's always been part of Nintendo's dna to introduce a new ways to play with new hardware...Nintendo normally is not a company that does "just" graphical upgrades (even if I know there are some exceptions). This use case would also explain the need for a lighter body and a better CPU.
This is my guess. I was always thought (as I a posted before) this revision would be VR related and, even if I look like fool, I'm still believing it, sort of...
Before I thought that a Switch Pro could be an "TV only" device with a base dock and a VR headset...now I have a different idea.
Maybe if this new upgrade is powerful enough the new Switch could still be used as a VR headset but in "handheld mode", similarly to how the VR Labo kit works.
While entertaining this idea I thought what Nintendo may need in terms of graphics capability to run current Switch games in VR. My guess is about 3 times the performance of a "base" Switch in docked mode...the aim should be to double performance and give a small increase of base resolution (while the bigger boost would be obtained with the aid of DLSS).
If this new SoC has 12SMs (as some suggested previously), taking base Tegra X1 (4SMs) as reference I would expect 1,53 TF of performance @ 1,00Ghz. Now, Z0m3le previously reported a hidden 1,267 Ghz GPU clock speed setup in the last Switch firmware. Assuming this would be the docked clock of the new device, I would expect the handheld clock to drop around 60% - 760 Mhz (which by the away is in the same ballpark as Ampere Max-Q mobile GPUs). If my rough math is correct 12SMs @760Mhz would be around a 1,18 TF performance, so around 2.5 to 3 times the performance of the current Switch while docked, without factoring DLSS.
The reason why I still think about it, is simply because it's just a too good opportunity for Nintendo to pass. Right now they are in the best position to launch mass market VR games: they have the right IPs (looking at their Mario Kart VR cabinet made me realize that) as opposed to Oculus/Facebook; they have the "customer expectations" advantage, in a sense their user base is usually not expecting hi-end graphics and realistic looking games (which we know are taxing to render in VR); lastly this is the right timing since MS/Sony have just launched their new consoles, it will take some time before they could convince their user base to buy another expensive add-on.
Plus, it's always been part of Nintendo's dna to introduce a new ways to play with new hardware...Nintendo normally is not a company that does "just" graphical upgrades (even if I know there are some exceptions). This use case would also explain the need for a lighter body and a better CPU.
Last edited:
I think what you're saying makes sense, and thanks for sharing what you have heard.
If you're wrong - that's okay. It's a fun ride.
As much as I'd like to see magnesium alloy being used for the housing, it's probably unlikely since DigiTimes is the only source that's talking about a magnesium alloy housing and DigiTimes tends to be hit or miss when it comes to Nintendo rumours.
Does anyone know the viability of Nintendo putting a Mclassic-style upscaler (A variation of the Realtek Chip in the Firmware) in the Switch 2021 itself in order to potentially act as a way to help games that don't use DLSS to upscale to 4k more easily? If not 4k, at least take it from a potential 1080p (assuming it locks most DRS games to 1080) to 1440p.
VR just seems like a dying trend to try and chase, even Sony hasn't mentioned much of anything about the next-gen PSVR. 12 Ampere SM's on 8nm in a Switch form factor would probably produce to large of an SoC that Nintendo would prefer to use. Also the Maxwell architecture that TX1 is based on hase 128 cuda cores per SM, so the Switch has 2 SMs that makes up the 256 cuda cores. 12 SMs @760Mhz would roughly equal- 2.3Tflops, which also on 8nm would probably draw way more wattage than Nintendo would like even in docked mode.
The interesting part about the Magnesium alloy rumor is that in the production of this device, this will be the first thing to leak out if there's any truth to it...
I think it's technically possible for Nintendo to integrate a chip similar to Marseille's VTV-122x chip into the next Nintendo Switch model's dock's motherboard. But whether or not Nintendo would actually integrate a chip similar to Marseille's VTV-122x chip into the next Nintendo Switch model's dock's motherboard is a different story. And Mike Heskin mentions that the Realtek chip mentioned in Aula is probably a DisplayPort 1.4 to HDMI 2.1 converter chip (e.g. RTD2173 or similar).
considering it's software-based, they wouldn't need a new chip for it
Ok, thanks for clarifying things!
Although, I don't agree VR is a dying trend on the contrary I believe we may have just scratched the surface.
The problem is, as any other new tech product (imo), VR needs to meet at least 3 qualities to become mass market:
- Tech maturity to make it viable with reasonable quality at a reasonable price point: we are almost there with Oculus Quest
- A compelling value proposition: again Oculus Quest being an all-in-one device is almost there
- A compelling application/content: PC headsets / PSVR have decent libraries but they are simply too expensive (PC) or both expensive and low quality (PSVR)
Nintendo is in the best position to combine all these 3 qualities: a reasonable compromise between price and quality, a compelling value proposition and killer apps. I belive the potential is there.
Everytime I see you post about what you've heard and theorize this device will be, I can't help but get insanely excited haha.
Sorry if it's been asked before, but assuming this beauty is coming later this year (fall perhaps), when should we expect to get some leaks about it's manufacturing?
Unknown at this time, but if I were to guess, no. Doesn't HDR add to the overall bandwidth of the GPU output signal? (Unless it's enforced by 4k HDMI standards.)
Based on what was said in the Ubisoft earnings call today, it seems plausible that more "AAA" franchise games will be made for mobile. If the Switch Pro specs can match or surpass the flagship phones, we probably will see more ports from these "AAA" franchise games. After all Switch and smartphones are brothers in ARM, amirite? *duck*
Basic 4k output really shouldn't be a problem with an improved HDMI controller in the dock. I don't think they're going to complicate it with extra image processing like that.
don't know about that tweet. GI still has the usual monitizing elements of mobile games (which is even making their way to console games). if anything, I foresee even bigger disappointments
Many of these AAA console games basically already have mobile-style monetization at this point.
I agree that there are many interesting things that could still be done with VR and maybe Switch 2 will be designed with a HMD in mind by having a much better and higher-resolution screen. Nintendo are definitely the company that would even go further out on the limb to incorporate many VR functions into their higher profile games such as Zelda and Mario, but has the public interest stayed the same as it was back when PSVR was first announced and how much of an investment would be needed on top of regular game development to make this happen.
It should be noted first off that the A53's were fused off the design of the original TX1, and remained unused by any product. Currently there is no evidence that they remain on TX1+ so they may have ditched them already.
With the propised design I feel it would be best to hold back on more then 6 cores, in order to save die space for the larger GPU and memory bus. Due to the failings of Samsung, Nvidia is left with a more restrictive space budget so it has to be designed carefully.
I personally think that 4 A78's would be enough for a Pro, especially if they were run a 2Ghz; in which case you'd get a 6x jump from the TX1. Which is pretty good and similar to the N3DS jump.
The A78C is very interesting. However I believe it actually ties into a Qualcomm project to push further into the laptop market, specifically the 8cx successor.
- Status