• Hey everyone, staff have documented a list of banned content and subject matter that we feel are not consistent with site values, and don't make sense to host discussion of on Famiboards. This list (and the relevant reasoning per item) is viewable here.
  • Do you have audio editing experience and want to help out with the Famiboards Discussion Club Podcast? If so, we're looking for help and would love to have you on the team! Just let us know in the Podcast Thread if you are interested!

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

Lapsus-Ransom-NVIDIA.png

These people will be in jail by Friday.
 
NVN2 has added lots of documentation relating to ray tracing, so that's confirmed for the next model, as well as relating to DLSS. You know, just in case anybody was still wondering whether that's confirmed.
I may be asking too much but..........Where are you seeing this? I haven't seen any new leaked data so far.
 
Saying that, I wonder how much more information is going to be unveiled about Drake on Friday, assuming Nvidia's not going to capitulate to the hacker's demands.
Well they say 250GB of Hardware info is in there so Drake's exact specs (As of when the data was pulled) likely are in there too.
 
0
I've been wondering if Nintendo has been seeing the silicon shortage as a reason to delay Switch 2 and maybe that gave them and Nvidia time to migrate Dane to a new process node, and with that may have prompted a codename change. Possibly taking the process node change in balance to increase the number of cores and/or clockspeed for the CPU and/or GPU.

On the same note, with Dane all but done, that could still make it into a product like the Shield.
 
I've been wondering if Nintendo has been seeing the silicon shortage as a reason to delay Switch 2 and maybe that gave them and Nvidia time to migrate Dane to a new process node, and with that may have prompted a codename change. Possibly taking the process node change in balance to increase the number of cores and/or clockspeed for the CPU and/or GPU.

On the same note, with Dane all but done, that could still make it into a product like the Shield.
I think that depends on when Nintendo and Nvidia decide to change the process node being used for the fabrication of Drake since that's not a process that's done in a matter of months, but in a matter of years (a minimum of half a year or 6 months). (One example is that GH100 was mentioned to be fabricated using TSMC's N5 process node almost 2 years ago.)
 
I've been wondering if Nintendo has been seeing the silicon shortage as a reason to delay Switch 2 and maybe that gave them and Nvidia time to migrate Dane to a new process node, and with that may have prompted a codename change. Possibly taking the process node change in balance to increase the number of cores and/or clockspeed for the CPU and/or GPU.

On the same note, with Dane all but done, that could still make it into a product like the Shield.
The T239 part hasn't changed, and that's the chip's official designation. It's unlikely to have changed, it was just probably always Drake and Kopite got that wrong for whatever reason.
 
I may be asking too much but..........Where are you seeing this? I haven't seen any new leaked data so far.
These are all things people are finding in the leak. There's 75 GB of data, so the handful of screencaps on Twitter is certainly not all of it.

The T239 part hasn't changed, and that's the chip's official designation. It's unlikely to have changed, it was just probably always Drake and Kopite got that wrong for whatever reason.
One interesting thing I saw was that, out of several lists of forbidden words (that are meant to be scanned for before something is released to the public) which had "drake" on them, one of them also had "dane." So that name was considered sensitive for some reason at some point.
 
One interesting thing I saw was that, out of several lists of forbidden words (that are meant to be scanned for before something is released to the public) which had "drake" on them, one of them also had "dane." So that name was considered sensitive for some reason at some point.
Well now I'm confused all over again.
 
I think that depends on when Nintendo and Nvidia decide to change process node since that's not a process that's done in a matter of months, but in a matter of years. (One example is that GH100 was mentioned to be fabricated using TSMC's N5 process node almost 2 years ago.)
Exactly. Chip shortage was pretty evident in mid-2020. Let's say that they're level headed and by late 2020 made a decision:

1. Split the current development effort into releasing the OLED that was being saved for Switch 2 as an upgraded Switch 1 in 2021.
1a. Took 239 and started the work to port it to a smaller node and increase the GPU core count by 2.
2. Release the OLED Switch in late 2021.
3. Plan to release Switch 2 late 2022 or 6 year anniversary or Late 2023 depending on a number of factors.

This way they get two things.

1. They get the cost benefits of moving to OLED.
2. If there was some unhappiness about what hardware fit into the power envelope, they get a chance to adjust it.

It may have been that they were worried about an anemic and launch of Switch 2 anyway.

It just seems like a whole jumble of costs and benefits to me, but I feel like it's worth thinking about.
 
Well now I'm confused all over again.
The codename for Dane leaked early last year, them going with a new one doesn't surprise me. It could have also been a change to some element of the chip. GA10F is another name for the chip that we know now, considering the alphabet, it is interesting they went all the way to F for the naming of this SoC.
 
Exactly. Chip shortage was pretty evident in mid-2020. Let's say that they're level headed and by late 2020 made a decision:

1. Split the current development effort into releasing the OLED that was being saved for Switch 2 as an upgraded Switch 1 in 2021.
1a. Took 239 and started the work to port it to a smaller node and increase the GPU core count by 2.
2. Release the OLED Switch in late 2021.
3. Plan to release Switch 2 late 2022 or 6 year anniversary or Late 2023 depending on a number of factors.

This way they get two things.

1. They get the cost benefits of moving to OLED.
2. If there was some unhappiness about what hardware fit into the power envelope, they get a chance to adjust it.

It may have been that they were worried about an anemic and launch of Switch 2 anyway.

It just seems like a whole jumble of costs and benefits to me, but I feel like it's worth thinking about.
This doesn't add up because Aula was datamined in April 2020, meaning in the works long before the pandemic hit.
 


Its interesting because the post speaks about FSR and how this tecnology bring a 1080p on the resolution quality of Life is Strange 3
 
This doesn't add up because Aula was datamined in April 2020, meaning in the works long before the pandemic hit.
I thought this leaked today? I don't remember Drake from 2020. Maybe the catalyst was earlier. Maybe it was going to be Aula and Dane side-by-side, but Dane got delayed. Maybe Aula was originally just a Switch Lite revision 2DS Style.

Please remember that I'm not making claims, but instead exploring a line of reasoning. I'm not trying to prognosticate.
 
considering it's not there now, I doubt enabling support is a thing on their minds
I was talking about theoretically speaking rather in actually speaking. But yeah, I personally don't expect Nintendo to enable HDR support for the OLED model anytime soon.
 
So that means Nintendo could enable HDR support for the OLED model in the future?
It's not clear to me from what's been shown whether NVN2 is supported on pre-Ampere GPUs, but I'm leaning toward no. The language around GPU/arch support is all related to the Windows implementation (i.e. things to be aware of since your development PC might not behave exactly like the NX hardware you're actually targeting), and I haven't yet seen anywhere that's like "this model of NX can do this, but this other one can't." So that implies the NVN2 documentation is only concerned with one NX model, which would have to be the new one since the old one (as far as we know) can't do DLSS and ray tracing.
 
It's not clear to me from what's been shown whether NVN2 is supported on pre-Ampere GPUs, but I'm leaning toward no. The language around GPU/arch support is all related to the Windows implementation (i.e. things to be aware of since your development PC might not behave exactly like the NX hardware you're actually targeting), and I haven't yet seen anywhere that's like "this model of NX can do this, but this other one can't." So that implies the NVN2 documentation is only concerned with one NX model, which would have to be the new one since the old one (as far as we know) can't do DLSS and ray tracing.
Thanks for the clarification. That does make sense.
 
0
Don't think about "revision" or "successor" think about "cross-gen support" as a spectrum.

On one end of the spectrum, all games released going forward run on both devices, and when support stops for the Old Switch it stops for the Nu Switch at the same time (close to New 3DS, or DSi level revision).

On the other end of the spectrum, there are a few cross-gen games, but they are different versions that you have to buy seperately, and all first party dev moves to the "new" machine (the WiiU).

There are lots of options in between. Like, for example, lots of cross-gen games for a few years, but eventually all games are exclusive to the Nu Switch which slowly takes over for a "generation" that runs like 12 years - the Game Boy Color model.

Lots of folk saying "Successor" or "revision" is a marketing term, but if the device is backwards compatible, then it's as much about "how does first party support shift" which is a decision that Nintendo can change* after the launch of the device.

*I recognize that it is really hard to turn that ship, just that Nintendo can retool if things go wrong. Just as they kept 3DS on life support until they were sure Switch took off, or GBA until the DS was solid
Thank you for that explanation, honestly I agree entirely and I just never knew how to put it in words where it made sense haha. A cross-gen spectrum absolutely makes the most sense for how the Switch and any future iterations would be supported.

So that means Nintendo could enable HDR support for the OLED model in the future?
Even if HDR isn't implemented on the OLED for gaming, is it at all possible if streaming platforms like Netflix or Disney+ get added in the future, HDR could be enabled for OLED consoles via firmware updates given their docks also support HDMI 2.0?
 
Truly don't know, and honestly something I would wonder too. It seems AMD releases these Zen chipsets every year or every other year. The Nvidia SoC that people reference, I was under the impression come out every 7 or 8 years, do they not?
It's been a while since Nvidia produced an SoC that would be appropriate for a game console, but they have produced a few between TX1 and Orin. It's probably mostly a function of most of Nvidia's non-Nintendo SoC customers having pretty specialized needs.
One interesting thing I saw was that, out of several lists of forbidden words (that are meant to be scanned for before something is released to the public) which had "drake" on them, one of them also had "dane." So that name was considered sensitive for some reason at some point.
Hard to say what this means without knowing how Nvidia treats codenames internally.
 
I thought this leaked today? I don't remember Drake from 2020. Maybe the catalyst was earlier. Maybe it was going to be Aula and Dane side-by-side, but Dane got delayed. Maybe Aula was originally just a Switch Lite revision 2DS Style.

Please remember that I'm not making claims, but instead exploring a line of reasoning. I'm not trying to prognosticate.
I'm talking about Aula, not Drake. Aula was datamined in 2020 so the OLED model (which is Aula) was planned far before the pandemic hit. And everything in that original datamined remained accurate when the OLED model launched, so plans for the OLED model did not change at all after the onset of the pandemic and the chip shortage.

So the theory that the OLED model was some stopgap or branch thrown together due to the pandemic cannot be accurate with what we know.


Dane/Drake was likely never planned for 2021 since Orin, the chip which Dane/Drake is a cut down version of, was not set to release until 2022. This roadmap placing Orin in 2022 was from 2019, before the pandemic hit.
 
Even if HDR isn't implemented on the OLED for gaming, is it at all possible if streaming platforms like Netflix or Disney+ get added in the future, HDR could be enabled for OLED consoles via firmware updates given their docks also support HDMI 2.0?
I think theoretically speaking, yes. But whether or not Nintendo would actually enable HDR support for media is another story altogether.
 
0
Had a whole thing about this typed up but it was all speculation and rambling. So instead I'll just say (a) the NVN2 documentation has a new section on enabling G-Sync, and (b) despite that, the contents of that section and certain other things make me question whether G-Sync is really a feature the NX can use with NVN2.

It's a bit unfortunate that most of the people who already know about NVN and the Switch's internals really can't/shouldn't be looking at the leaked materials, since doing so while participating in reverse engineering efforts on a system the leak is about opens you up to serious liability. But I'd love if somebody who knows more about this stuff could see it firsthand and interpret it. Like I said, I can speculate all day about how the stuff in the leak does or doesn't suggest G-Sync as a real feature, but it's barely-educated guessing.
 
You hate to see this news come out this way but I remember stating some time ago that I feared hackers leaking the information before Nintendo even gets a chance to reveal it themselves. Sure these people getting this information is more detrimental to Nvidia, but the byproduct will be that anything surrounding the next Switch hardware will be what creates headlines...

As someone mentioned above already, I don't see Nvidia giving into any demands and someone will be in jail soon behind this (or given a high paying IT job) you never can tell how things will pan out in this capitalism over everything society. 🤔
 
0
I don't fault him on the codename mix up at all, but the ? By the process node is the most interesting out of all of this.
Simply because he was pretty assertive that all the Tegra derivatives would be on 8nm as well...
I think the question marks are just referring to things this leak has not yet confirmed.
 
0
I'm going to again preface this by saying I have a shallow understanding of the things I'm seeing at this point. But this certainly seems relevant to the thread.

NVN v1 source has a constants file with the following:

C:
// Number of warps per SM on TX1 hardware
#define __NVN_NUM_WARPS_PER_SM_TX1                  128

// Number of SMs on TX1 hardware
#define __NVN_NUM_SMS_TX1                           2

NVN v2 source replaces them with this:

C:
// Number of warps per SM on ga10f
#define __NVN_NUM_WARPS_PER_SM_GA10F                48

// Number of SMs on on ga10f
#define __NVN_NUM_SMS_GA10F                         12
 
I'm going to again preface this by saying I have a shallow understanding of the things I'm seeing at this point. But this certainly seems relevant to the thread.

NVN v1 source has a constants file with the following:

C:
// Number of warps per SM on TX1 hardware
#define __NVN_NUM_WARPS_PER_SM_TX1                  128

// Number of SMs on TX1 hardware
#define __NVN_NUM_SMS_TX1                           2

NVN v2 source replaces them with this:

C:
// Number of warps per SM on ga10f
#define __NVN_NUM_WARPS_PER_SM_GA10F                48

// Number of SMs on on ga10f
#define __NVN_NUM_SMS_GA10F                         12
Dangit, so close to the GPU core count!

Btw, link to the files? XD

EDIT: Stupid Mobile, umm..is the latter Core count a typo? The Number of SMs seems to line up right (12SM Drake!?) But the"Wraps" is off
 
Last edited:
Quoted by: LiC
1
Dangit, so close to the GPU core count!

Btw, link to the files? XD

EDIT: Stupid Mobile, umm..is the latter Core count a typo? The Number of SMs seems to line up right (12SM Drake!?) But the"Wraps" is off
Those are the numbers. I can't speak to any implications.

I have seen enough to say now that there's no chance of NVN2 running on an old Switch, regardless of feature set. Even on Windows the oldest GPUs supported are Turing (with the note already posted earlier that says Ampere is preferred because it's most compatible with the real NX). Anything older is unsupported in the sense of "it bails/crashes/doesn't compile." Not really surprising, but just since it was asked earlier about HDR in NVN.
 



This is also one of those "this is probably something, just don't ask me what" posts. Kopite was replying to the first account, so figure it's got some credibility. And that account replied to the second tweet...
 
12 SMs is 6 times Switch power right? In RAW performance and same clocks (400GFLOPs vs 2400 GFLOPS) on docked mode.

And in portable is 200 GFLOPS vs 1200 GFLOPS

My expectation is half of that but a little higher clock.
 
So I'm just learning this terminology - SM means streaming multiprocessors, when I look up how many SMs the Tegra X1 has it's 2, which lines up with the code above. So the new chip has... 12? I am assuming there is no difference in SM count between devkit and final hardware, so that seems like quite a lot more GPU power.
 
12 SMs is 6 times Switch power right? In RAW performance and same clocks (400GFLOPs vs 2400 GFLOPS) on docked mode.

And in portable is 200 GFLOPS vs 1200 GFLOPS

My expectation is half of that but a little higher clock.
You can not compare FLOPs across uArchs directly.

12SM Dane at 10-15W Docked would most likely be running right into that gap between the PS4 Pro and Series S tbh
 
I'm going to again preface this by saying I have a shallow understanding of the things I'm seeing at this point. But this certainly seems relevant to the thread.

NVN v1 source has a constants file with the following:

C:
// Number of warps per SM on TX1 hardware
#define __NVN_NUM_WARPS_PER_SM_TX1                  128

// Number of SMs on TX1 hardware
#define __NVN_NUM_SMS_TX1                           2

NVN v2 source replaces them with this:

C:
// Number of warps per SM on ga10f
#define __NVN_NUM_WARPS_PER_SM_GA10F                48

// Number of SMs on on ga10f
#define __NVN_NUM_SMS_GA10F                         12
That 48 seems... weird. If that's the number of shaders in the SM as implied, that seems weirdly low.
Those are the numbers. I can't speak to any implications.

I have seen enough to say now that there's no chance of NVN2 running on an old Switch, regardless of feature set. Even on Windows the oldest GPUs supported are Turing (with the note already posted earlier that says Ampere is preferred because it's most compatible with the real NX). Anything older is unsupported in the sense of "it bails/crashes/doesn't compile." Not really surprising, but just since it was asked earlier about HDR in NVN.
Yeah, it sounds like this will be the API that native Dane/Drake games will use.
 
Quoted by: LiC
1
So I'm just learning this terminology - SM means streaming multiprocessors, when I look up how many SMs the Tegra X1 has it's 2, which lines up with the code above. So the new chip has... 12? I am assuming there is no difference in SM count between devkit and final hardware, so that seems like quite a lot more GPU power.
The caveat here is that SMs have different configurations in different architectures. So 1 Maxwell SM will have a different number of CUDA cores then 1 Ampere SM.
 
The caveat here is that SMs have different configurations in different architectures. So 1 Maxwell SM will have a different number of CUDA cores then 1 Ampere SM.
Yeah but we already know the amount of CUDA in a Ampere/Orin SM, 128 cores.

So 12SMs of that would be 1536 CUDA cores.

If GA10F inherits the Cache boost from GA10B (Orin) then 12SM Drake would likely be running right near the Laptop 3050 before DLSS in raster

And that is before we consider Drake in any config likely will have way better memory management due to having 8GBs versus 4 at minimum (as we see with the Steam Deck versus RX6500XT)
 
The caveat here is that SMs have different configurations in different architectures. So 1 Maxwell SM will have a different number of CUDA cores then 1 Ampere SM.
Yeah, the thing is, if you do the math with those numbers, the chip would only have 576 CUDA cores, which is only a bit more than double the current Switch's 256.

Unless the warp count doesn't directly correspond to CUDA cores. I'll admit this is pretty out of my depth on GPU architecture.
 
Please read this staff post before posting.

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


Back
Top Bottom