• 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!)

Or like the last time they pretended it was a new generation and it was, internally spec-wise, just a GameCube pro 😝.

Nintendo has options on what they can do with this. Though as for the other convos going on about it, a platform can have a next generation device without having a clean break of the platform.
yes. This.

Are you on the right link? This is what it's showing me right now.

Shipped and sold by Amazon. Plenty in stock.

4t0DF4q.png
Same for me out west ... though stores are a bit more scarce... its still mostly in stock at best buy and such




samsung 5nm would be awesome ... but I guess for me ... its 8nm until it isn't
 
Last edited:
0
Saw a place in NVN2 that defines the L2 cache as 1 MB, not 4 MB. There were several other sources for 4 MB, but it's possible those were places where GA10F wasn't sufficiently differentiated from GA10B.

As a reminder, the things I've been posting shouldn't be taken as concrete evidence of anything beyond the literal fact "the source code for a driver/tool has a constant that says X spec has Y value." I have a decent idea whether something is relevant to T239/GA10F and NVN or not, but beyond that, who knows if values could have changed, or were dummy values, or simulation-only, or what.
 
Saw a place in NVN2 that defines the L2 cache as 1 MB, not 4 MB. There were several other sources for 4 MB, but it's possible those were places where GA10F wasn't sufficiently differentiated from GA10B.

As a reminder, the things I've been posting shouldn't be taken as concrete evidence of anything beyond the literal fact "the source code for a driver/tool has a constant that says X spec has Y value." I have a decent idea whether something is relevant to T239/GA10F and NVN or not, but beyond that, who knows if values could have changed, or were dummy values, or simulation-only, or what.
Thank you for being so clear and candid with your findings.
 
I am starved for information. Another leak, a Nate podcast, Bloomberg article....anything.

I guess the good news is that if this device is coming out for holiday of this year, we'll know about in a few months even without an official Nintendo announcement. Bloomberg or someone else will get wind of things heating up on the manufacturing side of things.
 
Oldpuck! I'm back!
Oh hey, Thread Lurker. How you doing?

Pissed off. At you.
What? Why? Last time we talked you seemed pretty jazzed.

That was when you told me that the Omega Switch Would Be The Best Graphics Evar!
I... I did not actually say that

But now I hear that The Oemga Switch-
Don't call it that, please don't let that catch on

-that the Omega Switch won't be backwards compatible!!!!
Oh gods, not this again. You know what? Don't worry about it

What? So it WILL be backwards compatible???
I said don't worry about it. There isn't anything to worry about. Chill. It's cool. Promise. Your games are going to work

So MVG is wrong?
I didn't actually say that.

MVG IS RIGHT? MY GAAAAAAAMES!
Okay, you know what? I'm going to explain it as best I can but here is the deal. You can only read farther if you don't freak out.

I can't promise anything, but I will try
Okay. Big deep breath

The Switch is built on an Nvidia architecture called Maxwell. The Omega Switch -

It's catching on!
Hush! The Omega Switch, according to this leak, is built on an Nvidia architecture called Ampere. Well before this leak, we've known backwards compatibility for an Ampere console would be an issue.

But-
Stop freaking out. Breath.

Okay, okay. I'm listening
There almost definitely will be a backwards compatibility solution so good you'll never notice. There is a decent chance that MVG is right and the solution isn't compatible hardware but a software solution. That software solution is definitely going to be automated in some way, and that every game you care about will work.

Maybe we'll see a patch like MVG describes. Xbox does this for some games, and Xbox is the gold standard for preserving BC across hardware. There is also a possible hardware solution, that doesn't require patching, but also is probably not perfect-forever-and-always BC. It's just BC so good you don't care about the problems.

Even the Game Boy Color didn't exactly replicate what the Game Boy hardware could do. The processor was slightly faster, but that's not something that ever mattered to a video game

Games are all I care about it
Exactly. Perfect BC isn't the same as "BC so good..."

So good you'll never notice, right.
Besides you don't want perfect backwards compatibility do you?

What? Yes I do!
Perfect BC means no improvements in old games. If you want "perfect" BC you basically need to shove the entire Classic Switch hardware into the Omega Switch's case, and run old games on there. That means no framerate improvements, no resolution increases, nothing. It also means the hardware is hella expensive.

What you want is to play all your old games, with no major new bugs, and better graphics

The Best Graphics Evar!!!!
Right. So what does this mean

It means don't worry about it?
It means don't worry about it.
 
It means don't worry about it?
It means don't worry about it.
Simple, elegant, and perfectly concluded. “Don’t worry about it”

You’ll still get great games person who lurks ;).

Saw a place in NVN2 that defines the L2 cache as 1 MB, not 4 MB. There were several other sources for 4 MB, but it's possible those were places where GA10F wasn't sufficiently differentiated from GA10B.

As a reminder, the things I've been posting shouldn't be taken as concrete evidence of anything beyond the literal fact "the source code for a driver/tool has a constant that says X spec has Y value." I have a decent idea whether something is relevant to T239/GA10F and NVN or not, but beyond that, who knows if values could have changed, or were dummy values, or simulation-only, or what.
That’s a bit disappointing but it’s still pretty potent of a device.

If it had 4MB, it would sit comfortably for a long time.
 
0
I think it would be a good idea for Nvidia to do what they can to get Nintendo on the best node possible to help promote their brand. Do you guys think they would be willing to foot a large amount of the bill to get Nintendo on 5nm? I think the media hype behind the performance of the Switch 4K would do good for their stock price and brand relevance. I've heard the word Tegra so many times because of the Switch lol. An even more powerful switch upgrade than expected will allow DLSS to be more likely to be used and marketed. But i don't how big that bill will have to be and if the tradeoff is worth it.

I wondered something like this awhile back and thought maybe we could see a return of the licensed logo on the box and device from Nvidia (similar to what ATi did back in the day with GameCube). Yes Nvidia are the #1 GPU provider in the industry, but having your logo on the one of the most popular gaming devices in the world can surely only help with keeping things that way.
Also as others stated with getting Nvidia specific technologies adopted, the Switch could be a Trojan horse for that to in the future...
5e776a9eaa2eb10030753b11
 
Saw a place in NVN2 that defines the L2 cache as 1 MB, not 4 MB. There were several other sources for 4 MB, but it's possible those were places where GA10F wasn't sufficiently differentiated from GA10B.

The more we learn form the leak, the worse the specs sound after the initial hype.

Not "worse" as in actualy bad - but it's obvious that we don't have a full picture and the hardware might be less powerful than the PS4 Pro expectations we read about here sometimes.
 
The more we learn form the leak, the worse the specs sound after the initial hype.

Not "worse" as in actualy bad - but it's obvious that we don't have a full picture and the hardware might be less powerful than the PS4 Pro expectations we read about here sometimes.
This is just one place with others mentioning the 4.. I wouldn't call this definitive for anything.
 
The more we learn form the leak, the worse the specs sound after the initial hype.

Not "worse" as in actualy bad - but it's obvious that we don't have a full picture and the hardware might be less powerful than the PS4 Pro expectations we read about here sometimes.

Keeping expectations in check is always a good thing, and we have no idea what the final chip will really be like. But I don't understand this post, as this is the first time I've had to correct anything after initially seeing it in the leak, and it's a minor thing at that.
 
Drake is the answer, Nvidia secured TSMC 5nm not that long ago, and in the leak we can see that the codename changed from Dane to Drake, the confirmation of this is that Dane is a banned search word (at least this is what I remember from last week). There is nothing else Nvidia has Samsung's 5nm node as a use for, and offering the node to Nintendo, also solves die size mystery that we have been working with, as this could literally be half the size of 8nm die, and would also solve how this chip is so big and still able to be portable, as we have yet to see any evidence of Thraktor's working theory on disabling SMs, (there is nothing listed in the document for the GPU other than 12SM, which is odd if the portable mode does disable them, that is a huge issue for developers trying to make games for the platform).

Anyways, again this is a theory, but I do think that it is correct, I was 50:50 on a die shrink, but now I favor it, I think Nvidia had the same issue here as when they moved from 20nm Maxwell desktop line (that never happened) to 16nm finfet, and offered the node to Nintendo (who they had worked with as early as 2014), this avoids a contract breach worth potential 100s of millions of dollars, just like it did then.
I'm thinking the Switch OLED screen and dock were originally meant for Dane but Dane went back to the drawing board and became Drake. Explains why everything about the Switch OLED is glaringly more advanced than the SOC as well. Also jibes with what brainchild might have been alluding to about a hypothetical SOC redesign. It does fit.
 
This was always going to be to the PS4 Pro what the XB1 was to the PS4 in difference of performance.

Not literally a PS4 Pro docked.
 
0
Saw a place in NVN2 that defines the L2 cache as 1 MB, not 4 MB. There were several other sources for 4 MB, but it's possible those were places where GA10F wasn't sufficiently differentiated from GA10B.

As a reminder, the things I've been posting shouldn't be taken as concrete evidence of anything beyond the literal fact "the source code for a driver/tool has a constant that says X spec has Y value." I have a decent idea whether something is relevant to T239/GA10F and NVN or not, but beyond that, who knows if values could have changed, or were dummy values, or simulation-only, or what.
Honestly think that it just has the 4MB of L2 from Orin still and that if that 1MB L2 mention is something that isn't a typo it's something like 1MB of L2 reserved for the OS for things like the recording functionality
 
Quoted by: LiC
1
Keeping expectations in check is always a good thing, and we have no idea what the final chip will really be like. But I don't understand this post, as this is the first time I've had to correct anything after initially seeing it in the leak, and it's a minor thing at that.
I agree and it's not meant as a complaint, I just feel that some users are getting a bit too excited about the power of this and I still believe it won't be a leap to Series S level.

At the same time I also don't think that's necessary a sling as it's a balanced machine that works well for devs that want to port cross gen titles.
 
0
Honestly think that it just has the 4MB of L2 from Orin still and that if that 1MB L2 mention is something that isn't a typo it's something like 1MB of L2 reserved for the OS for things like the recording functionality
It's not a typo, and it's not a reserved portion. It's the "total size of L2 cache in the GPU core."

Anyone can go see these sources for themselves and draw some conclusions, but as it stands I would say "Drake has either 1 MB or 4 MB of L2."
 
It's not a typo, and it's not a reserved portion. It's the "total size of L2 cache in the GPU core."

Anyone can go see these sources for themselves and draw some conclusions, but as it stands I would say "Drake has either 1 MB or 4 MB of L2."
Still find it weird, as that would reduce the L2 below even Ampere.
Which is super weird and would kill the GPU's Efficiency especially at 12SMs
 
0
The formula is just 2 x clock x number of CUDA cores. Drake has 12 SM with 128 cores each = 1536 total.

Twice that is 3,072. So, for Drake, just multiply the clock by 3,072 (or just 3k for a quick rough number, like 400MHz => 1.2 TF, 1GHz => 3 TF, 1.2GHz => 3.6 TF).
It's a bit depressing to realize that these exercises are ultimately futile for now since we don't know what consumption Nintendo will go for for the succ. 15W portable ? 15W docked? 5W portable and 20W docked? Even after we plot the consumption curve of the RAM and the CPU candidates, we will have no base reference to compare our data with.

Also, will it be a portable at all? How efficient will the Bluetooth module be? How much energy will the screen sip? There are so many unknowns...
 
It's a bit depressing to realize that these exercises are ultimately futile for now since we don't know what consumption Nintendo will go for for the succ. 15W portable ? 15W docked? 5W portable and 20W docked? Even after we plot the consumption curve of the RAM and the CPU candidates, we will have no base reference to compare our data with.

Also, will it be a portable at all? How efficient will the Bluetooth module be? How much energy will the screen sip? There are so many unknowns...
I think a reasonable assumption is that it will be a hybrid (cause everything else would be dumb, business wis), and it will use the same screen as the switch oled (if it ain’t broke, don’t fix it).
 
I think a reasonable assumption is that it will be a hybrid (cause everything else would be dumb, business wis), and it will use the same screen as the switch oled (if it ain’t broke, don’t fix it).

So there is one thought I had, just thinking outside the box for a moment, what if the device they're releasing in the next year or so is a Switch Home, TV-only console powered by Drake on 8nm, and then in a couple of years they release Switch 2, a hybrid device with a die-shrunk version of Drake on 5nm or whatever? So the Switch Home is basically the TV-only version of Switch 2, it just comes out a couple of years before Switch 2 does. It makes the marketing a bit easier, where Switch Home is very clearly a supplementary device rather than a new generation, and it largely plays the same library but with some exclusives, and then when Switch 2 comes around they just have to say that you can play all your Switch 2 games on Switch Home.

The awkward part would probably be that all the games made in the years between Switch Home and Switch 2 releasing would be missing a performance profile: Drake handheld. Therefore while they'd work on Switch 2, and look great in docked mode, they'd probably have to revert to the base Switch handheld profile, which would mean a huge gap in quality between the two. Not sure if there's an easy way around that other than forcing all developers to support a completely unused profile in the interim.

Anyway, I'd still be very surprised if it's anything other than a hybrid, but with the leak shaking things up I figured I'd share ones of my more out-there thoughts.

Edit: Ok, so for the issue of no Drake handheld profile, they could set up Switch Home so that when it's connected to a 1080p display it automatically clocks down to the planned Drake handheld clocks, and they just call it 1080p mode and claim it's to save power. I'm sure it would piss off devs to no end, and probably some users too, who would prefer to run at 4K and get the benefits of downscaling, but it would force developers to ensure every game would end up being compatible with Switch 2 handheld mode.
 
Last edited:
So there is one thought I had, just thinking outside the box for a moment, what if the device they're releasing in the next year or so is a Switch Home, TV-only console powered by Drake on 8nm, and then in a couple of years they release Switch 2, a hybrid device with a die-shrunk version of Drake on 5nm or whatever? So the Switch Home is basically the TV-only version of Switch 2, it just comes out a couple of years before Switch 2 does. It makes the marketing a bit easier, where Switch Home is very clearly a supplementary device rather than a new generation, and it largely plays the same library but with some exclusives, and then when Switch 2 comes around they just have to say that you can play all your Switch 2 games on Switch Home.

The awkward part would probably be that all the games made in the years between Switch Home and Switch 2 releasing would be missing a performance profile: Drake handheld. Therefore while they'd work on Switch 2, and look great in docked mode, they'd probably have to revert to the base Switch handheld profile, which would mean a huge gap in quality between the two. Not sure if there's an easy way around that other than forcing all developers to support a completely unused profile in the interim.

Anyway, I'd still be very surprised if it's anything other than a hybrid, but with the leak shaking things up I figured I'd share ones of my more out-there thoughts.

Edit: Ok, so for the issue of no Drake handheld profile, they could set up Switch Home so that when it's connected to a 1080p display it automatically clocks down to the planned Drake handheld clocks, and they just call it 1080p mode and claim it's to save power. I'm sure it would piss off devs to no end, and probably some users too, who would prefer to run at 4K and get the benefits of downscaling, but it would force developers to ensure every game would end up being compatible with Switch 2 handheld mode.
After the switch lite there is no reason to split the modes. The public has spoken and they want a system that can do both. A docked only mode is a colossal waste of everything
 
Saw a place in NVN2 that defines the L2 cache as 1 MB, not 4 MB. There were several other sources for 4 MB, but it's possible those were places where GA10F wasn't sufficiently differentiated from GA10B.

As a reminder, the things I've been posting shouldn't be taken as concrete evidence of anything beyond the literal fact "the source code for a driver/tool has a constant that says X spec has Y value." I have a decent idea whether something is relevant to T239/GA10F and NVN or not, but beyond that, who knows if values could have changed, or were dummy values, or simulation-only, or what.
Question, how do these constants get used in the driver? Is it something that would cause obvious bottlenecks/issues in performance if say, if one part of the driver things only 1 MB exists vs 4? Mind you, I don't know how aware drivers have to be about cache size and such.

While NVN is the main API used to interact with the Switch graphics hardware, there are games that use OpenGL and Vulkan. Are there any OpenGL/Vulkan implementations for the T239 hardware?
 
Question, how do these constants get used in the driver? Is it something that would cause obvious bottlenecks/issues in performance if say, if one part of the driver things only 1 MB exists vs 4? Mind you, I don't know how aware drivers have to be about cache size and such.

While NVN is the main API used to interact with the Switch graphics hardware, there are games that use OpenGL and Vulkan. Are there any OpenGL/Vulkan implementations for the T239 hardware?
Even if it's not listed Vulkan and OpenGL will be fully compatible.
 
It's a bit depressing to realize that these exercises are ultimately futile for now since we don't know what consumption Nintendo will go for for the succ. 15W portable ? 15W docked? 5W portable and 20W docked? Even after we plot the consumption curve of the RAM and the CPU candidates, we will have no base reference to compare our data with.

Also, will it be a portable at all? How efficient will the Bluetooth module be? How much energy will the screen sip? There are so many unknowns...
Exactly. Performance is quite a lot about TDP + used manufacturing node (when architecture is decent). My guess is 15W SoC TDP docked 5W (max.) portable. With TSMC N5 / N5P / N4 that would probably mean about 2 TFLOPS (just for some comparison).

So close to rumoured PS4 level. But with DLSS (later 3.0?) the preceivable gaming performance would be closer to that of PS4 Pro / Xbox Series S.

They could go with 25W / 30W SoC, whitch would be great, but that requires serious active, forced cooling from the dock.
 
So the Switch Home is basically the TV-only version of Switch 2, it just comes out a couple of years before Switch 2 does. It makes the marketing a bit easier, where Switch Home is very clearly a supplementary device rather than a new generation
I'd say it makes the marketing a lot harder. You want it to be a complementary premium device to the old generation, until it becomes a cut-down version of the new generation. That's completely reversing the positioning of the hardware.

Then they have to minimize the damage made to their main product appeal, which isn't the best way to play Nintendo games anymore, and to the next generation, which will have 2 years old specs rather than being hyped over.

And that's not mention smaller issues like the Home not having new controller features.
 
I was wondering more if the Vulkan/OpenGL implementations would collaborate the same info involving the GPU specs.
nah, I wouldn't think so. there's nothing about Vulkan/OpenGL that would necessitate specific things like cache sizes and whatnot. specific versions of Vulkan would point us in every direction from Turing, to Ampere, to RDNA2, to Adreno, to Arc
 
What if Drake is the SoC Nvidia is creating for Mediatek? Probably why is more powerful than Dane, Made for laptops with better cooking while Dane is a cut down version of it?
 
What if Drake is the SoC Nvidia is creating for Mediatek? Probably why is more powerful than Dane, Made for laptops with better cooking while Dane is a cut down version of it?
There wouldn't be references to a chip made for Mediatek in an API Nvidia created for a Nintendo console.
 
nah, I wouldn't think so. there's nothing about Vulkan/OpenGL that would necessitate specific things like cache sizes and whatnot. specific versions of Vulkan would point us in every direction from Turing, to Ampere, to RDNA2, to Adreno, to Arc
I see.

So no way to confirm what's the deal with the different L2$ values mentioned in NVN2.
 
0
Question, how do these constants get used in the driver? Is it something that would cause obvious bottlenecks/issues in performance if say, if one part of the driver things only 1 MB exists vs 4? Mind you, I don't know how aware drivers have to be about cache size and such.

While NVN is the main API used to interact with the Switch graphics hardware, there are games that use OpenGL and Vulkan. Are there any OpenGL/Vulkan implementations for the T239 hardware?
I personally (a) don't have the experience/context to talk confidently about what the GPU driver is doing, and (b) wouldn't want to discuss specific details of the proprietary code anyway.

As far as conflicts go, there does seem to be a place that validates at runtime whether various constants defined by NVN match the actual values. If the defines I saw before really were for GA10F, then this would seem to conflict, and one of the two definitions would need to be updated.
 
So there is one thought I had, just thinking outside the box for a moment, what if the device they're releasing in the next year or so is a Switch Home, TV-only console powered by Drake on 8nm, and then in a couple of years they release Switch 2, a hybrid device with a die-shrunk version of Drake on 5nm or whatever? So the Switch Home is basically the TV-only version of Switch 2, it just comes out a couple of years before Switch 2 does. It makes the marketing a bit easier, where Switch Home is very clearly a supplementary device rather than a new generation, and it largely plays the same library but with some exclusives, and then when Switch 2 comes around they just have to say that you can play all your Switch 2 games on Switch Home.

The awkward part would probably be that all the games made in the years between Switch Home and Switch 2 releasing would be missing a performance profile: Drake handheld. Therefore while they'd work on Switch 2, and look great in docked mode, they'd probably have to revert to the base Switch handheld profile, which would mean a huge gap in quality between the two. Not sure if there's an easy way around that other than forcing all developers to support a completely unused profile in the interim.

Anyway, I'd still be very surprised if it's anything other than a hybrid, but with the leak shaking things up I figured I'd share ones of my more out-there thoughts.

Edit: Ok, so for the issue of no Drake handheld profile, they could set up Switch Home so that when it's connected to a 1080p display it automatically clocks down to the planned Drake handheld clocks, and they just call it 1080p mode and claim it's to save power. I'm sure it would piss off devs to no end, and probably some users too, who would prefer to run at 4K and get the benefits of downscaling, but it would force developers to ensure every game would end up being compatible with Switch 2 handheld mode.

Yeah, if Drake is the SoC that's going to be powering Switch devices going forward, I feel like Nintendo is in kind of a weird spot if they're releasing something in the hybrid form factor in 2022. In a post-OLED market, how do they price it? Do they ever intend on selling Drake Switches at the $300-$350 price point, or is the $400-$500 range the new normal for the flagship device?

It kinda plays into my pet theory that I've had for a while that a transition to the next gen SoC is a good time to experiment with the form factor and shoot for a "premium" price point. I doubt they'll go this route, but my personal favorite idea is an Oculus Quest style Switch standalone VR headset/console launching in 2022 at a $450-$500 price point, followed by the official Switch 2 hybrid launch ~2-3 years later with a die shrink, rock solid battery life and various other improvements at $350 announced alongside genuine next gen exclusives like a new Mario Kart, 3D Mario, whatever.

Like some have said, the high end expectations for this SoC's performance seem like overkill if the main intention is to run current Switch games at 4K. Portable VR would obviously change that equation. They could probably squeeze some pretty sweet VR out of this device compared to what the Quest can do and maintaining the OG Switch as the target platform for traditional flat games should leave them with plenty of overhead to tack on no-compromise VR modes for some of these titles.

Games like Tetris Effect, Rez: Infinite, Moss, Astro Bot, etc have shown that VR can augment traditional gameplay experiences in worthwhile ways. I think it'd be pretty cool to be able to Switch back and forth from playing Mario Odyssey in 4K on my TV to playing it in the headset surrounded by a Mushroom Kingdom themed environment, spatial audio and looking at gameplay through a virtual screen with a stereoscopic 3D effect. If Nintendo could pump out 1 or 2 VR exclusives per year alongside a handful of "VR modes" for their flat games, the software support for this system would blow away what Sony and Facebook have provided for their VR platforms. I'd also expect flat games, even if not adapted to VR, to be playable in theater mode in the headset so there could be continuity for a "portable Drake" profile for traditional games when the real Switch 2 comes.

Use this opportunity to play around with VR. Maybe you eat Zuckerberg's lunch and establish Nintendo as an unlikely titan in the next digital frontier. Maybe VR ends up being a fad and Nintendo finds after a few years that basically nobody unplugs it from their TVs anymore. Either way, the pivot back to the hybrid form factor is always there for them to deploy whenever they want.
 
TechPoweredUp's testing for the A2000 is very illuminating. when chips are selected for low power, they can get really low, while matching the 3050

relative-performance_1920-1080.png
power-gaming.png


clock-vs-voltage.png


Clock Frequencies & Voltage
GPU Clock​
Memory Clock​
GPU Voltage​
Idle
210 MHz​
101 MHz​
0.652 V
0.650 to 0.656 V​
Multi-Monitor
210 MHz​
101 MHz​
0.656 V​
Video Playback
227 MHz
210 to 727 MHz​
105 MHz
101 to 203 MHz​
0.662 V​
Furmark
826 MHz
772 to 1200 MHz​
1500 MHz​
0.653 V
0.650 to 0.725 V​
Gaming
(Cyberpunk 2077)
1265 MHz
1237 to 1282 MHz​
1500 MHz​
0.725 V
0.718 to 0.731 V​
V-Sync
(Cyberpunk 2077)
1342 MHz
1327 to 1357 MHz​
1500 MHz​
0.750 V
0.743 to 0.756 V​
Gaming
(23 Games)
1342 MHz
1192 to 1890 MHz​
1500 MHz​
0.737 V
0.656 to 1.056 V​

That's amazing. A2000 is on a 12nm node, right? I wonder how the power efficancy would be ok a 5nm node.
 
I personally (a) don't have the experience/context to talk confidently about what the GPU driver is doing, and (b) wouldn't want to discuss specific details of the proprietary code anyway.

As far as conflicts go, there does seem to be a place that validates at runtime whether various constants defined by NVN match the actual values. If the defines I saw before really were for GA10F, then this would seem to conflict, and one of the two definitions would need to be updated.
Oh, ok.

Another question then. Are these L2$ values assigned to different constants or the same one? If different, which constant is used more throughout the API. If only a small part of the API is using the 1 MB L2$ value it might be something that isn't critical to have the exact value. Of course, I'm just trying to make sense. Especially if the API is in use and the software developed hasn't behaved outside performance expectations. I can't believe that such a different between constants would be left in the code unintentionally.
 
Quoted by: LiC
1
I’m going to go off the limb and say that, the soc has 4MB of L2$ for the GPU, simply due to the structure it has and supposedly being a derivative of an existing chip. I’m not exactly sure how or what it refers to the 1MB, but it would be odd I think. Usually a derivative follows a similar layout to the former original product.


derivative in this sense being cut down of the other one.
 
0
Oh, ok.

Another question then. Are these L2$ values assigned to different constants or the same one? If different, which constant is used more throughout the API. If only a small part of the API is using the 1 MB L2$ value it might be something that isn't critical to have the exact value. Of course, I'm just trying to make sense. Especially if the API is in use and the software developed hasn't behaved outside performance expectations. I can't believe that such a different between constants would be left in the code unintentionally.
They're not the same constant because they're in different sources. And there's definitely not some place where it would be okay to just have the wrong value for, well, anything, but certainly not the total size of a memory component.
 
After the switch lite there is no reason to split the modes. The public has spoken and they want a system that can do both. A docked only mode is a colossal waste of everything
The only reason to do a docked mode only is to throw a bone at less customers then the Lite. But I doubt it would be a product they are happy with since it would most likely be 200$ & just use the dock mode profile of whatever Switch is currently out.
 
That's amazing. A2000 is on a 12nm node, right? I wonder how the power efficancy would be ok a 5nm node.
A2000 is 8nm since it's an Ampere product
The only reason to do a docked mode only is to throw a bone at less customers then the Lite. But I doubt it would be a product they are happy with since it would most likely be 200$ & just use the dock mode profile of whatever Switch is currently out.
$200 ain't gonna cover the r&d and production of this chip, especially with the heavily truncated audience a docked only switch would sell to
 
They're not the same constant because they're in different sources. And there's definitely not some place where it would be okay to just have the wrong value for, well, anything, but certainly not the total size of a memory component.
Does this area also include 12SM? This could be a BC mode for the device, running Switch games with 2 to 4 SMs enabled and 1MB L2 cache would be enough for Maxwell's uarchitecture and save on battery life, but if it does list 12SM, this wouldn't be the case.
 
Quoted by: LiC
1
After the switch lite there is no reason to split the modes. The public has spoken and they want a system that can do both. A docked only mode is a colossal waste of everything

Have I been magically transported to an alternate dimension where the Lite was some kind of crazy failure? It's sold around 18 million units so far, which is already more than the 2DS and 2DS XL combined (which would be the closest comparable devices), and it will have comfortably outsold several of Nintendo's entire consoles by the time all is said and done. Would people only have counted it as a success if it just straight-up outside the main model?

Yes, I wouldn't expect a home-only console to sell better than a hybrid, all other things being equal. But, I also wouldn't expect a highly-priced "pro" model to sell that much better, either. Neither Sony nor MS have reported PS4 Pro or XBO X sales, but what sales figures from Japan show the PS4 Pro accounted for just over 17% of all PS4 sales in the country by the time PS5 had launched. Even if it was higher in other regions, that still puts PS4 Pro sales at maybe between 20-25 million, and Xbox One X sales certainly well below that. The Switch Lite, incidentally, has almost certainly already outsold the One X, and will probably end up outselling the PS4 Pro too.

This is why I'm talking about crazier strategies like releasing a more powerful home console first, then following up with a hybrid to match. We've got a chip that's too powerful for a "pro" console, but arguably too early for a Switch 2, and much bigger (and therefore potentially expensive and power-hungry) than we would have expected for either. We've also got consistent reports of the device launching late this year or early next year, and the leaks seem to contain a pretty complete set of drivers for the device. Things don't really line up cleanly for a straight-up Switch 2 or Switch Pro as I see it, so it's worth considering what more unusual plans Nintendo might have.

I've seen a few people suggest that Nintendo release Drake in a more expensive "pro" style model at anywhere up to $500, and then in a couple of years bring the price down and effectively replace the current Switch, but I don't think the audience for that is necessarily hugely more than a $300 ~3.7TF home console version. In both cases they're likely selling primarily to existing owners, in any case.

And again, I don't actually expect this, I'm very much expecting some kind of hybrid system, but I do think it's interesting to consider the alternatives.

I'd say it makes the marketing a lot harder. You want it to be a complementary premium device to the old generation, until it becomes a cut-down version of the new generation. That's completely reversing the positioning of the hardware.

Then they have to minimize the damage made to their main product appeal, which isn't the best way to play Nintendo games anymore, and to the next generation, which will have 2 years old specs rather than being hyped over.

And that's not mention smaller issues like the Home not having new controller features.

I don't think it's a big issue in terms of positioning, it's just a separate product line, and it doesn't have to be positioned as a "cut-down" version of the Switch 2, just an alternative option. Gaming hardware (and electronics of all kinds) often transition between being high-end to being a budget model in any case; the PS4 was cutting edge technology when it was launched, but is the cheaper model now. If there's any kind of cross-gen period, then Nintendo will already have to have clear markings on what hardware a given game runs on (they actually already do with the Lite, with both labelling on boxes, and warnings if you try to buy software like 1-2 Switch on the Switch Lite eShop), so the Switch Home would be included in that.

Regarding hardware, I don't see Switch 2 being "2 year old specs" being an issue. Firstly, because it's two year old console specs in a hybrid (do you think that people would give out if MS launched a device in the Switch form-factor later this year that had a die-shrunk XBSS chip in it?), but also because Nintendo isn't going to reveal the specs to either device anyway. Not that most Nintendo console owners care that much about having the latest specs in any case, as if they did they probably wouldn't be buying Nintendo consoles.

If there are new controller features on Switch 2, then yeah there's a break in compatibility there, but we already have a similar situation with Switch Lite. It would only be an issue if Nintendo made a big push for some new kind of interface or paradigm for the Switch 2, but this theory is really just assuming a very straight-forward successor to the Switch. Otherwise they'd really need more of a clean break.
 
Does this area also include 12SM? This could be a BC mode for the device, running Switch games with 2 to 4 SMs enabled and 1MB L2 cache would be enough for Maxwell's uarchitecture and save on battery life, but if it does list 12SM, this wouldn't be the case.
12 SM is the only number I've seen, including within NVN. And I don't think so, since the 1 MB number is in the main documentation and public header and is the only L2 value as defined by NVN. It's GA10F sources entirely outside of NVN that have 4 MB.
 
After the switch lite there is no reason to split the modes. The public has spoken and they want a system that can do both. A docked only mode is a colossal waste of everything
I'll entertain Thraktor's theory.

Nintendo wants keep the switch around for as long as possible and claims it's halfway through, perhaps until they release an powerful and affordable switch 2 hybrid with the right node (assuming we don't get a 5nm node). 🤔

An 8nm samsung switch home could help with that and would be a perfect stop gap. Because let's be honest, switch is showing it's age. Performance for 1st and 3rd party are really struggling. Nitnendo knows fans want better performance and more third party games.

So this home could be treated as a "pro console" with shared higher performance 1st party games, while gaining some third party games (goal of increasing third party support) for x years as exclusives, and then when switch 2 hybrid comes out officially ( 4 or 5nm), then it will be paired with that and come out with 1st party exclusives and be treated as next gen.

This model could get current gen games too, and would be close to x box series s. Perhaps an xbone vs PS4 difference in power in GPU performance before DLSS.

IF Drake really is 12 SMs and it's also a hybrid, i really am thinking it needs 5nm for that kind of power and power draw. I dunno about 8nm...

if it ends up being a Home model, it can still have multiple profiles, because it could have backwards compatibility for current switch profiles too.

A2000 is 8nm since it's an Ampere product

$200 ain't gonna cover the r&d and production of this chip, especially with the heavily truncated audience a docked only switch would sell to
Oh whoops. Of course.

Yeah, if Drake is the SoC that's going to be powering Switch devices going forward, I feel like Nintendo is in kind of a weird spot if they're releasing something in the hybrid form factor in 2022. In a post-OLED market, how do they price it? Do they ever intend on selling Drake Switches at the $300-$350 price point, or is the $400-$500 range the new normal for the flagship device?

It kinda plays into my pet theory that I've had for a while that a transition to the next gen SoC is a good time to experiment with the form factor and shoot for a "premium" price point. I doubt they'll go this route, but my personal favorite idea is an Oculus Quest style Switch standalone VR headset/console launching in 2022 at a $450-$500 price point, followed by the official Switch 2 hybrid launch ~2-3 years later with a die shrink, rock solid battery life and various other improvements at $350 announced alongside genuine next gen exclusives like a new Mario Kart, 3D Mario, whatever.

Like some have said, the high end expectations for this SoC's performance seem like overkill if the main intention is to run current Switch games at 4K. Portable VR would obviously change that equation. They could probably squeeze some pretty sweet VR out of this device compared to what the Quest can do and maintaining the OG Switch as the target platform for traditional flat games should leave them with plenty of overhead to tack on no-compromise VR modes for some of these titles.

Games like Tetris Effect, Rez: Infinite, Moss, Astro Bot, etc have shown that VR can augment traditional gameplay experiences in worthwhile ways. I think it'd be pretty cool to be able to Switch back and forth from playing Mario Odyssey in 4K on my TV to playing it in the headset surrounded by a Mushroom Kingdom themed environment, spatial audio and looking at gameplay through a virtual screen with a stereoscopic 3D effect. If Nintendo could pump out 1 or 2 VR exclusives per year alongside a handful of "VR modes" for their flat games, the software support for this system would blow away what Sony and Facebook have provided for their VR platforms. I'd also expect flat games, even if not adapted to VR, to be playable in theater mode in the headset so there could be continuity for a "portable Drake" profile for traditional games when the real Switch 2 comes.

Use this opportunity to play around with VR. Maybe you eat Zuckerberg's lunch and establish Nintendo as an unlikely titan in the next digital frontier. Maybe VR ends up being a fad and Nintendo finds after a few years that basically nobody unplugs it from their TVs anymore. Either way, the pivot back to the hybrid form factor is always there for them to deploy whenever they want.
it will be priced at least at $399 for sure. OLED could drop to $300 and V2 drop to $250. That's how many of us see it.
 
Last edited:
12 SM is the only number I've seen, including within NVN. And I don't think so, since the 1 MB number is in the main documentation and public header and is the only L2 value as defined by NVN. It's GA10F sources entirely outside of NVN that have 4 MB.
NVN or NVN2? It does sound like it could be a hint at disabled SMs for BC, it would be very weird to remove L2 cache on Ampere, as we know it's a bottleneck for the desktop GPU hardware and was spotted by the engineers when designing Orin, I won't argue about leaked specs, but we possibly don't know the context here. I'm just going to wait on L2 cache until it's pared with the rest of the GPU specs.
 
12 SM is the only number I've seen, including within NVN. And I don't think so, since the 1 MB number is in the main documentation and public header and is the only L2 value as defined by NVN. It's GA10F sources entirely outside of NVN that have 4 MB.
That's pretty interesting, as it couldn't be a holdover from previously released silicon; both GM20B and GV10B (potentially relevant if Xavier-based dev kits exist, which IIRC was a common theory) don't match.

Whatever that number was, it was likely entered intentionally at some point. I wonder what's caused the discrepancy.
 
Okay let’s assume this 12SM device is a Nintendo Home Console. With the added space for bigger better heatsink and fan, what do we think this will be clocked at? Could we be potentially looking at a 4TF machine?

Punching above its weight with DLSS this may actually rival a Series S, no?
 
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