• 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!)

I feel people are ignoring just how small Orin NX is despite it having the same die as Orin?


It's 70x45mm.

That is smaller than the Jetson TX1's board.


And Dane would be able to fit into a board of same SoC to PCB proportions as Orin NX within a board of 45.22.5mm (1012.5mm2)

For reference, here is what a 40x30 (1200mm2) board would look overtop the RAM and SoC section of the Switch's OG Motherboard

Overlay_Mobo.png


Fits i'd say considering Orin NX has SoC and RAM only, and that part of the Switch Motherboard is in charge of that, and that square is 200mm2 over the general size half of Orin NX would be.
For further reference, the Orin NX chip package by itself appears to be around 38x38mm, with the die (~21x21?) maybe about as big as the whole TX1 package if your box there is accurate. It's a big boy. I can maybe see an argument for the same specs (cause chopping CPU and particularly GPU down takes out significant chunks), but not the same chip/die as shown by Nvidia, due to both size constraints for Nintendo and the wasted die space for Nvidia (particularly with Nintendo's volumes). It's not an ideal setup for either party.
Assuming that the successor to the Snapdragon 8cx Gen 2 is using the octa-core (8) configuration of the Cortex-A78C, which seems to be the case going by how the CPU cores are described in the rumour from Roland Quandt, which seems to be vindicated by the information written on "CPU Information" (here and here) in the Geekbench 5 benchmarks, and using the Geekbench 5 benchmarks for the AMD 4700S, which is based on the PlayStation 5's APU, the octa-core configuration of the Cortex-A78C is overall theoretically very close to the Zen 2 CPU in the PlayStation 5 in terms of single-core performance, but the octa-core configuration of the Cortex-A78C can theoretically range from being ~43.63% to ~64.36% slower than the Zen 2 CPU in the PlayStation 5 in terms of multi-core performance.
That multicore performance is weird. Sounds like it's still a 4+4 type setup (even if it seems to be 4p+4p) rather than true equivalent 8 core, which probably has something to do with it, if not just basic thermals.
 
I have to ask why we would have to take 120mm2 as a limit (beyond "because it was used for erista"). 384 cores might be fine, but I still think that 2RT cores might not be enough acceleration performance for even software RT. hypothetically speaking, of course
 
I have to ask why we would have to take 120mm2 as a limit (beyond "because it was used for erista"). 384 cores might be fine, but I still think that 2RT cores might not be enough acceleration performance for even software RT. hypothetically speaking, of course
Especially since this will likely be priced at $400-$450, if not higher. A bigger more expensive die is certainly possible.
 
I have to ask why we would have to take 120mm2 as a limit (beyond "because it was used for erista"). 384 cores might be fine, but I still think that 2RT cores might not be enough acceleration performance for even software RT. hypothetically speaking, of course
Especially since this will likely be priced at $400-$450, if not higher. A bigger more expensive die is certainly possible.
I think it's mostly to set a baseline of expectation, much like the TDP being capped at 15W like Erista.
I'll suggest things could go higher on the TDP, but I set 20W as the uppermost limit because of thermal dissipation constraints within the form factor as we understand it. Die size is similar in this respect, there's only so big it can go with a motherboard size for a similar form factor to an OLED Switch.

Keeping within constraints of the current form factor and design keeps discussion within a wholly reasonable/realistic expectation by providing guide-rails for which all other factors are centred around. With so much of this new hardware being unknowable, staying close to where Switch already is for certain metrics like die size, average battery life and TDP (and, to a lesser extent, cost of materials) prevents things from flying wildly into pure fantasy territory. And I don't consider it a bad thing to define some clear boundaries of realistic expectation in discussions like this. That way, even if people postulate other more extravagant ideas, there's at least a baseline to harken back to.
 
I think it's mostly to set a baseline of expectation, much like the TDP being capped at 15W like Erista.
I'll suggest things could go higher on the TDP, but I set 20W as the uppermost limit because of thermal dissipation constraints within the form factor as we understand it. Die size is similar in this respect, there's only so big it can go with a motherboard size for a similar form factor to an OLED Switch.

Keeping within constraints of the current form factor and design keeps discussion within a wholly reasonable/realistic expectation by providing guide-rails for which all other factors are centred around. With so much of this new hardware being unknowable, staying close to where Switch already is for certain metrics like die size, average battery life and TDP (and, to a lesser extent, cost of materials) prevents things from flying wildly into pure fantasy territory. And I don't consider it a bad thing to define some clear boundaries of realistic expectation in discussions like this. That way, even if people postulate other more extravagant ideas, there's at least a baseline to harken back to.
not that keeping to known board sizes isn't reasonable, but there's still a lot of wiggle room to work with. thinks like high density motherboards and whatnot. all a function of how much Nintendo wants to spend and heat, of course, but plus some millimeters probably isn't out of the question. the X1 wasn't initially a Nintendo-oriented chip, so they just happened onto 120mm2
 
Considering that Orin's using the Cortex-A78AE, and Dane's likely to use the Cortex-A78 (4 Cortex-A78 cores & 4 Cortex-A55 cores or the hexa-core (6) configuration of the Cortex-A78C or the octa-core (8) configuration of the Cortex-A78C) by extension, using the Cortex-A53 is definitely not an option.
you misread my question. i asked about
1) increased density of process since Erista/TX1 (20nm to 8nm presumed)
2) removing the unused a53 cores freeing up some extrra space
3) requirement to stick to the TX1 die size since we have no information on what Nintendo is paying for the new chip, but we mostly agree it will likely cost more than X1 SoC at launch.
 
I have to ask why we would have to take 120mm2 as a limit (beyond "because it was used for erista"). 384 cores might be fine, but I still think that 2RT cores might not be enough acceleration performance for even software RT. hypothetically speaking, of course

This exactly, I'm definitely leaning towards the next Switch not including a fan in the device and one in the dock...
Just look at how much real-estate that fan area is taking up in that cut out of the current Switch's motherboard.
 
you misread my question. i asked about
1) increased density of process since Erista/TX1 (20nm to 8nm presumed)
2) removing the unused a53 cores freeing up some extrra space
3) requirement to stick to the TX1 die size since we have no information on what Nintendo is paying for the new chip, but we mostly agree it will likely cost more than X1 SoC at launch.
My bad and my apologies.

Going by the rough annotations of the Tegra X1, the Renesas R-Car H3, and the Snapdragon 835, the amount of space freed up by removing the Cortex-A53 cores in a 8 nm** chip is probably fairly negligible.

** → a marketing nomenclature used by all foundry companies
 
0
This exactly, I'm definitely leaning towards the next Switch not including a fan in the device and one in the dock...
Just look at how much real-estate that fan area is taking up in that cut out of the current Switch's motherboard.
they can always go with a denser battery. the problem with that is, they'd have to buy a new battery rather than keep the same order
 
That's a fair point regarding manufacturing process, and looking into it Samsung did announce they were working on an automotive-grade 8nm process a couple of years ago. I'd still say with density pretty much identical to their 8N process, it's probably very close to standard 8N.

Regarding clock speeds, I don't think that's a process issue, I think it's just the reality of using a really big chip within the power/thermal limitations of the Jetson AGX form factor they've chosen (which is carried over from Xavier). They're limited to 50W for the full board, which means they're simply not going to be able to run a 460mm2 chip at full clocks. Consider that the GA104, which is a smaller die, consumes well over 200W when running at full clocks. The Drive AGX Orin doesn't seem to have the same limitation, as it advertises 254 TOPS (rather than 200), which would suggest a GPU clock as high as 1.5GHz (assuming the DLA clocks don't change much).

I actually wouldn't be surprised to see a Dane in the next Switch running higher clocks than Jetson Orin, perhaps up to about 1.2GHz docked, simply because with a smaller die and smaller GPU it'll have the capacity to.



I don't know what you mean by "unknown numbers" or what Xavier has to do with it, but here's my point: The TX1 was used in the original Switch and is about 121mm2 in size. I don't think it's reasonable to expect that their new chip for the next Switch (Dane) will be substantially bigger than that. We know from die shots of TX1 that Nvidia could spend 12mm2 of that 121mm2 on SMs. We also have a reliable source telling us that Dane is very similar in architecture to Orin, and we now know that an Orin SM takes up about 4.4mm2 of die space each (and that's the old Orin, they may be larger on the new 21b transistor Orin).

I'm saying that Nvidia might be able to squeeze 4 Orin SMs onto Dane, which would take up 17.6mm2, almost a 50% increase over TX1, and even then they're going to have to find that space somewhere else on the die. When people are saying they expect something like the Orin NX GPU for Dane, they're talking about tripling the die area of SMs over TX1, which I think is unrealistic unless you expect a significantly larger die, which is just not something I expect.



Nintendo account for more of Nvidia's revenue than the entire automotive industry. The amount they'd lose by producing tens of millions of dies at twice the size they need to be would absolutely dwarf the R&D costs of using a different GPU configuration (which hey, they're going to have to do anyway, because they're clearly not using full Orin).



Note I said maybe some A55s. I'm not ruling out a full A78 setup, but I'm also not ruling out a couple of A55s. It could be two A55s, which should be sufficient to run Switch's OS, would take up less die space and less power than an A78 core, and would allow the system to run much more efficiently in low-power states (eg sleep mode).



Orin NX isn't Orin S. You can see from Nvidia's photos that the die size of the Orin AGX and Orin NX chips is identical, so it's clearly a binned full Orin die.
So in your own measurements, 8SM would require the chip to be ~139mm² instead of ~121mm². Considering Wii U' s MCM cost twice what Switch's SoC cost and both sold for $299 at launch, while the Wii U's GPU was nearly 160mm² alone, I don't see a hard number on die size, the most important metric is performance per watt, and that the range for the Dane model docked is 10 to 15 watts, which is also performance modes found on Orin NX.

Orin NX, disregarding all of the extra logic, clocks would likely only need to be reduced by 20-25% to reach 15watts, and maybe 35% to reach 10watts.

That 1.3-1.6tflops for the GPU and 1.3-1.6GHz for the cpu. Regardless of the configuration, Orin NX gives us a performance per watt.

I don't think the die size needs to match TX1 because the TX1 was designed on an expensive 20nm process in 2014 the same year Apple was using it, on TSMC. Samsung's 8nm in 2021 is nothing like that process node, Nintendo bought an existing die and costs were whatever it was going to be, Nintendo ended up with a much cheaper SoC than they originally thought they would pay as we know they were working on TX1 back in August 2014. Why do you think Nintendo is stuck at ~120mm²?
 
Last edited:
So in your own measurements, 8SM would require the chip to be ~139mm² instead of ~121mm². Considering Wii U' s MCM cost twice what Switch's SoC cost and both sold for $299 at launch, while the Wii U's GPU was nearly 160mm² alone, I don't see a hard number on die size, the most important metric is performance per watt, and that the range for the Dane model docked is 10 to 15 watts, which is also performance modes found on Orin NX.

Orin NX, disregarding all of the extra logic, clocks would likely only need to be reduced by 20-25% to reach 15watts, and maybe 35% to reach 10watts.

That 1.3-1.6tflops for the GPU and 1.3-1.6GHz for the cpu. Regardless of the configuration, Orin NX gives us a performance per watt.

I don't think the die size needs to match TX1 because the TX1 was designed on an expensive 20nm process in 2014 the same year Apple was using it, on TSMC. Samsung's 8nm in 2021 is nothing like that process node, Nintendo bought an existing die and costs were whatever it was going to be, Nintendo ended up with a much cheaper SoC than they originally thought they would pay as we know they were working on TX1 back in August 2014. Why do you think Nintendo is stuck at ~120mm²?
Pretty much my stance here

That in tandem with the whole "It would mostly just fit where the Eristra model did in the OG Motherboard with RAM did" point too considering how NVIDIA very much is making compact motherboards.
 
This exactly, I'm definitely leaning towards the next Switch not including a fan in the device and one in the dock...
Just look at how much real-estate that fan area is taking up in that cut out of the current Switch's motherboard.
I don't know if newer, passive cooling technologies for smartphones are designed for heavy workloads running for long, sustained periods of time in mind, considering that smartphones generally run at high CPU frequencies in short bursts of time rather than for long, sustained periods of time. And CPU frequency for the DLSS model* in handheld mode will likely be the exact same as the CPU frequency for TV mode.

Of course, Xiaomi's Loop LiquidCool Technology does seem impressive, especially with regards to Genshin Impact. But I don't know if newer, passive cooling technologies, like Xiaomi's Loop LiquidCool Technology, can thermally cope with games much more demanding than Genshin Impact, which is what the DLSS model* will be dealing with. I do think Xiaomi's Loop LiquidCool Technology needs to be vigorously tested for real world scenarios before any conclusions can be made.
 
they can always go with a denser battery. the problem with that is, they'd have to buy a new battery rather than keep the same order

There's multiple options Nintendo could go with in order to use a larger SoC
Not saying that Nintendo are now going to use such a large design, but we can't act like something slightly bigger than the TX1 is unimaginable.

Edit: due to inaccurate information
 
Last edited:
I don't know if newer, passive cooling technologies for smartphones are designed for heavy workloads running for long, sustained periods of time in mind, considering that smartphones generally run at high CPU frequencies in short bursts of time rather than for long, sustained periods of time. And CPU frequency for the DLSS model* in handheld mode will likely be the exact same as the CPU frequency for TV mode.

Of course, Xiaomi's Loop LiquidCool Technology does seem impressive, especially with regards to Genshin Impact. But I don't know if newer, passive cooling technologies, like Xiaomi's Loop LiquidCool Technology, can thermally cope with games much more demanding than Genshin Impact, which is what the DLSS model* will be dealing with. I do think Xiaomi's Loop LiquidCool Technology needs to be vigorously tested for real world scenarios before any conclusions can be made.

One of the reasons I hope Nintendo just goes full out with an all aluminum or magnesium design is because, plastics while great for being non conducting when it comes to energy are also a crutch in such a compact design because they also trap heat and fail to assist in the overall dissipation of such build up. The Switch never reaches temperatures of where it's to hot touch and most likely wouldn't ever get close to normal thermals if it were using premium materials instead.
 
This exactly, I'm definitely leaning towards the next Switch not including a fan in the device and one in the dock...
Just look at how much real-estate that fan area is taking up in that cut out of the current Switch's motherboard.
Er, I’m not so sure about that tbh. I wouldn’t bank on them removing the fan really. Fan helps this device do what it does.


I would think, however, that the rock has a more complex cooling system that helps the switch but not outright remove the fan from the chip itself.
 
It's said the cooling the switch is overkill, so they can always move to a smaller one of need be. We don't know if we can say the same for Dane yet. If the clocks are close to Mariko's the they wouldn't need such a big fan
 
Er, I’m not so sure about that tbh. I wouldn’t bank on them removing the fan really. Fan helps this device do what it does.


I would think, however, that the rock has a more complex cooling system that helps the switch but not outright remove the fan from the chip itself.

I don't agree at all, the real-estate a dedicated fan needs in such a compact design is not only a waste of space but Nintendo can surely source much cheaper fans to fit inside the dock that also encourages more airflow that a much smaller fan could never compete with.
Nintendo will want to do as much as possible to remove constant drain on battery life and fans are a major culprit of this.

Also most modern current pieces of tech manufactured in the Switch's TDP range are all going passively cooled and Nintendo will already do a couple of things in their favor to benefit such a design. The CPU and GPU cores will be under clocked along with RAM as well, so the chances of a 6-8 watt handheld profile needing to throttle would be pretty overkill...
 
It's said the cooling the switch is overkill, so they can always move to a smaller one of need be. We don't know if we can say the same for Dane yet. If the clocks are close to Mariko's the they wouldn't need such a big fan
Yeah, and they can forward that to a fan in the dock
(If they want to get crazy, they could even put a Piezoelectric cooler in the dock)

 
I don't agree at all, the real-estate a dedicated fan needs in such a compact design is not only a waste of space but Nintendo can surely source much cheaper fans to fit inside the dock that also encourages more airflow that a much smaller fan could never compete with.
Nintendo will want to do as much as possible to remove constant drain on battery life and fans are a major culprit of this.

Also most modern current pieces of tech manufactured in the Switch's TDP range are all going passively cooled and Nintendo will already do a couple of things in their favor to benefit such a design. The CPU and GPU cores will be under clocked along with RAM as well, so the chances of a 6-8 watt handheld profile needing to throttle would be pretty overkill...
The thing would be on the edge of throttling which is what you don’t want for a console. It’s because of the fan that it’s clocked high enough to maintain a stable load. You get either something clocked really low and no fan to prevent throttling or you get a thing that’s clocked high enough and you would need a fan to maintain stability.

It should be kept in mind that A78 is not optimized for the SEC 10/8nm process but for the TSMC 7 and the 5nm process, that’s where a good amount of the lower heat and power savings actually comes from.

Mariko went with a smaller fan despite being significantly cooler of a design to work with for the OLED model. Last thing they want is a throttle to happen for this device in any possible way.

If they worry about battery life they can go for a 5000 mAh battery life or a 5500mAh battery life.


And from personal experience (well, my nephew), even with the fan the lite did get pretty toasty playing Minecraft.


Now, whether they need such a big fan for the device is one thing, outright removal of the fan is another thing entirely.
 
Last edited:
0
Nintendo can always go crazy with a thin vapor chamber

279239


or they can keep it simple and make a large metal contact patch to distribute heat
 
0
Do we have the BOM for the Switch OLED? I feel like that's a trial run for the next Switch in terms of buying stuff they know they will need like OLED screens, in bulk so it's a smaller step up for them.
 
0
not that keeping to known board sizes isn't reasonable, but there's still a lot of wiggle room to work with. thinks like high density motherboards and whatnot. all a function of how much Nintendo wants to spend and heat, of course, but plus some millimeters probably isn't out of the question. the X1 wasn't initially a Nintendo-oriented chip, so they just happened onto 120mm2
It's all a function of what those millimetres are capable of adding, though.
I don't think it's unreasonable to suggest it could be a slightly larger die (emphasis mine on slightly), but it's a matter of what it adds, whether that addition within the scope of an acceptable TDP, if the performance per watt addition makes a decent material difference, if the cost-benefit analysis of things like a denser motherboard work in their favour, etc.

I think there's a case to be made, but it would need to be made with all considerations in mind before it could fully supplant the current baseline expectation many here are operating under.
One of the reasons I hope Nintendo just goes full out with an all aluminum or magnesium design is because, plastics while great for being non conducting when it comes to energy are also a crutch in such a compact design because they also trap heat and fail to assist in the overall dissipation of such build up. The Switch never reaches temperatures of where it's to hot touch and most likely wouldn't ever get close to normal thermals if it were using premium materials instead.
There's also a significant cost consideration with a full aluminum/magnesium body. Plastic is dirt-cheap by comparison to milled metal, to the tune of half the cost or less.
Metal bodies also demand exposed/exterior radio antennae for ideal BT/Wifi signal in the design, aren't as good as plastic at shock absorption when it's made too thin and adds significant weight/physical bulk the thicker you mill it.
I don't agree at all, the real-estate a dedicated fan needs in such a compact design is not only a waste of space but Nintendo can surely source much cheaper fans to fit inside the dock that also encourages more airflow that a much smaller fan could never compete with.
Nintendo will want to do as much as possible to remove constant drain on battery life and fans are a major culprit of this.

Also most modern current pieces of tech manufactured in the Switch's TDP range are all going passively cooled and Nintendo will already do a couple of things in their favor to benefit such a design. The CPU and GPU cores will be under clocked along with RAM as well, so the chances of a 6-8 watt handheld profile needing to throttle would be pretty overkill...
The problem with passive cooling design like you suggest is that the moment you take the device off the dock, its cooling capacity is severely diminished and becomes unable to adequately ventilate the heat it was just generating before it was removed from the dock, while still generating more (albeit at a reduced rate). You'll notice this when playing games with a high demand on the SoC in docked mode, as the fan runs quite strong for a minute or two after it's removed, to continue ventilation until a base expected temperature for handheld mode is achieved. That cooling capacity it requires in that transition disappears in a passive cooling design assisted by dock fans.

iPhone 13 Pro hits a 6-7W TDP before it tries to throttle down to 3W if sustained for longer than a handful of minutes. That's a long ways away from a 15W TDP like Switch in docked mode, a device that demands sustained performance. And Android devices are no different in this regard; they trade performance for efficiency,, particularly thermal efficiency, with performance improvements in cell phones coming about from far greater performance per watt as process nodes shrink. So you'll have to show me a modern device that's hitting a 15W TDP at sustained performance using passive cooling, as passively cooled devices I'm aware of don't hit a sustained load of 15W TDP.
 
Last edited:
Especially since this will likely be priced at $400-$450, if not higher. A bigger more expensive die is certainly possible.
Good point; one of the assumptions we were working with back then was a price tag of <= $400. Now we've generally shifted more in favor of >= $400.

-

I'm a bit surprised that Qualcomm's 8cx gen 3 hasn't been announced yet. It's already been more than a year since gen 2.
Also, it's been a little more than a year since ARM announced the A78C variant and I don't recall any product announced to use it yet.
 
iPhone 13 Pro hits a 6-7W TDP before it tries to throttle down to 3W if sustained for longer than a handful of minutes. That's a long ways away from a 15W TDP like Switch in docked mode, a device that demands sustained performance. And Android devices are no different in this regard; they trade performance for efficiency,, particularly thermal efficiency, with performance improvements in cell phones coming about from far greater performance per watt as process nodes shrink. So you'll have to show me a modern device that's hitting a 15W TDP at sustained performance using passive cooling, as passively cooled devices I'm aware of don't hit a sustained load of 15W TDP.
I believe the switch doesn't actually use 15W in docked mode at all times, it taps out at 12-13W, but closer to 12W.

The only time if brushes over 15W is if the joy cons are attached and the thing is dead or nearly dead, reaching a peak of 18W theoretically, but not really crossing far over 15-16w range.

Mariko on the other hand is less than 10W in docked mode with the controllers attached I think.

I forgot where I read that but I will find the source and edit my post


 
Last edited:
One of the reasons I hope Nintendo just goes full out with an all aluminum or magnesium design is because, plastics while great for being non conducting when it comes to energy are also a crutch in such a compact design because they also trap heat and fail to assist in the overall dissipation of such build up. The Switch never reaches temperatures of where it's to hot touch and most likely wouldn't ever get close to normal thermals if it were using premium materials instead.
Unfortunately, one caveat with having a complete aluminium or magnesium housing is that metals in general reflect and absorb radio waves, which can interfere with wireless connections (Wi-Fi, Bluetooth), resulting with poor wireless connections. And that can be very problematic for devices with poor antennae placement, such as the Nintendo Switch, HTC 10, etc.

Of course, Nintendo does seem to take the Nintendo Switch's issue with poor Wi-Fi reception seriously, considering Nintendo replaced the PCB antenna used on the Nintendo Switch with a LCD antenna on the OLED model. But in order to maintain reasonably consistent wireless connections on the OLED model, Nintendo opted to have most of the back of the OLED model use plastic instead of metal since the wireless antennae are located in the back of the OLED model.

Based on my personal experience, the Nintendo Switch can be very warm to the touch when taking the Nintendo Switch out of the dock after long sessions on TV mode. And trying to remove a device with a fully metal housing from a dock after long sessions on TV mode can be very difficult since it can be too hot to touch, especially with metals generally being very good heat conductors, which is very problematic, unless Nintendo plans on installing fans on the DLSS model*'s dock.

~

Yeah, and they can forward that to a fan in the dock
(If they want to get crazy, they could even put a Piezoelectric cooler in the dock)


I wouldn't be opposed to Nintendo using a Peltier cooler fan for the DLSS model*'s dock.

~

I'm a bit surprised that Qualcomm's 8cx gen 3 hasn't been announced yet. It's already been more than a year since gen 2.
Also, it's been a little more than a year since ARM announced the A78C variant and I don't recall any product announced to use it yet.
There was a rumour from Roland Quandt about the successor to the Snapdragon 8cx Gen 2, which going by the description of the CPU cores seems to be using the octa-core (8) configuration of the Cortex-A78C, which seems to be vindicated by the information written on "CPU Information" (here and here) in the Geekbench 5 benchmarks. Hopefully, Qualcomm will reveal the successor to the Snapdragon 8cx Gen 2 during Snapdragon Tech Summit 2021.
 
I believe the switch doesn't actually use 15W in docked mode at all times, it taps out at 12-13W, but closer to 12W.

The only time if brushes over 15W is if the joy cons are attached and the thing is dead or nearly dead, reaching a peak of 18W theoretically, but not really crossing far over 15-16w range.

Mariko on the other hand is less than 10W in docked mode with the controllers attached I think.

I forgot where I read that but I will find the source and edit my post
Even at 12W, it's a huge gulf from 6-7W. But I'll use the listed TDP just for the sake of simplicity, since that seems to be what the Switch's original thermal setup was designed for (even if it doesn't hit that number) and it seems it was still in excess of that figure by some accounts.
And we definitely should not compare against the TDP of a process node change that Mariko was, done 2 years into the life cycle, when I feel confident Dane will also get one at some point in its life cycle, as well. The launch edition, the way Switch was originally designed and its thermal design, is all that is worth comparing here, especially when Orin only scales down to 10W at BEST so far.
 
0
Hi everyone

Long time follower of this thread. First thank you for putting this out there, it's an interesting read, even if I don't know s*** about it. Sometimes when reading through this I feel like a monkey reading through a quantum physics paper.

Anyway, couple of questions that I've been wondering about for some time and was hoping you could answer so I could make a bit more sense of what's discussed here:
1) I keep reading that the Dane or the SoC is not "taped out", in a way that seems to suggest that such SoC it won't be used until it is. So my questions is, what does it mean, being taped out?
2) 8nm die is apparently the general consensus if I understand correctly. I gather this has something to do with chip size and that smaller is better? Is this correct, ie a smaller node will mean a stronger chip in a smaller form which, obviously, would be better?
 
There's multiple options Nintendo could go with in order to use a larger SoC and just for reference sake (because we keep showing blown up pictures of very small chips and things may get lost in translation. Here's essentially Microsoft's SQ1/2 processor which is derived from the Snapdragon 8cx manufactured on TSMC's 7nm process and is 300mm² in size...
Not saying that Nintendo are now going to use such a large design, but we can't act like something slightly bigger than the TX1 is unimaginable.


snapdragon-8cx-compute-platform-chip-front.jpg
8cx/SQ is 120mm2. Not 300mm2.

I'm starting to realize that Orin's performance is disappointing considering the size of the chip when compared to AMD APUs for current generation consoles. This chip really seems to be aimed at a niche AI field that really need so much TOPS for so little raw GPU compute power.

IMHO, T239 will be really different from Orin because for now, the most reasonable choice would be to use a chip like the S870 instead of T239.

S870 manages to reach 1.3 FP32 TFLOPs with a 10.3BTr budget in 83mm2 with a 10W TDP. This chip is already available on Poco F3 at €279 with taxes. Of course it would not be 4K compatible but it seems that Nvidia mobile SoCs seem to suffer from the lack of raw GPU performance at the price of a wider GPU footprint and a larger transistor budget compared to its competitors (Qualcomm/Apple). It would be a nice surprise if Dane would benefit from a better density (the full Orin NX GPU logic in a 120mm2 chip) or a better efficiency (the full Orin NX GPU perf with a smaller 4SM GPU) but for now, the most cost effective would still be to use a competitor's chip even if it would mean no 4K (with DLSS). Actually RT and Tensor cores are probably the reason for this low TFLOPs/Tr ratio especially with low SM count mobile chips.

It would be a shame if the new model would be less powerful than most $300 smartphones of its time when the original model was the best mobile gaming platform when it came out.
 
I see no issues with the die size. The most important aspect, by far, is power consumption. You can have very large chips with lower clocks, like Apple does. Also, I don't think that there's an issue with the chip cost, since this would be the only advantage of going Samsung 8nm over the other options.
 
I see no issues with the die size. The most important aspect, by far, is power consumption. You can have very large chips with lower clocks, like Apple does. Also, I don't think that there's an issue with the chip cost, since this would be the only advantage of going Samsung 8nm over the other options.
S870 is delivering 65% of the Orin NX FP32 GPU performance in a 83mm2 chip.
 
8cx/SQ is 120mm2. Not 300mm2.

I'm starting to realize that Orin's performance is disappointing considering the size of the chip when compared to AMD APUs for current generation consoles. This chip really seems to be aimed at a niche AI field that really need so much TOPS for so little raw GPU compute power.

IMHO, T239 will be really different from Orin because for now, the most reasonable choice would be to use a chip like the S870 instead of T239.

S870 manages to reach 1.3 FP32 TFLOPs with a 10.3BTr budget in 83mm2 with a 10W TDP. This chip is already available on Poco F3 at €279 with taxes. Of course it would not be 4K compatible but it seems that Nvidia mobile SoCs seem to suffer from the lack of raw GPU performance at the price of a wider GPU footprint and a larger transistor budget compared to its competitors (Qualcomm/Apple). It would be a nice surprise if Dane would benefit from a better density (the full Orin NX GPU logic in a 120mm2 chip) or a better efficiency (the full Orin NX GPU perf with a smaller 4SM GPU) but for now, the most cost effective would still be to use a competitor's chip even if it would mean no 4K (with DLSS). Actually RT and Tensor cores are probably the reason for this low TFLOPs/Tr ratio especially with low SM count mobile chips.

It would be a shame if the new model would be less powerful than most $300 smartphones of its time when the original model was the best mobile gaming platform when it came out.
it's mostly unfortunate that it's 8nm. I would have no doubt we would get a 2 TFLOPs machine on 5nm TSMC. It is what it is, but it should be able to get really close to the Steam Deck in GPU performance though.
 
it's mostly unfortunate that it's 8nm. I would have no doubt we would get a 2 TFLOPs machine on 5nm TSMC. It is what it is, but it should be able to get really close to the Steam Deck in GPU performance though.
That's the problem. 2TFLOPs machines are a thing on 7 nm (A12X, A12Z, 8cx, 8cx2, SQ1, SQ2, Renoir) since 2018. Most of them being smaller than 130mm2 with Renoir being the biggest one with 156mm2.

I would find it hard for Dane to reach steam Deck or any of the mentioned chip above in a >120mm2 footprint on 8 nm. Especially with Orin density.
 
0
But size is not that important. What is important is performance/W.
What is important is perf/$. Any newer node would beat 8N in a perf/W comparison. And considering the price of poco phones, I would expect N7P to win at the perf/$ battle too.

Xiaomi is able to deliver 120Hz screen/UFS 3.1/LPDDR4X in a €179 price budget (Poco X3) and 120Hz OLED screen/UFS 3.1/LPDDR5 in a €279 price budget. Both with 7 nm SoCs. Both without gaming revenues and third party royalties.

I'm not impressed by chips being beaten by 3yo mobile SoCs. People are expecting DLSS to double the theorical GPU perfs in order to deliver 4K gameplay for free while it might actually do the tricks for 1080p only with third party games.
 
I dunno, The lite model isn't even that good. Battery life is only barely better than V1 switch.
. If the battery was similar to V2, j would only maybe recommend it.

With my hypothetical scenario, I'm suggesting a reason for people to buy a docked only mode for those on the fenced and that's more performance in docked. Higher resolution, framerates etc. Essentially a switch pro with more marching performance for 3rd party games to the of the x series s and PS5.

or they could make it 50-100 cheaper with the same specs as hybrid.

I was probably not that coherent in my post.

I think it is unlikely a Switch Heavy would offer notably better performance. They wouldn't want the flagship model to be the one that compromises of the gaming experience.

Thinking about it though, I could see some optimisations a Heavy could provide. They could provide more than one cart slot and maybe extra internal storage.
 
0
What is important is perf/$. Any newer node would beat 8N in a perf/W comparison. And considering the price of poco phones, I would expect N7P to win at the perf/$ battle too.

Xiaomi is able to deliver 120Hz screen/UFS 3.1/LPDDR4X in a €179 price budget (Poco X3) and 120Hz OLED screen/UFS 3.1/LPDDR5 in a €279 price budget. Both with 7 nm SoCs. Both without gaming revenues and third party royalties.

I'm not impressed by chips being beaten by 3yo mobile SoCs. People are expecting DLSS to double the theorical GPU perfs in order to deliver 4K gameplay for free while it might actually do the tricks for 1080p only with third party games.
$/perf is secondary to perf/wat, specially on a handheld. And yes, 7nm is better than 8nm in all metrics, but price. And no, 7nm is not cheaper. Specially on today's global chip shortage. There's a reason why Nvidia as a whole went to 8nm, and that is price.

I agree that, over all, 8nm is worse than 7nm, and much worse thn 5nm. I would love for Nintendo to release a 5nm device. But my main points remain unchanged, perf/W is the most important metric by far. $/perf is secondary, and that one is favours 8nm. Die size is irrelevant when accounting the previous 2.

About Qualcom vs Nvidia. The 2015, 20nm TX1 was beating 2018 10nm android phones in real life performance. The theoretical performance of phones doesn't match up, specially with all the driver issues and trottling. I fully expect that an 8nm orin-based device to compare favourably with 2022 flagship phones. Let alone 2018 phones.

If I was forced to choose between a downclocked orin at 8nm and a top of the line Snapdragon 888 at 5nm, I would go for the orin.
 
Last edited:
What is important is perf/$. Any newer node would beat 8N in a perf/W comparison. And considering the price of poco phones, I would expect N7P to win at the perf/$ battle too.

Xiaomi is able to deliver 120Hz screen/UFS 3.1/LPDDR4X in a €179 price budget (Poco X3) and 120Hz OLED screen/UFS 3.1/LPDDR5 in a €279 price budget. Both with 7 nm SoCs. Both without gaming revenues and third party royalties.

I'm not impressed by chips being beaten by 3yo mobile SoCs. People are expecting DLSS to double the theorical GPU perfs in order to deliver 4K gameplay for free while it might actually do the tricks for 1080p only with third party games.
I'm supposed to be impressed by this?

 
$/perf is secondary to perf/wat, specially on a handheld. And yes, 7nm is better than 8nm in all metrics, but price. And no, 7nm is not cheaper. Specially on today's global chip shortage. There's a reason why Nvidia as a whole went to 8nm, and that is price.

I agree that, over all, 8nm is worse than 7nm, and much worse thn 5nm. I would love for Nintendo to release a 5nm device. But my main points remain unchanged, perf/W is the most important metric by far. $/perf is secondary, and that one is favours 8nm. Die size is irrelevant when accounting the previous 2.
Die size is the link between perf/W and perf/$. Especially in late 2019-early 2020 when TSMC and Samsung were both unable to make massive GPU chipsets due to their toolings lacking EUV pellicules for the manufacturing of those big >200mm2 dies which has been addressed by ASML and TSMC in 2020 and with M1p/m being the first 'big' chip on an EUV process.

The theory that newer 7 nm or 5 nm nodes are more expensive than 8 nm while true has never stopped mobile company to put 7 nm SoCs in €279 phones in 2020 and 2021. Smaller mobile chipsets have suffered less from the price increase induced by EUV nodes as opposed to their GPU counterparts that are yet to be seen on EUV process which should change in 2022 with Lovelace and RDNA3.

What's wrong with Orin is the fact it is only able to push for 2 FP32 TFLOPS with a 25W power budget on 8N which is low compared to 1.3 for S870 on N7P (83mm2) and 1.8 for S888 on 5LPE probably with a small <100mm2 die size both consuming less than 10W.
 
I'm supposed to be impressed by this?

Poco X3 pro - S860 (refined version of S855) - N7 - €179
Poco F3 5G - S870 (refined version of S865) - N7P - €279

Both chipsets were the top of the line (android) mobile chipsets from their time. Whereas, at the moment, we can only reasonably expect Dane to have a 4+4 CPU configuration with a 4SM GPU configuration that would be less powerful than the 2019 S865.

People don't seem to understand that the Orin GPU upgrade came at the cost of a massive die size increase and a lower than expected density on the less dense 'high end' node available. It will probably be hard to cram 8SM and 8 big cores in a small mobile oriented chipset.
 
Last edited:
Poco X3 pro - S860 (refined version of S855) - N7 - €179
Poco F3 5G - S870 (refined version of S865) - N7P - €279

Both chipsets were the top of the line (android) mobile chipsets from their time. Whereas, at the moment, we can only reasonably expect Dane to have a 4+4 CPU configuration with a 4SM GPU configuration that would be less powerful than the 2019 S865.

People don't seem to understand that the Orin GPU upgrade came at the cost of a massive die size increase and a lower than expected density on the less dense 'high end' node available. It will probably be hard to cram 8SM and 8 big cores in a small mobile oriented chipset.
I don't think mobile chipsets are useful for head to head comparison with Dane, mainly due to thermal throttling. I searched around and found a thread on xda-dev: https://forum.xda-developers.com/t/...ottling-test-results-for-the-poco-f3.4274331/.
Under sustained load, Poco F3 can throttle really hard, down to 73% peak perf while staying at around 86°C. To make the comparison more on point (perf/W/$), I think it's better to look for another device with same chipset but better build quality at least. Otherwise we should only look at those SoCs to get a performance ballpark for Dane.
 
I don't think mobile chipsets are useful for head to head comparison with Dane, mainly due to thermal throttling. I searched around and found a thread on xda-dev: https://forum.xda-developers.com/t/...ottling-test-results-for-the-poco-f3.4274331/.
Under sustained load, Poco F3 can throttle really hard, down to 73% peak perf while staying at around 86°C. To make the comparison more on point (perf/W/$), I think it's better to look for another device with same chipset but better build quality at least. Otherwise we should only look at those SoCs to get a performance ballpark for Dane.
There would be no thermal throttling with Switch CPU clocks and cooling solution.
 
There would be no thermal throttling with Switch CPU clocks and cooling solution.
Yes, my point is: those mobile SoCs look strong on paper but we won't know how they fare in practice when thermal throttling kicks in, mean while Orin (and Dane) will perform exactly as advertised even under sustained load. This makes paper spec comparison less meaningful imo.
 
Yes, my point is: those mobile SoCs look strong on paper but we won't know how they fare in practice when thermal throttling kicks in, mean while Orin (and Dane) will perform exactly as advertised even under sustained load. This makes paper spec comparison less meaningful imo.
In practice, we all have been disappointed to see TX1 run at 50% the advertised CPU clocks and 80% the advertised GPU clocks (33% in handheld mode) while having an active cooling solution.
 
In practice, we all have been disappointed to see TX1 run at 50% the advertised CPU clocks and 80% the advertised GPU clocks (33% in handheld mode) while having an active cooling solution.
to make the battery last longer. the video I posted shown a 10% loss in battery over 20 mins, despite being on smaller nodes. Dane should be able to run the bench at a sustained clock and with more battery life left
 
to make the battery last longer. the video I posted shown a 10% loss in battery over 20 mins, despite being on smaller nodes. Dane should be able to run the bench at a sustained clock and with more battery life left
Dane won't be able to run the bench with more battery life left if those SoCs were running at the same CPU clocks.


Here are tests made on 8cx gen3. Not sure if they are legits but it shows what we could expect from 8 big cores. Compared to the video posted earlier, an 8*A78C configuration (with two different sets of clocks and memory configuration) would be 30% faster than S870 in MT and 25% faster than S888. That would probably make a higher clocked 7 nm or 5 nm 4+4 CPU faster than a 1.2GHz 8 cores CPU on 8 nm. Moreover, having 8 big CPU cores would limit the space left on the die for the GPU.

I would be pleased to see Orin NX performances being somehow transferred to Dane in a smaller package but numbers are showing that it won't be possible with Orin density and power consumption. They will either have to use uHD libraries to cram that much transistors in a smaller place with 8 nm, using a newer and denser node in order to reduce the die footprint and power consumption or either reduce the numbers of CPU/GPU cores.
 
Last edited:
8cx/SQ is 120mm2. Not 300mm2.

I'm starting to realize that Orin's performance is disappointing considering the size of the chip when compared to AMD APUs for current generation consoles. This chip really seems to be aimed at a niche AI field that really need so much TOPS for so little raw GPU compute power.

IMHO, T239 will be really different from Orin because for now, the most reasonable choice would be to use a chip like the S870 instead of T239.

S870 manages to reach 1.3 FP32 TFLOPs with a 10.3BTr budget in 83mm2 with a 10W TDP. This chip is already available on Poco F3 at €279 with taxes. Of course it would not be 4K compatible but it seems that Nvidia mobile SoCs seem to suffer from the lack of raw GPU performance at the price of a wider GPU footprint and a larger transistor budget compared to its competitors (Qualcomm/Apple). It would be a nice surprise if Dane would benefit from a better density (the full Orin NX GPU logic in a 120mm2 chip) or a better efficiency (the full Orin NX GPU perf with a smaller 4SM GPU) but for now, the most cost effective would still be to use a competitor's chip even if it would mean no 4K (with DLSS). Actually RT and Tensor cores are probably the reason for this low TFLOPs/Tr ratio especially with low SM count mobile chips.

It would be a shame if the new model would be less powerful than most $300 smartphones of its time when the original model was the best mobile gaming platform when it came out.

My bad, this was where I got that information from, it might have been a typo on their part in discussing both 8cx gen2 plus the possibilities of 8cx gen3 in comparison to Apple's M1.

"We do, however, believe that the RAM will be soldered on these future notebooks, so plan ahead before purchasing. The Apple M1 competitor in development reportedly features a die size measuring 20 x 17 millimeters, while the Snapdragon 8cx measures 20 x 15 millimeters. A larger die size means Qualcomm will have ample room to incorporate more performance cores, an addition that might give it a fighting chance against the M1. Then again, Qualcomm might have different plans, according to the latest report."

 
Dane won't be able to run the bench with more battery life left if those SoCs were running at the same CPU clocks.


Here are tests made on 8cx gen3. Not sure if they are legits but it shows what we could expect from 8 big cores. Compared to the video posted earlier, an 8*A78C configuration (with two different sets of clocks and memory configuration) would be 30% faster than S870 in MT and 25% faster than S888. That would probably make a higher clocked 7 nm or 5 nm 4+4 CPU faster than a 1.2GHz 8 cores CPU on 8 nm. Moreover, having 8 big CPU cores would limit the space left on the die for the GPU.

I would be pleased to see Orin NX performances being somehow transferred to Dane in a smaller package but numbers are showing that I won't be possible with Orin density and power consumption. They will either have to use uHD libraries to cram that much transistors in a smaller place with 8 nm, using a newer and denser node in order to reduce the die footprint and power consumption or either reduce the numbers CPU/GPU cores.

Fair enough, going from Orin NX to Dane will probably raise issues with thermal/power consumption. I'm still open to an uplift im power/thermal/W trade-off from removing the AI-oriented functions, but we'll see.
 
0
Hi everyone

Long time follower of this thread. First thank you for putting this out there, it's an interesting read, even if I don't know s*** about it. Sometimes when reading through this I feel like a monkey reading through a quantum physics paper.

Anyway, couple of questions that I've been wondering about for some time and was hoping you could answer so I could make a bit more sense of what's discussed here:
1) I keep reading that the Dane or the SoC is not "taped out", in a way that seems to suggest that such SoC it won't be used until it is. So my questions is, what does it mean, being taped out?
2) 8nm die is apparently the general consensus if I understand correctly. I gather this has something to do with chip size and that smaller is better? Is this correct, ie a smaller node will mean a stronger chip in a smaller form which, obviously, would be better?
1) A tape out is when a chip design has been finalised and mass manufacturing can begin.

2) Yes. But a more advanced process node also means that the size of the transistors has decreased, which means more transistors can be crammed into smaller sized chips.

~

Xiaomi is able to deliver 120Hz screen/UFS 3.1/LPDDR4X in a €179 price budget (Poco X3) and 120Hz OLED screen/UFS 3.1/LPDDR5 in a €279 price budget. Both with 7 nm SoCs. Both without gaming revenues and third party royalties.
The theory that newer 7 nm or 5 nm nodes are more expensive than 8 nm while true has never stopped mobile company to put 7 nm SoCs in €279 phones in 2020 and 2021. Smaller mobile chipsets have suffered less from the price increase induced by EUV nodes as opposed to their GPU counterparts that are yet to be seen on EUV process which should change in 2022 with Lovelace and RDNA3.
I've read that Chinese smartphones companies, including Xiaomi, are generally willing to sell smartphones at no profit, which does explain how Xiaomi was able to sell smartphones running on 7 nm** SoCs at prices as low as €179. And the fact that the smartphone market is cut-throat in China and India does force smartphone companies to sell smartphones at lower prices in China and India at the expense of profit.

Nintendo on the other hand wants to sell products at a profit, no matter how small, unless Nintendo's forced to sell products at a loss.

$/perf is secondary to perf/wat, specially on a handheld. And yes, 7nm is better than 8nm in all metrics, but price. And no, 7nm is not cheaper. Specially on today's global chip shortage. There's a reason why Nvidia as a whole went to 8nm, and that is price.
I think as far as Samsung is concerned, yields are potentially an issue with Samsung's 7LPP process node. Andrei Frumusanu from Anandtech mentioned that there's a possibility that the yields for the Exynos 990 are bad, considering that he's noticed that smartphones equipped with the Exynos 990 has bins as bad or worse compared to the Samsung Galaxy S20+ and the Samsung Galaxy S20 Ultra smartphones he's reviewing, which means there's a possibility Samsung's 7LPP process node has worse yields compared to Samsung's 8N process node. (Nvidia did originally plan on having most Ampere GPUs fabricated using Samsung's 7LPP process node. However, Nvidia did ultimately decide to have most Ampere GPUs fabricated using Samsung's 8N process node.)

As far as TSMC is concerned, securing enough capacity for TSMC's 7 nm** process nodes is definitely an issue, considering how ridiculously high demand for TSMC's 7 nm** process nodes are.

~

My apologies for ranting here, but I don't understand the argument that a 7 nm** process node is too old. TSMC's N5 process node was announced at around the same time that TSMC's N6 process node and TSMC's N7P process node were announced. And by that logic, the PlayStation 5, the Xbox Series X|S, and the Steam Deck are too old, because the APUs for the respective consoles are fabricated using TSMC's N7P process node, which makes zero sense to me.

~

About Qualcomm vs Nvidia. The 2015, 20nm TX1 was beating 2018 10nm android phones in real life performance. The theoretical performance of phones doesn't match up, specially with all the driver issues and trottling. I fully expect that an 8nm orin-based device to compare favourably with 2022 flagship phones. Let alone 2018 phones.
I agree in terms of the GPU. I don't agree in terms of the CPU, especially with the CPU being one of the Nintendo Switch's biggest bottlenecks, and with the Cortex-A75 being significantly more performant and power efficient in comparison to the Cortex-A57.
 
My bad, this was where I got that information from, it might have been a typo on their part in discussing both 8cx gen2 plus the possibilities of 8cx gen3 in comparison to Apple's M1.

"We do, however, believe that the RAM will be soldered on these future notebooks, so plan ahead before purchasing. The Apple M1 competitor in development reportedly features a die size measuring 20 x 17 millimeters, while the Snapdragon 8cx measures 20 x 15 millimeters. A larger die size means Qualcomm will have ample room to incorporate more performance cores, an addition that might give it a fighting chance against the M1. Then again, Qualcomm might have different plans, according to the latest report."

It's funny because I just posted about 8cx gen3. I don't know the size of gen3 but gen2 is definitely not bigger than 120mm2. Qualcomm has shown a wafer full of gen1 chipsets during a show and gen2 is an refined version of gen1.

That's funny that they were expecting gen3 to close the performances with M1 while it will probably end up slower than A14.
 
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