I expect next Switch undocked performance to match the current one docked, perhaps on a 1080p screen. Docked it should obviously scale higher, maybe offering some of 4kHD support on select titles.
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
Based on how tech has progressed since the Switch, how powerful could a Switch Pro in early 2021 actually be?
- Thread starter ItWasMeantToBe19
- Start date
- Status
I've got one of these, the intel HD 620 is good, but it is basically still adept at last gen games around medium spec. I like mobile tech is on the cusp but not quite there yet.
That's good right? cpu shouldn't be an issue with regards to handling ps4/x1 ports.
I must admit, it's hard to get too excited about the prospects of Nintendo using an a76. It feels like they always go for something that's more conservative than expected.
This is what I'd like to see
A76 @ 1.5Ghz
8GB (1GB-1.5GB for OS)
500Gflops/1TF(docked)
What we'll probably end up with
A73 @1.3Ghz
6GB (1GB for OS)
300Gflops/ around 700Gflops (docked)
I'd like to believe the first scenario is possible, but I've been burned too many times. 😔
By the way, was the dock gpu patent real?
Ok, so lets look at a real world product and what it can achieve at a very low wattage.
The Snapdragon 855 is a 73mm², has the Adreno 640 at 600MHz gives 1tflop fp32. This is a passively cooled device that can play fortnite for hours on end.
With active cooling and docked for power consumption, as well as more modest CPU clocks, you could easily increase the GPU clock by 60% to about 1GHz, to get to that 1.6TFLOPs I've mentioned in the past, and this would be a very powerful upgrade for the Switch.
Yeah I think they are more consistent with their handhelds, GB to GBA is pretty big, GBA to DS is big, DS to 3DS is also pretty big and made a pretty straight forward upgrade path for all of these, then from 3DS to Switch, the jump is absolutely massive, one of the biggest we've ever seen from any company.
The supplementary computing device patent is real yes, and if you think about the Switch Lite, a dock that has it's own power to run games and cost like $199 could be very compelling. If it had its own SoC and RAM, it would only increase the price a small amount over a GPU with RAM, but could play all Switch games digitally, and allow you to dock your Switch Lite to play your games on your TV, since the Switch Lite would just act like an account pass through and allow its game cards to be read by the dock.
Would also be interesting to just see a combo product where you get a Switch Lite and this eDock that is basically a switch standalone console for $399 or $499 (depending how powerful they want to make the dock). Not that I expect the product to exist, but with the Lite not being able to do TV out, this would be an interesting way for Nintendo to add it, and because it's just USB C, it could support the current base Switch models too.
I'm not very sure if this is the right place to ask this question, but if the "Nintendo Switch Pro" or the "Nintendo Switch 2" use a magnesium body (if that rumour have some validity), is there a possibility that there would be problems with antenna placement (for Wi-Fi reception) and heat (as in will the "Nintendo Switch Pro" or the "Nintendo Switch 2" get hot to the touch when playing graphically intensive games?)?
Sure. I'm just pointing out that any prediction must be based on power first and everything else later.
Switch 2 as a mobile device is 100% limited by SoC's power consumption which simply can't go higher than it was in launch model (which has relatively shitty battery life and an active cooler).
Die sizes are more about costs than what is possible to do in mobile space since you can theoretically put a 850 mm^2 GV100 into mobile device but chances are that it won't work off battery for more than ten minutes.
Is the question how powerful could a 2021 Switch model be, or how powerful would it be? If Nintendo went nuts on a "money is no object (and to hell with battery life while we're at it)" approach, then they could theoretically fit something like an 1280-core Ampere/Hopper GPU, an 8-core Hercules CPU and 16GB of 410GB/s HBM2E RAM for something that could be pretty comfortably more capable than PS4 in docked mode (although probably not PS4 Pro level). Portable clocks would have to be dialled down to miniscule levels to get more than an hour of battery life, but if we're ignoring cost we may as well ignore that too.
If the question is what a more powerful Switch model in 2021 would be, the most likely answer is "non-existent", but if they do release something, then realistically I'd expect somewhere between 2-3x more raw performance in each mode. That should be achievable on a 7nm process in 2021 for a reasonable price and retaining batter life somewhere between the launch 2017 and 2019 Switch models.
The interesting question is what the purpose of the new model is. Had a more powerful model been released in 2020, I would have assumed it would effectively replace the existing model and act as a Switch 1.5, a halfway point until the Switch 2 in 2023 or so, breaking up the traditional generational reset. In 2021 that makes less sense, as it's much closer to Switch 2 than Switch 1 in release timing (although perhaps something was planned for this year and pushed back). Previous Nintendo late-gen upgrades like the DSi, n3DS and GBC all had a particular focus feature driving their development, and although Switch doesn't really fit the mould of previous hardware in many other ways, I'm wondering if they release an updated model in 2021 would there be a similar push for a particular feature.
They could just do the same as Sony and Microsoft, and focus on running the same games at higher resolutions. As the Digital Foundry video pointed out, DLSS could be pretty useful here, as they could get away with perhaps only a 2x or so raw performance increase and use DLSS to scale up to 1080p portable and 4K docked. I'm not sure how much die area you'd have to dedicate to tensor cores to achieve this, or what the power consumption would be, but it seems a pretty useful piece of technology if Nintendo want to sell a "4K" model.
Just to play devil's advocate, but in general for a given fixed level of power consumption, a larger GPU operating at a lower frequency will outperform a smaller GPU at a higher frequency, hence why a RTX2080 Max-Q will outperform a mobile RTX2060 despite running at the same TDP. As dynamic power is exponential with respect to clock speed, slower, wider architectures will either outperform faster, narrower architectures at the same power consumption, or equivalently consume less power for the same performance. You'll get gains from additional width as long as the dynamic power is greater than the static power, but static power does decrease linearly with die area, so 7nm should be quite a bit more power efficient at low frequencies (which any Switch-style device would have to run at in portable mode), even if it's not massively more power efficient at much higher frequencies.
That being said, there are pretty rapidly diminishing returns from adding more width to a GPU at very low power levels, and I personally wouldn't expect more than say 512 ALUs being sensible for a Switch form-factor device on 7nm.
If the question is what a more powerful Switch model in 2021 would be, the most likely answer is "non-existent", but if they do release something, then realistically I'd expect somewhere between 2-3x more raw performance in each mode. That should be achievable on a 7nm process in 2021 for a reasonable price and retaining batter life somewhere between the launch 2017 and 2019 Switch models.
The interesting question is what the purpose of the new model is. Had a more powerful model been released in 2020, I would have assumed it would effectively replace the existing model and act as a Switch 1.5, a halfway point until the Switch 2 in 2023 or so, breaking up the traditional generational reset. In 2021 that makes less sense, as it's much closer to Switch 2 than Switch 1 in release timing (although perhaps something was planned for this year and pushed back). Previous Nintendo late-gen upgrades like the DSi, n3DS and GBC all had a particular focus feature driving their development, and although Switch doesn't really fit the mould of previous hardware in many other ways, I'm wondering if they release an updated model in 2021 would there be a similar push for a particular feature.
They could just do the same as Sony and Microsoft, and focus on running the same games at higher resolutions. As the Digital Foundry video pointed out, DLSS could be pretty useful here, as they could get away with perhaps only a 2x or so raw performance increase and use DLSS to scale up to 1080p portable and 4K docked. I'm not sure how much die area you'd have to dedicate to tensor cores to achieve this, or what the power consumption would be, but it seems a pretty useful piece of technology if Nintendo want to sell a "4K" model.
Just to play devil's advocate, but in general for a given fixed level of power consumption, a larger GPU operating at a lower frequency will outperform a smaller GPU at a higher frequency, hence why a RTX2080 Max-Q will outperform a mobile RTX2060 despite running at the same TDP. As dynamic power is exponential with respect to clock speed, slower, wider architectures will either outperform faster, narrower architectures at the same power consumption, or equivalently consume less power for the same performance. You'll get gains from additional width as long as the dynamic power is greater than the static power, but static power does decrease linearly with die area, so 7nm should be quite a bit more power efficient at low frequencies (which any Switch-style device would have to run at in portable mode), even if it's not massively more power efficient at much higher frequencies.
That being said, there are pretty rapidly diminishing returns from adding more width to a GPU at very low power levels, and I personally wouldn't expect more than say 512 ALUs being sensible for a Switch form-factor device on 7nm.
I think it's realistic to match the Snapdragon 855 via portable mode in both power consumption and performance. A 3x bigger Cuda Core count along with a 30% clock increase to 600MHz would get the raw performance we see in Snapdragon 855, with 1TFLOPs. Moving to docked mode, a 1.05GHz clock would give a 1.6TFLOPs of performance, which could easily out perform the PS4.
768 Cuda Cores also would mean 96 Tensor cores for DLSS, with the resolutions being so small on the Switch, it should be enough for effective DLSS. Launch Switch used 9.1 watts to play Zelda BotW and 11 to 12 watts (no screen) to play docked. The chip I'm pointing out is the one found in Galaxy 10 phones and as I've mentioned is just 73mm², the chip I'm talking about shouldn't be much bigger and should fall well within launch power consumption as well.
I believe that the 855 best represents the performance this thread is trying to target, if the Switch iteration launches next spring, it will be a little over 2 years since this level of performance hit the market. The only reason the docked version would out perform it is because of active cooling and DLSS being a very usable feature.
This level of performance would also feel 'next generation' as well, even if it's just a 4x performance increase, the DLSS has the real possibility to multiply that seeming performance again in terms of resolution.
I hope they wait until the leap in performance is significant. I'd also be interested in a device where the dock has extra computing power.
AMD now has a thunderbolt-approved mobo, so maybe they can use that. or just use USB4 protocol. but a "pro dock" might either be separate or part of a pro bundle
Without overthinking battery concerns better than base PS4 is more than doable. Surface Pro X which is also a portable device just larger has specs theoretically better but the emulation layer keeps it from reaching it. Give it Nvidia's newest GPU tech and it could even outperform that.
USB4 is basically thunderbolt, though the work around is to just add a cheap quad core ARM CPU to the dock and have the game run exclusively on the dock, this would allow Switch lite to use the device to display on TV, since it wouldn't be running on the lite at all, it's basically like how the super Gameboy worked, so it's actually a very Nintendo thing to do imo.
EDIT: the other benefit is that this would basically be a digital only, console version of the Switch platform, and thus every eDock sold would be another unit added to the Switch platform. They could also realistically sell it in a combo pack with the Switch Lite for say $399. The dumb part about this idea is that while it would be selling 2 switch units with every purchase, the Switch hybrid already does both of these things in a single product, so a combo with switch lite has very limited appeal imo. I don't really expect this type of device to ever come to market, I'm just pointing out that a stand alone console focused on digital, would be a better product than just a dock with a gpu in it, and the cost difference is basically the cost of an ARM CPU.
I think hitting 1tflops and using DLSS is much more likely for the portable performance of the device. A 60% increase in performance when docked is also a huge move for Nintendo here.
It would definitely feel like a next gen Switch if they are targeting the use of DLSS, because in portable mode they could look at 540p and use DLSS to hit 1080p resolution. When docked, using 720p they can use DLSS to hit 1440p. It shouldn't have an issue with next gen games if such targets and technology is being used.
Would a 4x increase + DLSS able to boost resolution an additional 4x be enough of a jump for you personally?
Last edited:
I don't expect 1TFLOP in handheld until Switch 2 in 2024. I don't know how 1 TFLOP could be done in the same power draw as OG switch in handheld mode at 7nm without being really pricey.
How much power draw does the snap dragon 855 even use anyway?
How much power draw does the snap dragon 855 even use anyway?
It's the SoC found in the Galaxy S 10, it's about 4 to 5 watts. Switch launch models used about twice that power consumption to deliver 3 hours of Zelda BotW to gamers, the performance is well within power and heat limits.
We are talking about a 4x performance increase moving from 20nm to 7nm, it's not unreasonable at all to expect this level of performance IMO. Especially because the Switch uses active cooling, so a 600mhz 7nm 1TFLOPs SoC would be very easy to cool. I'd also expect much lower CPU clocks than the SD855 chip has, which is 4 A76 cores and 4 A55 cores... With the A76 cores well over 2GHz, where I'd limit it to just 2GHz here to actually keep the wattage lower, since passive cooling devices like phones often only boost up to those high clocks for moments on the CPU side.
EDIT: If you are looking at the anandtech link where I get the power consumption from, sustained performance is really all you should care about in those charts, because the device throttles thanks to passive cooling, it would perform at peak performance the entire time in a Switch form factor thanks to the active cooling. I think you'd likely have to reduce the CPU clock speed to keep the power consumption where you want it, but when you are playing with 9 to 10 watts for a Switch 2, you should have plenty of room to play around with these CPU clocks here.
Last edited:
Realistically this could only be done in a way where the dock has an eGPU which takes over the iGPU inside the mobile device completely when it's inside the dock. This means that the dock will be very expensive since it'll basically be an external videocard, not only with GPU but with its own local RAM as well. And from there it's a quarter step to actually making this dock into a standalone home console - which is why I think that Nintendo won't pursue this option.
It would also be an interesting optimization problem, with games for such system working in both UMA and NUMA modes depending on how they are running. Not sure that many devs would appreciate this.
Yeah, I think the only way this could really work is if it was a stand alone console that you could dock a Switch into. The performance of the console would have to match say Switch 2 or be 4x stronger and target 4K visuals. You could then have 4K texture patches downloaded to the console's internal hard drive directly. Like you, I don't think Nintendo would do this, but the benefit is that Switch Lite owners could dock their Switch Lite and it could then play on the TV directly, you'd just have to wait a bit more time for the console to get the RAM data (state) downloaded so it can play from where you left off.
But if we're going this route why not just build a modern online backend which would allow you to transfer not only licenses but save games and profiles between a mobile and a home console? I mean, this seems like an easier way of handling such "integration" than plugging the mobile device into something which is essentially a standalone home console. Would also allow you to use both at the same time and the price difference will likely be negligible.
I am suggesting that it works like a stand alone console and you could play games on both at the same time. I'm also suggesting that it acts just like a dock, where you plug any Switch portable mode device into it and it displays your current game at that current moment onto the TV, that happens a lot faster doing a RAM transfer via USB-C (10GB/s for the one in current Switch models) then it would via the internet.
Nintendo's online network already allows you to sync your account to multiple Switch devices, allowing you to use multiple games, saves and profiles across both.
Yeah, a 4x increase in power definitely qualifies as "significant." I'm not particularly concerned with the next Switch being able to get to 4k. I still do all of my gaming at 1080p. Plus, I'd think attractive 1080p games would look fine on a 4k tv or monitor when viewed from the appropriate distance since it's an integer upscale.
IMO what I hope for and think is possible
Spring 2023
- 5nm or 5nm+ Nvidia SOC
- 8 CPU cores @1.5Ghz or 2Ghz (Ninty loves downclocking)
- GPU @1.5-2TFs (ideally only one mode if it's do-able at right energy use)
- 12GB LPDDR5 (probably 1-2GB used by OS)
- 64GB flash (microSD tho)
- 6.5 inch 1080p screen (ideally foldable)
- Same Joy Con idea, maybe smaller rails, improved controllers but conceptually similar
- 2.5-4 hours battery life
Apple A12X (correction as I labelled it the A13X earlier in this thread) is already roughly 1.4 TFLOP and that is a 2018 chip.
Nvidia not being able to give Nintendo something a bit better than that for 2022 is a stretch considering the Tegra X1 was equivalent/better than the Apple A9X in 2015.
Nvidia not being able to give Nintendo something a bit better than that for 2022 is a stretch considering the Tegra X1 was equivalent/better than the Apple A9X in 2015.
why have only one gpu mode? you'd be sqandering docked play with that
Idk, uniformity for devs? I mean if they can have one mode that tops out the SOC why not? Imagine if switch could have docked in portable from day one, full 400-500GF?
it'd be nice, but then you're not leveraging the form factor enough. if the those clocks are good for handheld, then they can be even better for docked. the devs already can prioritize one mode over the other and just scale resolution (and textures) as needed
in other news, there's a passively cooled 1650 coming out. call me an optimist, but I believe 1650-level performance is doable with the Switch 2 (probably with more cores/lower clocks)
All I'm saying is ideally the max GPU capability that the chip is designed for be the base development mode. I guess they could push docked a bit more if you're right. But if the SOC is rated to go say 1.5TF or 2TF (like Tegra was 500GF Max) I'd love for it to do that in portable modeit'd be nice, but then you're not leveraging the form factor enough. if the those clocks are good for handheld, then they can be even better for docked. the devs already can prioritize one mode over the other and just scale resolution (and textures) as needed
in other news, there's a passively cooled 1650 coming out. call me an optimist, but I believe 1650-level performance is doable with the Switch 2 (probably with more cores/lower clocks)
That would only really make sense if you couldn't use more power when it is plugged in, which because the Switch has active cooling, it shouldn't hit a thermal limit before a clock/voltage limit.
I don't think we'll get a Switch Pro, but if we did I think it'd look something like:
Process Node- 12nm TSMC
CPU- 4 A75 cores at 1.6Ghz
GPU- Turing 256-CUDA cores. Docked performance of about 800Gflops, Portable of about 400.
RAM- 8GB LPDDR4X. Split up between 2 4GB modules. Runs at 1661Mhz
Storage- 64GB
Screen- is the same.
Price- $349, Standard model reduced to $279
That'd give you the "Pro" experience, while holding back for a proper sequel.
Alternatively; just a TX1+ clocked higher so that it has the OG Model power draw, with more performance.
Process Node- 12nm TSMC
CPU- 4 A75 cores at 1.6Ghz
GPU- Turing 256-CUDA cores. Docked performance of about 800Gflops, Portable of about 400.
RAM- 8GB LPDDR4X. Split up between 2 4GB modules. Runs at 1661Mhz
Storage- 64GB
Screen- is the same.
Price- $349, Standard model reduced to $279
That'd give you the "Pro" experience, while holding back for a proper sequel.
Alternatively; just a TX1+ clocked higher so that it has the OG Model power draw, with more performance.
What are some of those tablets? I didn't know these were a thing, sounds cool.
I wonder how valuable a DLSS capable mobile chip that can be used in a switch, laptops, VR headsets, next gen shield products, and tablets would be for Nvidia to invest in the R&D to develop it?
raw numbers or equivalent to 1.4 TFLOPs xbone? I think you mean the former. To be fair, apple has always been at least 2 years ahead in performance vs the competition.
75 watt TDP though. I'm assuming its still the 3 TFLOPs equivalent. Certainly still 12nm. I think 3 TFLOPs in docked mode is possible in a more advanced nvidia architecture in 5nm+ or 3nm at maybe 20 watts. We'll see.it'd be nice, but then you're not leveraging the form factor enough. if the those clocks are good for handheld, then they can be even better for docked. the devs already can prioritize one mode over the other and just scale resolution (and textures) as needed
in other news, there's a passively cooled 1650 coming out. call me an optimist, but I believe 1650-level performance is doable with the Switch 2 (probably with more cores/lower clocks)
Anyway, I really hope we get a switch pro next year. Matching PS4 base would be fantastic. Just anywhere between xbone-ps4. I don't know how much longer I can hold out waiting 3 more years for switch 2.. when we can get pro next year and switch 2 in 2024 instead. Also 800 GFLOPs for a switch pro feels like such a waste of an upgrade. If it was between that and waiting 3-4 years of switch pro at 3 TFLOps, I'd choose switch 2 easily. LPDDR5 is coming out this year too, so sticking to LPDDR4X in 2021 would be awful considering the limitations.
In regards to RAM amount.. Nintendo could go with Samsung and do 6 or 12GB.. or micronex and be able to do 8GB.
Last edited:
Yea I posted this before I learned the mobile variant was 1024 cores at a lower clock and at 35W TDP. Closer to my expectations, surprisingly.
As for ram, Micron's LPDDR5 is available in 6GB, 8GB, and 12GB variants. All are, I assume, are single chip
I'm not an expert but I understand everything that makes it into a final console spec is usually decided upon at least a year before launch.
I do not think Nintendo will do anything to upset those 70 million Switch owners by the theoretical launch of March 2021. It would be higher than 3DS if they waited for a Q3 launch like Lite.
So whatever tech makes it in I don't expect exclusive content for it like DSi and new 3DS before it. That said, there are probably some games out now that are not compatible with Switch Lite so who knows. Nintendo is weird.
Edit: Can 1-2-Switch, Arms or Labo be played on Lite?
I do not think Nintendo will do anything to upset those 70 million Switch owners by the theoretical launch of March 2021. It would be higher than 3DS if they waited for a Q3 launch like Lite.
So whatever tech makes it in I don't expect exclusive content for it like DSi and new 3DS before it. That said, there are probably some games out now that are not compatible with Switch Lite so who knows. Nintendo is weird.
Edit: Can 1-2-Switch, Arms or Labo be played on Lite?
never said anything about dropping nvidia, only that Intel is starting to open up with Thunderbolt so that board makers can use it on non-intel boards
Yeah I did some research and noticed that too with micronex. 44 to 50GB/s bandwidth on 64 bit buswidth. I wonder when the first 128 bit will be, and if its something Nintendo is willing to put in the pro or switch 2. They absolutely need to for Switch 2 imo.
Would be interesting to see if Nintendo goes with micronex or samsung if they go with LPDDR5 Ram for a hypothetical switch pro next year. 6GB would be enough to match xbone/ps4 for games if Nintendo keeps the OS barebones like the current models.. But if they want to add extra features, they should go with Micronex's 8GB.
This is a good analysis, let me add some things:
Having the same die size on 7nm would make it more expensive than the original Switch, and that's without considering the rumors that said the reason Nintendo went with the Tegra X1 is because Nvidia had a lot of them unused because it was a commercial failure for them, so they got them really cheap. I think Nintendo will want to have a small chip that can be used to sell the console charging not too much and at a small profit, like they always did.
There is no need to have A55 cores because the A53 cores in the Switch are unused. If this new chip is being made for Nintendo, there is no need to have unused cores. Maybe only 4 cores are enough. Could they have compatibility problems if they go with Hercules?
How did you get your transistor count projections for Hercules? Are you considering that Ampere cores will be bigger than current Nvidia cores? Why add tensor cores? They don't seem worth the area for this
I think if I remember correctly A55s could run all the code of the OS on a new switch (since theyre compatible with A57 code) which would leave whatever big cores to games giving 4 A76 or whatever they want for games and 4 a55 for OS functions.
all of these ARM cores implement the same instruction set, so 4 Hercules cores should have no problem running code compiled for the A57. I think they would only include little cores if they wanted to use it for the OS, but i feel like Nintendo would go with the simpler route and just dedicate some of the primary cores CPU cycles to the OS.
that is what they did with the current switch ... but the X1 didn't allow for use of big and little at the same time ... which is likely why they just locked em.
It all depends on if OS features are going to become more prominent or not but I think it would be a good Idea to offload the OS to it's own little cores and give every available big core to the Devs for Games
Mariko still use context switching. the A53s don't do anything when gaming
right... small cores on a future switch revision could be active is what I'm saying.
I wonder if AMD could one day create a mobile SoC lineup that can be competitive against Nvidia's Tegra SoC lineup (like how AMD created the Ryzen lineup that can be competitive against Intel's Intel Core lineup).
I know that it's very unlikely at this time that Nintendo will drop Nvidia (for obvious reasons). And I'm not sure if AMD's mobile GPUs can be competitive against Nvidia's mobile GPUs right now.
But I was thinking that if AMD can create a mobile SoC that can be competitive against Nvidia's Tegra SoC, I imagine AMD would give Nintendo much more flexibility in terms of customisation (the type of CPU cores being used, etc.).
I know that it's very unlikely at this time that Nintendo will drop Nvidia (for obvious reasons). And I'm not sure if AMD's mobile GPUs can be competitive against Nvidia's mobile GPUs right now.
But I was thinking that if AMD can create a mobile SoC that can be competitive against Nvidia's Tegra SoC, I imagine AMD would give Nintendo much more flexibility in terms of customisation (the type of CPU cores being used, etc.).
I totally forgot about Samsung's and AMD's mobile GPU partnership. I'm very curious about how power efficient AMD's mobile GPUs (that I presume will run exclusively on Samsung's mobile devices, at least for now) considering that if I remember correctly, AMD's GPUs in general are not exactly well known for being power efficient.
But it sounds like Samsung might be as restrictive as Nvidia in terms of customisation of the SoC, if Nintendo, for some reason, decides to use Samsung's SoC instead of Nvidia's Tegra SoC (the newer iteration, of course) for future Nintendo Switch models (which I personally think is unlikely).
The 20nm Tegra X1 GPU is 1 TeraFlop, at 1 GHz. A 7nm revision could reach 2~3x that.
This, plus DLSS could run any Xbox One or PS4 game on handheld mode.
Not much better than what's in the switch now, rendering it mostly pointless. For Nintendo get a true leap they'd either have to do the opposite of the Lite which is a docked-only console or they'd be releasing an expensive device with 50 minutes of battery life.
if Qualcomm can turn Radeon into a mobile powerhouse, then surely AMD can too. as for the lack of semi-custom, that would mean Nintendo could, theoretically buy a Samsung/AMD chip (really, it's just an ARM/AMD chip as Samsung bowed out of custom ARMs
What I meant by power efficiency is power consumption. If I remember correctly, AMD's GPUs tend to have a much larger power consumption than Nvidia's GPUs.
So I'm curious to see how the power consumption of AMD's mobile GPUs (that I presume will be used exclusively on Samsung's mobile devices, at least for now) compares to the power consumption of Nvidia's mobile GPUs on future iterations of Tegra (that I presume will be used on future Nintendo Switch models and/or future mobile devices).
- Status