Nintendo Switch Technical discussion [SOC = Tegra X1]

Switch has now moved over 32 million units, and has rocked the bone stock Tegra X1 since launch. Seeing as how its likely the Switch will see a "New" SKU at some point in the next couple years, what is the cost/investment required with revising an existing chip compared to a brand new design? The Tegra X2 has even more hardware that a console wouldn't need, but would the investment be all that large to revise the Tegra X2, removing unneeded hardware and moving to the more popular 16nm FinFet process? Wouldn't the cost be similar to the revision the Xbox One saw moving to the S model?
 
Nintendo doesn’t want to break compatibility with existing games?

Or for those games to run faster on new hardware?
 
Switch has now moved over 32 million units, and has rocked the bone stock Tegra X1 since launch. Seeing as how its likely the Switch will see a "New" SKU at some point in the next couple years, what is the cost/investment required with revising an existing chip compared to a brand new design? The Tegra X2 has even more hardware that a console wouldn't need, but would the investment be all that large to revise the Tegra X2, removing unneeded hardware and moving to the more popular 16nm FinFet process? Wouldn't the cost be similar to the revision the Xbox One saw moving to the S model?
Now that we have a confirmed interest in the Nintendo HW from consumers, it ought to be worthwhile on the R&D side to rework the chip to remove the extraneous bits as they'd very much be looking at fairly substantial order of wafers. The twin Denver cores in the base TX2 design would free up a fair chunk of space, certainly. TX2 is already on 16nmFF, mind you, and I imagine they would actually just opt for the "12nm" density optimized process anyway since it's a drop in the bucket being the same node generation.

Alternatively, they just jump straight to 7nm because the chip would that much smaller, and the per-yielded-chip cost may end up being a wash in comparison despite it being on a more expensive node, and the pricing would only get better moving forwards. The benefits to power consumption are also a big win here considering they'd likely just stick with approximately iso-performance and look at reducing costs elsewhere in the console unit. At the same time, it might even be worthwhile to look at GDDR6 support, although I'm not sure how it compares to the latest LPDDR4 bins (power per pin etc). This would be extremely optimistic though.
 
Nintendo doesn’t want to break compatibility with existing games?

Or for those games to run faster on new hardware?

I would think that keeping comparability wouldn't be an issue, seeing as how many consoles, including the Wii, had a chip revision that didn't break comparability. I have been told that unlike the Xbox One S performance boost, that would be unlikely with Switch, because the API on the Switch is likely much closer to the metal.
 
I think if there’s a new soc it will be a simple shrink to a cheap variant of 14/16nm process.

No new features, no speed boost, nothing, just power consumption reduction so they can put it in a smaller device with a smaller battery.
 
If they reduce power consumption, I can see then allowing the player the option to play the game at "docked" quality even when playing portable. But preferably, the user would be allowed to chose to play it in the classical portable quality if he'd rather get the battery savings.
That is the easiest way to implement a use for extra efficiency.
 
New IOW, what's the cheapest thing Nintendo could do while boosting sales?
Either nothing or order a redesign on ultra-cheap 28nm that takes away all the blocks that aren't being used (ISP, I/O, Cortex A53, etc.), because 20nm is rather useless for SoCs.


Something somewhat exciting would be a new SoC that takes advantage of a newer process to clock at docked mode values when going mobile.
However, you should expect the most boring outcome possible.
 
Either nothing or order a redesign on ultra-cheap 28nm that takes away all the blocks that aren't being used (ISP, I/O, Cortex A53, etc.), because 20nm is rather useless for SoCs.


Something somewhat exciting would be a new SoC that takes advantage of a newer process to clock at docked mode values when going mobile.
However, you should expect the most boring outcome possible.

Depends, if its a chip revision that goes into the standard SKU, then yea, nothing but a cheaper solution is likely. If they release a "New" SKU like they did with the 3DS and DS before it, then a more substantial revision will take place. A larger 1080p screen that takes up the space currently occupied by the bezel could be accompanied by a new chip capable of running docked clock speeds in portable mode.
 
Make mobile only and docked only versions for cheaper.

Don’t they usually try to price it the same or higher when they put out new revisions of a popular system?

If there are cost savings, they may choose to pocket the savings while maintain8ng the price.

Depends, if its a chip revision that goes into the standard SKU, then yea, nothing but a cheaper solution is likely. If they release a "New" SKU like they did with the 3DS and DS before it, then a more substantial revision will take place. A larger 1080p screen that takes up the space currently occupied by the bezel could be accompanied by a new chip capable of running docked clock speeds in portable mode.

I think that’s what enthusiasts like us would like. But maybe a more streamlined design without performance or battery improvements would be enough to boost sales, including getting existing Switch owners to buy again for the new shiny, even if it doesn’t bring better performance or battery life.
 
I think that’s what enthusiasts like us would like.

No question about it. A lot of early adopters of consoles are more than willing to buy premium offerings a few years down the road. Xbox One X is a prime example. I would have to believe a large percentage of Xbox One X buyers already owned a base model. The DS and 3DS revisions almost certainly caused a lot of people to double dip. If I were Nintendo I wouldnt wait to long however. At some point there needs to be a clean break, and the true successor to the Switch needs to stand out. Coming out a couple years after the PS5 seems like a good strategy to me. A couple years of advancement in the mobile processor market helps close the gap, and like the Switch now, could allow for many current generation games to be ported to the console. Who knows, the true Switch successor may be able to match or exceed the PS4 Pro/Xbox One X capabilities if it were to release in 2023, giving the original Switch a 6 year life cycle. Of course the base model could sell for longer, but it would then be the value priced hardware selling on its legacy software.
 
upload_2019-4-30_15-38-39.png

Interesting to find out that Nintendo has more performance profiles than we originally knew about. Apparently they have a 460Mhz option for portable mode, and Zelda BoTW and Mario Odyssey use it. Mortal Kombat 11 is reportedly the first third party game to have access to this mode. They have also introduced a CPU boost mode that is used during loading scenes that has proven to drastically reduce loading times in Zelda BOTW and Mario Odyssey. From the footage I have seen from hackers overclocking their Switch, I would love to see a CPU boost mode for gameplay made available. Doom and Wolfenstein 2 framerate dips were nearly eliminated when boosting the CPU to 1400Mhz. Come on Nintendo, lets give that little cooling fan and workout and give developers some extra juice.
 
I would guess that the 0x9222..... modes are recently approved modes after extensive testing/QA.

Curious that they have modes where the RAM is underclocked considerably, though maybe that's just because it's just one of the typical clock modes (1066, 1333, 1600).

Still, no 1020/384/1600 nor 1020/230/1066 mobile modes, although one of the recent additions was 1020/460/1600...

Wonder what the power consumption measurements would show across all these modes.
 
Last edited:
From what I have read, Zelda BoTW has used the 460Mhz clock speed for portable mode, presumably the entire time. This would help explain why the game outperforms its docked performance in terms of framerate. The clock speeds between docked and undocked werent as different as we had thought. Typical Nintendo though, saving that profile for themselves until opening it up for MK11.
 
i wonder, when the new switch finallly comes out, and they unlock the boost clock for older switch.... maybe they'll also release "AIRPLANE LABO" and fly?

swooosh!


or strack it together with PS4 Pro for dual jet airplane!

but seriously, currently Switch fan spins way too slow. Even when the switch exterior feels hot, the fan still runs quiet.
 
Back
Top