• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Confirmed: The Nintendo Switch is powered by an Nvidia Tegra X1

Status
Not open for further replies.

Polygonal_Sprite

Gold Member
OK, SuperMetalDave.

LOL! Might not be at E3 but people are crazy to laugh at the suggestion we will see more form factors of varying power which all share the same architecture in the future. Nintendo restructured their entire development teams and partnered with Nvidia for this very reason imo.

That didn't make Knack look like Uncharted 4.

But it also meant that Shadow Fall, Second Son, The Order and Driveclub are still some of the best looking games available on PS4 only just being beaten recently and even then not by leaps and bounds. PS4 was close to maxed out much, much faster than PS3 was, that's not even a debate.
 

jon bones

hot hot hanuman-on-man action
Nvidia Tegra line:

Tegra APX
Tegra 2 <-------- 3DS almost used this, before Nintendo decided to go with DMP's PICA 200
Tegra 3 - "Kal-El"
Tegra 4 - "Wayne" - original Shield handheld
Tegra K1 - "Logan" (Kepler architecture) - Shield Tablet
Tegra X1 - "Erista" (Maxwell architecture) - Shield TV, updated Shield TV, Switch
Tegra "Parker" (Pascal architecture)

*Xavier (Volta architecture)

damn, the name Erista is a deep cut
 

Donnie

Member
LOL! Might not be at E3 but people are crazy to laugh at the suggestion we will see more form factors of varying power which all share the same architecture in the future. Nintendo restructured their entire development teams and partnered with Nvidia for this very reason imo.



But it also meant that Shadow Fall, Second Son, The Order and Driveclub are still some of the best looking games available on PS4 only just being beaten recently and even then not by leaps and bounds. PS4 was close to maxed out much, much faster than PS3 was, that's not even a debate.

I agree PS4 is far easier to develop for than PS3. Still took three and a half years to hit its graphical peak with Horizon though.
 

jon bones

hot hot hanuman-on-man action

everything's a very popular comic reference... except that one

Tegra 3 - "Kal-El" - Clark Kent's birth name
Tegra 4 - "Wayne" - Bruce's last name
Tegra K1 - "Logan" - Wolverine's most common name
Tegra X1 - "Erista" - Wolverine's extremely obscure son
Tegra "Parker" - Spider-Man's last name
*Xavier - of X-Men fame
 
I agree PS4 is far easier to develop for than PS3. Still took three and a half years to hit its graphical peak with Horizon though.
Are we sure that even Horizon is at the PS4's peak?

Anyway, not even a console that is very easy to develop for will not get tapped out during the first generation. Factor 5, for example, was able to surpass GCN's launch title, Rogue Leader.
 

Hermii

Member
What kind of goalpost moving is that?

That game is exactly what one can expect from SIE Japan Studio. Imfamous with its particle effects is still a stunning title.

I agree it was silly, I later modified it to Shadow Falls and Horizon.

Infamous wasn't a launch title and wasn't a PS3 remaster/ cross gen. Maybe Xenoblade or Odyssey will close to max out what the Switch can do, but I don't think Zelda is.
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
Speaking of Tegra, it seems ARM made some serious strides on the coherency fabric front recently, as their latest tech allows heterogeneous mixing (i.e. bigLITTLE) within the same cluster: http://www.anandtech.com/show/11213/arm-launches-dynamiq-biglittle-to-eight-cores-per-cluster

Pretty good articles on AMD architectures on anandtech that will give you those numbers, but yes, I said without optimizations that is what you'd get. I'm actually being generous with vliw5, which pushed 5 spus together and only saw an average of 3.4 utilized on average in games, and the 80% number I use is actually stated to be above what the architecture actually achieves generally. http://www.anandtech.com/show/4061/amds-radeon-hd-6970-radeon-hd-6950/4

It's a really good read if you haven't taken a look at it, the article explains why AMD started grouping 4 spus together in the 6000 series cayman+ cards. There is another article about GCN on there, but I haven't recently looked at it, iirc it could hit 92% efficiency in practice, which shows again how much performance gains these architectures achieve at a base level of code utilization.
The thing is, that article hypothesizes on best as well as average cases. But different workloads can deviate arbitrary from the best case as well as from the average scenarios. Apropos, best case for VLIW5 is definitely above 80% - back in the day when VLIW5 was still the norm some guys managed to extract above 90% at dense matrix multiplication workloads via hand-tuned assembly kernels, but I can't seem to find the thread on the old AMD dev forums. As I said, it really depends on the workload.
 

heringer

Member
I agree it was silly, I later modified it to Shadow Falls and Horizon.

Infamous wasn't a launch title and wasn't a PS3 remaster/ cross gen. Maybe Xenoblade or Odyssey will close to max out what the Switch can do, but I don't think Zelda is.

Launch titles are usually affected by other factors other than hardware.

Infamous is actually the perfect example here. Only a few months after launch and it's still one of the best looking PS4 titles. Open world too.

I'm sure we'll see some improvements in a couple of years, but far from a PS3/360 situation.
 

Hermii

Member
Speaking of Tegra, it seems ARM made some serious strides on the coherency fabric front recently, as their latest tech allows heterogeneous mixing (i.e. bigLITTLE) within the same cluster: http://www.anandtech.com/show/11213/arm-launches-dynamiq-biglittle-to-eight-cores-per-cluster

.
Does that mean the X2 can run the Denver cores at the same time as the A57s?

So a hypothetical X2 Switch can utilize all 6 cores at the same time? Or at least that the Denver cores wont be completely useless.
 

Hermii

Member
is it possible that any future revision would use pascal? or would they likely stick with the exact processor that is there?

It is very likely since its using stock hardware and the developer enviroment is made by Nvidia, that its future proofed enough so it wont be complicated to use a newer Tegra should Nintendo want.

Iwata talked about wanting their next hardware to be iterative like IOS and Android devices.
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
.Does that mean the X2 can run the Denver cores at the same time as the A57s?
ARM's latest developments are way beyond Parker - there's nothing using those yet. But, yes, Parker has the right CCI to support HMP.
 

ggx2ac

Member
.
Does that mean the X2 can run the Denver cores at the same time as the A57s?

So a hypothetical X2 Switch can utilize all 6 cores at the same time? Or at least that the Denver cores wont be completely useless.

Nvidia revealed that they can do Heterogeneous Multiprocessing with those CPUs back at Hot Chips when showing Parker.

This DyanmIQ shows that you can have a CPU cluster with a variable amount of cores of varying types of CPUs. The main use for it is for autonomous cars, the article makes an example where you could have an 8 core cluster with two high-powered cores, four mid-range cores and two low power cores.
 
The thing is, that article hypothesizes on best as well as average cases. But different workloads can deviate arbitrary from the best case as well as from the average scenarios.
Though such details are interesting, they're not really material. Even using z0m3le's more ideal estimates, the undocked Switch would have a ceiling around twice as powerful as an Xbox 360. This puts simple ports of current-gen games off the table, even if we don't take into account your more realistic info. A lot of careful rework will be necessary to get those games running well on Nintendo's hardware.
 

z0m3le

Banned
Though such details are interesting, they're not really material. Even using z0m3le's more ideal estimates, the undocked Switch would have a ceiling around twice as powerful as an Xbox 360. This puts simple ports of current-gen games off the table, even if we don't take into account your more realistic info. A lot of careful rework will be necessary to get those games running well on Nintendo's hardware.

Well I'm sure you could just reduce graphic fidelity, computers are already doing this in many current gen games. Intel integrated graphics and lower end amd apus have hardware around switch undocked. A closed system with high performance apis and dedicated resources can realize games on switch.

I've said it before, but there is nothing produced today as a game that couldn't work on switch, the reduction in fidelity is not as drastic as one would expect and what people are willing to accept on the go on a 6 inch screen, 2 or 3 times 360? Sure why not, and docked they are getting twice that performance.
 
I've said it before, but there is nothing produced today as a game that couldn't work on switch, the reduction in fidelity is not as drastic as one would expect and what people are willing to accept on the go on a 6 inch screen, 2 or 3 times 360? Sure why not, and docked they are getting twice that performance.

I generally agree but there could be issues for CPU intensive games that max out all 6 Jaguar cores. I'm betting there are ways around that in most situations but it might not be possible for some games (such games might not even be made yet, but it's possible given the hardware).

Anyway if this sells like the Wii did (or better- unlikely) then I fully expect the majority of AAA multiplats to make it, though obviously downgraded. Not downgraded to the extent the Wii versions were of course.
 

Donnie

Member
Though such details are interesting, they're not really material. Even using z0m3le's more ideal estimates, the undocked Switch would have a ceiling around twice as powerful as an Xbox 360. This puts simple ports of current-gen games off the table, even if we don't take into account your more realistic info. A lot of careful rework will be necessary to get those games running well on Nintendo's hardware.

The amount of RAM and feature set will certainly help. 360 isn't just limited by pure performance after all. Its also missing features from current gen and has a very small amount of RAM, neither is the case for Switch. As far as pure performance, well Switch in handheld mode can get away with using a far lower resolution than 360 ever could.
 

z0m3le

Banned
I generally agree but there could be issues for CPU intensive games that max out all 6 Jaguar cores. I'm betting there are ways around that in most situations but it might not be possible for some games (such games might not even be made yet, but it's possible given the hardware).

Anyway if this sells like the Wii did (or better- unlikely) then I fully expect the majority of AAA multiplats to make it, though obviously downgraded. Not downgraded to the extent the Wii versions were of course.

I meant Today as in 3/21/2017. I know of no game that is actually pushing CPU general compute beyond what is capable on Switch's CPU. Now maybe a game will come along that wouldn't work on Switch, but that game doesn't exist atm afaik and Switch doesn't need to get every game to be successful. There is also the chance that Nintendo frees up the 4th core and or increases clocks to 1.2ghz on the CPU side down the road. (around half a watt increase in power consumption) We are still looking at vulkan reducing CPU overhead by a ton, and Nvidia's custom API should be managing this too. All I'm saying is it's not like last year's Call of Duty or BF1 couldn't run on the Switch with some reduction to graphical fidelity, and there is no reason to expect the next one to.
 

Pasedo

Member
So I was reading this article around porting Tomb Raider to Shield TV which was recently released. Not sure where they get their info from but they state 'pushing the boundaries of the powerful SHIELD hardware with its visual prowess.' Seeing as Switch has the same chip is this the potential we can expect and is it enough for today's AAA games? I mean Tomb Raider reboot is still 3+ years old and if it's being pushed by this game does it give us a good guage how it can handle today's games?

https://www.google.nl/amp/phandroid.com/2017/03/07/tomb-raider-nvidia-shield/amp/
 

atbigelow

Member
So I was reading this article around porting Tomb Raider to Shield TV which was recently released. Not sure where they get their info from but they state 'pushing the boundaries of the powerful SHIELD hardware with its visual prowess.' Seeing as Switch has the same chip is this the potential we can expect and is it enough for today's AAA games? I mean Tomb Raider reboot is still 3+ years old and if it's being pushed by this game does it give us a good guage how it can handle today's games?

https://www.google.nl/amp/phandroid.com/2017/03/07/tomb-raider-nvidia-shield/amp/
If not better, similar. Switch still has more RAM and probably runs similarly to when the Shield is throttling. Vulkan would also help out a lot. That article is also a bit mum on details like TR rendering resolution.
 

Pasedo

Member
If not better, similar. Switch still has more RAM and probably runs similarly to when the Shield is throttling. Vulkan would also help out a lot. That article is also a bit mum on details like TR rendering resolution.

They also say Shield TV can run Crysis 3 natively. An old game yet a game that brought PC's to their knees for many years and still used as a benchmark today...and it runs it natively. Does this give more credence to Switch being able to handle current and future AAA ports?
 
They also say Shield TV can run Crysis 3 natively. An old game yet a game that brought PC's to their knees for many years and still used as a benchmark today...and it runs it natively. Does this give more credence to Switch being able to handle current and future AAA ports?

Any game can theoretically run on switch if scaled back enough
 

Pasedo

Member
Any game can theoretically run on switch if scaled back enough

Battlefield 1 I think is what people today consider a very graphically intensive game. Id love to see at what level they will need to scale down to get it working on docked and undocked mode. Will it look like shit lol
 
Battlefield 1 I think is what people today consider a very graphically intensive game. Id love to see at what level they will need to scale down to get it working on docked and undocked mode. Will it look like shit lol

Treyarch actually faithfully ported Modern Warfare 1 to the Wii without any cutbacks to the gameplay side of things, it didn't actually look too shabby. If Ubisoft are porting Steep to the Switch, the system can't be that hard to port down to.
 
Battlefield 1 I think is what people today consider a very graphically intensive game. Id love to see at what level they will need to scale down to get it working on docked and undocked mode. Will it look like shit lol

Scale back visuals, player counts, destruction, physics and level complexity enough and it will run on switch
 

sfried

Member
Yeah ummmmm no. It absolutely destroys the switch from both a cpu and gpu perspective

http://www.anandtech.com/show/10286/the-97-ipad-pro-review/2

You can take roughly half of the pixel c scores for graphics benchmarks to replicate switch in portable mode. For cpu benchmarks its half in portable and docked mode

We are talking weakest iPads, like iPad Mini 3rd Gen. IPad Pro is considerably more powerful, yes, but also considerably larger than Switch too. Pros are more in line with laptops.
 

Instro

Member
Yeah ummmmm no. It absolutely destroys the switch from both a cpu and gpu perspective

http://www.anandtech.com/show/10286/the-97-ipad-pro-review/2

You can take roughly half of the pixel c scores for graphics benchmarks to replicate switch in portable mode. For cpu benchmarks its half in portable and docked mode

Without knowing what clocks these devices throttle to under extended heavy load, there's no way to get a good comparison. The Pixel C certainly wouldn't actually beat out the Switch in real world performance once it started throttling.
 
We are talking weakest iPads, like iPad Mini 3rd Gen. IPad Pro is considerably more powerful, yes, but also considerably larger than Switch too.

He's also taking into account the Switch's underclocking when the fact of the matter is that most mobile devices run at their best for the limited amount of time they're doing a benchmark, with that type of performance rarely being reached in most other circumstances. According to a poster on here from a couple of months ago, the Shield TV runs at anywhere between 614MHz (and actually stayed at that level when benchmarking) and 1GHz, with an average at 768MHz (which is the Switch's dock mode clock I think), and that's inside of an AC powered box with a fan in it. Is there anything that suggests that the Pixel C and iPad Pro (both of which use passive cooling, batteries, and are each in a much thinner chassis than the Shield TV) are able to sustain those speeds for most of the time that they're running? If not, I think it's a bit unfair to compare down clocked but consistent hardware to something that can vary in clocks as much as those other two devices by simply dividing the (questionably if it really does run at 614MHz when benchmarking) full clock speeds of another device.
 
He's also taking into account the Switch's underclocking when the fact of the matter is that most mobile devices run at their best for the limited amount of time they're doing a benchmark, with that type of performance rarely being reached in most other circumstances. According to a poster on here from a couple of months ago, the Shield TV runs at anywhere between 614MHz (and actually stayed at that level when benchmarking) and 1GHz, with an average at 768MHz (which is the Switch's dock mode clock I think), and that's inside of an AC powered box with a fan in it. Is there anything that suggests that the Pixel C and iPad Pro (both of which use passive cooling, batteries, and are each in a much thinner chassis than the Shield TV) are able to sustain those speeds for most of the time that they're running? If not, I think it's a bit unfair to compare down clocked but consistent hardware to something that can vary in clocks as much as those other two devices by simply dividing the (questionably if it really does run at 614MHz when benchmarking) full clock speeds of another device.

Unless in low power mode the pixel c certainly isnt throttling below an undocked switch. Apple in particular is good at maintaining good clock speeds during extended use. It will always be destroying switch. Apple is just so far ahead of everyone in the mobile tech space
 

Alchemy

Member
Unless in low power mode the pixel c certainly isnt throttling below an undocked switch. Apple in particular is good at maintaining good clock speeds during extended use. It will always be destroying switch. Apple is just so far ahead of everyone in the mobile tech space

What is the most CPU/GPU intensive tasks the iPads can even run? I'm actually really curious if anyone has done any long term testing on Apple stuff running those long enough to possible introduce throttling and to see how well the battery lasts. If something like Zelda was running on an iPad I can't imagine it wouldn't run into throttling at some point due to it being passively cooled.
 
What is the most CPU/GPU intensive tasks the iPads can even run? I'm actually really curious if anyone has done any long term testing on Apple stuff running those long enough to possible introduce throttling and to see how well the battery lasts. If something like Zelda was running on an iPad I can't imagine it wouldn't run into throttling at some point due to it being passively cooled.

theres just no way an ipad pro wouldnt run BotW better than switch given an equal development effort for each platform. people love to say "show me a game in iphone better than BotW" when its a meaningless comparison.
 

Goo

Member
We are talking weakest iPads, like iPad Mini 3rd Gen. IPad Pro is considerably more powerful, yes, but also considerably larger than Switch too. Pros are more in line with laptops.

The new iPad with the old SOC uses Apple's A9. Same as iPhone 6s. It's one of the fastest ARM devices: http://www.anandtech.com/show/9686/the-apple-iphone-6s-and-iphone-6s-plus-review/6

Not sure how it compares GPU vs maxwell but CPU it's faster than a57 in Switch. Apple uses custom versions of PowerVR and their OS has a fast API for graphics called metal.

The Switch has buttons and Nintendo games so that's a huge plus.
 

Alchemy

Member
theres just no way an ipad pro wouldnt run BotW better than switch given an equal development effort for each platform. people love to say "show me a game in iphone better than BotW" when its a meaningless comparison.

Its generally meaningless because mobile devices are designed to handle power draw, throttling, and and heat dissipation differently because the devices have different priorities. I'm not trying to draw any conclusions here I'm legitimately curious if anyone has tried to stress test any iPads for long term gaming and recorded any performance differences. I know Android hardware typically throttles really hard but I dunno about Apple stuff.

Also theres the issue of app size restrictions which are always going to make the comparisons a little tricky. I don't believe app sizes can even exceed 4GB total data so BotW simply would never be allowed to run on an iPad.
 
Its generally meaningless because mobile devices are designed to handle power draw, throttling, and and heat dissipation differently because the devices have different priorities. I'm not trying to draw any conclusions here I'm legitimately curious if anyone has tried to stress test any iPads for long term gaming and recorded any performance differences. I know Android hardware typically throttles really hard but I dunno about Apple stuff.

Also theres the issue of app size restrictions which are always going to make the comparisons a little tricky. I don't believe app sizes can even exceed 4GB total data so BotW simply would never be allowed to run on an iPad.

no developer is going to put nearly as much effort/time/money into a mobile game as nintendo will a switch game.
 
How long before we'll find out the exact clock speeds and all the other stuff to know for certain of its potential in raw numbers?

How long did it take for us to get all the numbers for Wii U? It was rather quick no?
 

Hermii

Member
How long before we'll find out the exact clock speeds and all the other stuff to know for certain of its potential in raw numbers?

How long did it take for us to get all the numbers for Wii U? It was rather quick no?
It doesn't matter, we know the exact hardware. If they tweaked the clocks after the eurogamer leak, it's very minor.
 
no developer is going to put nearly as much effort/time/money into a mobile game as nintendo will a switch game.

Maybe that's why we'll never see Pixel C or iPad throttle to below Switch undocked speeds, because those pairs of devices are designed to run smaller tasks than a systems driven open world game. I assure you that a Pixel C and even an iPad would throttle to a rather low point if it had to run Breath of the Wild for more than 10 minutes, because playing complex games for sustained periods of time is not what those two devices were designed for. My iPhone gets pretty hot and dies rather quickly playing much less complex games, but its benchmarks say that it can run Zelda.
 

Alchemy

Member
no developer is going to put nearly as much effort/time/money into a mobile game as nintendo will a switch game.

Theres Unreal 4 stuff running on iOS though, I'm not sure Icarus-M is released but visually its pretty impressive for mobile stuff. Also I believe the Nether Realm games all run Unreal on mobile? They could potentially stress the hardware. I think theres a difference between stressing the hardware and larger scale game experiences like Breath of the Wild.
 

Goo

Member
Its generally meaningless because mobile devices are designed to handle power draw, throttling, and and heat dissipation differently because the devices have different priorities. I'm not trying to draw any conclusions here I'm legitimately curious if anyone has tried to stress test any iPads for long term gaming and recorded any performance differences. I know Android hardware typically throttles really hard but I dunno about Apple stuff.

Also theres the issue of app size restrictions which are always going to make the comparisons a little tricky. I don't believe app sizes can even exceed 4GB total data so BotW simply would never be allowed to run on an iPad.

Apples A9 holds throttles to 1.4ghz under stress from stock 1.8ghz according to this review https://arstechnica.com/apple/2015/09/a-3d-touch-above-the-iphone-6s-and-6s-plus-reviewed/4/

Might be better in iPad because of the larger form factor.
 
Status
Not open for further replies.
Top Bottom