• 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.

Nintendo Switch: Powered by Custom Nvidia Tegra Chip (Official)

A

AzaK

Member
Zil33184 said:
We'll see, but if they do go with 128 bit lpddr4 then even an underclocked X1 could hit PS4 quality at 960x540. That would also easily upscale to 1080p and rumours are that the target resolution for NS is 540p. The only bottleneck at that point would be CPU performance.
Click to expand...

What? Where are these rumours? All the ones I've heard from people who have nailed everything so far say 720.
 
P

Principate

Saint Titanfall
Zil33184 said:
We'll see, but if they do go with 128 bit lpddr4 then even an underclocked X1 could hit PS4 quality at 960x540. That would also easily upscale to 1080p and rumours are that the target resolution for NS is 540p. The only bottleneck at that point would be CPU performance.
Click to expand...
Didn't the Nintendo PR mention a HD screen.

Edit: Checked yes it did the screen size is 720p minimum unless they were incorrect on their PR.
 
P

Principate

Saint Titanfall
Branduil said:
Nintendo's not going to set a sub-native resolution as the target resolution. That rumor is bunk.
Click to expand...
Yeah makes absolutely no sense they might has well just used the wii u screens if they were or at least some for cheaper than 720p screens for no real advantage outside of videos. Would have been cheaper and Nintendo historically has absolutely no problem with low resolution screens on handhelds.
 
P

Peltz

Member
lenovox1 said:
The "target" resolution that a video game runs at and the actual screen's full resolution are two different things.
Click to expand...

Nintendo has never rendered at anything other than native resolution on their handheld screens (unless we are talking about backwards compatibility with lower res handheld systems which doesn't count).
 
Cheeseman Battitude

Cheeseman Battitude

Member
Perhaps the Switch can upscale to a very high quality on the tablet. I have no idea about this, but what if a really high quality method of upscaling from 540p to the screen's 720p is less taxing than rendering at 720p outright? They might have found a way of doing it that makes the difference between upscale and native really minor.
 
Astral Dog

Astral Dog

Member
Cheeseman Battitude said:
Perhaps the Switch can upscale to a very high quality on the tablet. I have no idea about this, but what if a really high quality method of upscaling from 540p to the screen's 720p is less taxing than rendering at 720p outright? They might have found a way of doing it that makes the difference between upscale and native really minor.
Click to expand...

no, i dont think so, Nintendo no magic
 
K

KingSnake

The Birthday Skeleton
Cheeseman Battitude said:
Perhaps the Switch can upscale to a very high quality on the tablet. I have no idea about this, but what if a really high quality method of upscaling from 540p to the screen's 720p is less taxing than rendering at 720p outright? They might have found a way of doing it that makes the difference between upscale and native really minor.
Click to expand...

Yeah, Nintendo uses stars dust.
 
A

AzaK

Member
lenovox1 said:
The "target" resolution that a video game runs at and the actual screen's full resolution are two different things.
Click to expand...

And you really think Nintendo will make a handheld that has a screen of 720, but expect people to render at 540 and upscale?
 
N

nikatapi

Member
Seriously, i don't think there's even a remote possibility for sub-native resolution at least on Nintendo games. Some third party demanding games could be sub-native or utilizing dynamic resolution but i believe it's probably going to be 99% native.
 
S

sir_bumble_bee

Member
nikatapi said:
Seriously, i don't think there's even a remote possibility for sub-native resolution at least on Nintendo games. Some third party demanding games could be sub-native or utilizing dynamic resolution but i believe it's probably going to be 99% native.
Click to expand...

This is my take. Only the most demanding games on the console I could see being rendered at 540p in handheld mode, which is something I'm okay with if the performance is still appropriate.
 
N

NateDrake

Member
nikatapi said:
Seriously, i don't think there's even a remote possibility for sub-native resolution at least on Nintendo games. Some third party demanding games could be sub-native or utilizing dynamic resolution but i believe it's probably going to be 99% native.
Click to expand...

Most Nintendo made Wii U games were native 720p, so Nintendo should have no problem reaching that resolution for Switch titles on more powerful hardware.
 
K

KingSnake

The Birthday Skeleton
I don't see Nintendo not targeting 720p in mobile mode with their games.And I fully expect them to render at 1080p on the TV.

And I don't see anybody upscaling from 540p to 1080p on the tv. That would be unplayable.

I wonder if Nintendo/Nvidia can provide some kind of checkerboard rendering on the TV for third party AAAs, something like PS4 Pro is doing.
 
N

nikatapi

Member
NateDrake said:
Most Nintendo made Wii U games were native 720p, so Nintendo should have no problem reaching that resolution for Switch titles on more powerful hardware.
Click to expand...

Yeah for sure, i have no doubt about that, and we know how great performing almost all Nintendo-developed games are. Maybe if there's a very demanding 3rd party game it could render a little lower. If i recall correctly the COD games were not 720p on WiiU, or at least not all the time, so i could see something like that happening on Switch. It is a developer decision anyway, so we are just speculating.
 
B

BuggyMike

Member
Considering the dev kits are supposedly overclocked X1s (which use fully ARM processors) I wonder if Nintendo avoided using Denver cores for fear of it being drastically different to work with. I'm not sure how different Denver cores would be to regular ARM cores, but I've seen it suggested that Nintendo may just want to stick with A57. How likely would an ARM A72 be? Are those significantly more expensive than A57 cores?
 
S

sixteen-bit

Member
nikatapi said:
Yeah for sure, i have no doubt about that, and we know how great performing almost all Nintendo-developed games are. Maybe if there's a very demanding 3rd party game it could render a little lower. If i recall correctly the COD games were not 720p on WiiU, or at least not all the time, so i could see something like that happening on Switch. It is a developer decision anyway, so we are just speculating.
Click to expand...

Pokken Tournament was sub-HD too; 960 x 720
 
P

Persona7

Banned
Really hope they can get Smash Bros running at 1080p AND provide anti aliasing. Don't see why this would not be possible if the chip is vastly more powerful than the Wii U.
 
I

I Stalk Alone

Member
BuggyMike said:
Considering the dev kits are supposedly overclocked X1s (which use fully ARM processors) I wonder if Nintendo avoided using Denver cores for fear of it being drastically different to work with. I'm not sure how different Denver cores would be to regular ARM cores, but I've seen it suggested that Nintendo may just want to stick with A57. How likely would an ARM A72 be? Are those significantly more expensive than A57 cores?
Click to expand...


Might as well ask for A73's, but I don't know if it will be that customised. They do need something for emulation so maybe keep Denver in there maybe? For those that have a bit of knowledge on this, how customised can the design be away from x1 or x2 base get realistically, especially if a low thermal envelope and low battery drain is crucial? Yes they can get rid of all the redundant car navigation security parts but what can be added?
 
blu

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
BuggyMike said:
Considering the dev kits are supposedly overclocked X1s (which use fully ARM processors) I wonder if Nintendo avoided using Denver cores for fear of it being drastically different to work with. I'm not sure how different Denver cores would be to regular ARM cores, but I've seen it suggested that Nintendo may just want to stick with A57. How likely would an ARM A72 be? Are those significantly more expensive than A57 cores?
Click to expand...
Unless the design was already in advanced stages by the time A72 became viable, there are no viable reasons nintendo would not have preferred A72 before anything else NV could have offered. A72 are better than A57 in every possible metric - power, performance, area.
 
B

BuggyMike

Member
blu said:
Unless the design was already in advanced stages by the time A72 became viable, there are no viable reasons nintendo would not have preferred A72 before anything else NV could have offered. A72 are better than A57 in every possible metric - power, performance, area.
Click to expand...

Thanks. If they do go with A72, I'm curious to see if they'll go full A72, or some variation of A72 and A57.

I Stalk Alone said:
Might as well ask for A73's, but I don't know if it will be that customised. They do need something for emulation so maybe keep Denver in there maybe? For those that have a bit of knowledge on this, how customised can the design be away from x1 or x2 base get realistically, especially if a low thermal envelope and low battery drain is crucial? Yes they can get rid of all the redundant car navigation security parts but what can be added?
Click to expand...

I'm not sure if Denver is more capable at emulation than an ARM A57 or A72. Complete noob here obviously.
 
Schnozberry

Schnozberry

Member
BuggyMike said:
I'm not sure if Denver is more capable at emulation than an ARM A57 or A72. Complete noob here obviously.
Click to expand...

Denver is something of an enigma. Nvidia seems to have found a place for it in cars with deep learning instructions for automotive AI. I have hard time believing it would be used in a gaming scenario, particularly because it draws a lot of power compared to other ARM Cores at a similar speed, and wouldn't provide much benefit over, say, adding a few low power cores to run the OS and background tasks. Parker apparently uses a more advanced version of the original Denver code analysis and conversion, and it now can decide which threads should be placed on which cores, and executes them based on which one is the best for the job. This can apparently be done while all the cores have access to the same pool of cache, and threads can be transferred seamlessly. How much latency this would create in a gaming scenario when everything needs to be immediate I don't know.

I still hope they go with four A72 cores and four A53 cores. It would give them great multitasking and performance, even when undocked.
 
dr_rus

dr_rus

Member
ozfunghi said:
Sorry, it didn't come from Nvidia, but the factory in Taiwan that produces the chips:

http://www.tsmc.com/english/dedicatedFoundry/technology/16nm.htm
Click to expand...

Well, that's some basis to hope for power improvements but in the end it all depends on the implementation. If it'll provide close density but higher clockspeeds then the question is will NV use all the clockspeed for performance benefits or will they opt for some power benefits as well meaning less performance increase but better power efficiency.
 
T

TunaLover

Member
The following info comes from Japanese journalist, Nishikawa Zenji...
Nishikawa Zenji said:
-Switch will likely use the NVIDIA Pascal ‘Parker’ architecture for its GPU
-The Tegra chip inside the platform could help the Switch evolve just like the PS4 and the PS4 Pro
-There are no indications that Nintendo will opt for the ‘Maxwell’ architecture within the Switch
-The custom NVIDIA chip will feature a floating-point performance around 1 TFLOPS
-There is almost “no possibility” that the Switch will perform above 1.5 TFLOPS due to the battery drive inside the Switch
Click to expand...
http://wccftech.com/nintendo-switch-nvidia-tegra-pascal/

It's only a rumor, as always take it as it.
 
I

I Stalk Alone

Member
mitchman said:
1 TF FP16 or FP32 performance, that's the question.
Click to expand...

Says in tha article single floating point. Handheld mode as it goes on about battery drive.


We wont know what this thing will do docked without melting itself ofcourse. But 750gflops handheld is a beast especially considering the architectural advancements. Look at what the wii u could accomplish with 2 screens and 176gflops compared to 240 for the 360.

If they can manage to get this clocked higher in docked mode without issues overheating it would be very impressive considering its power draw.
 
S

Skittzo0413

Member
I Stalk Alone said:
Says in tha article single floating point. Handheld mode as it goes on about battery drive.
Click to expand...

He says single precision which refers to FP32, which is very unlikely to reach 1TF. FP16 is referred to as half precision, so it's likely just either a translation issue or he's misunderstanding something.

If it does make it to 1TF in FP32 that's much higher than any of the previous estimates we've seen.
 
I

I Stalk Alone

Member
Skittzo0413 said:
He says single precision which refers to FP32, which is very unlikely to reach 1TF. FP16 is referred to as half precision, so it's likely just either a translation issue or he's misunderstanding something.

If it does make it to 1TF in FP32 that's much higher than any of the previous estimates we've seen.
Click to expand...

I think he made the same mistake as i did thinking half but sayung single.
 
LordOfChaos

LordOfChaos

Member
viHuGi said:
People afraid this won´t be powerful, lads it is Nvidia doing the god damn chip, Nvidia is known for absolutely INSANE graphics cards and was the company who blamed Ps4 and called it a low end Pc, Switch will be powerfull, more than most think, Nvidia isn´t the type of company that does UP stuff, let alone graphics card :D

http://www.zdnet.com/article/nvidia-calls-ps4-hardware-low-end/

i mean let´s be serious here for a moment, Switch will be the most powered handheld device on the market by HUGE HUGE margins, 720p 60fps on the go, 1080p at home!
Click to expand...

IBM makes the Power8 therefore the Wii U can't be weak!


(I don't think the NX is for a mobile tablet, but this post, this logic :p )
 
M

MisterXDTV

Member
Skittzo0413 said:
He says single precision which refers to FP32, which is very unlikely to reach 1TF. FP16 is referred to as half precision, so it's likely just either a translation issue or he's misunderstanding something.

If it does make it to 1TF in FP32 that's much higher than any of the previous estimates we've seen.
Click to expand...

Wasn't X2 0.75 Tflops in FP32?
 
S

Skittzo0413

Member
MisterXDTV said:
Wasn't X2 0.75 Tflops in FP32?
Click to expand...

X2 (which isn't really a thing, it's called 'Parker') is 768GFlops in FP32, yes. But the Japanese journalist is saying "single float precision (aka FP32) performance around 1 TFlop" which Parker doesn't do. Unless he considers 768GFlops to be "around 1 TFlop"
 
LordOfChaos

LordOfChaos

Member
Zedark said:
Notwithstanding uncertainty about RAM, and I will assume the chips will run clocked to their standard clock which is perhaps possible when it is docked, the power question boils down to three uncertainties:

1. What Tegra chip will be used (Tegra X1 or Tegra Parker)?
2. How does the power measurement method (known as FLOPS) translate between AMD and NVIDIA cards (NVIDIA's FLOPS tend to be 'stronger', i.e. the communication allows for the FLOPS to be realised more efficiently, giving, and I found this ratio in another Neogaf thread, a ratio of approximately 4:3 in favour of NVIDIA, and this translates approximately one on one to an increase in FLOPS as compared to AMD. We want the AMD FLOPS, because we talk about FLOPS for Xbox One and PS4 as well, being 1.31(?) TFLOPS and 1.84 TFLOPS, but they use AMD rather than NVIDIA).
3. There are two ways (actually, there are more, but without loss of much accuracy we can say there are two) to do computations: FP32 and FP16. The first is slower but more accurate and the second is faster (roughly twice, in fact) but less accurate. Finding a balance between accuracy and speed in this method can increase the FLOPS rate (using only FP16 would give twice the number as compared to using only FP32, for example).

Tegra X1 has 512 GFLOPS for FP32, and Tegra Parker has roughly 750 GFLOPS FP32. If we could, for example, use FP16 for 1/3 of all computations, then we gain a 20% increase in FLOPS.* So you see there can be a significant increase. Xbox One and PS4 cannot use this so-called mixed-precision computation (PS4 Pro can, that's where the rumours about PS4 Pro doing 8.4 TFLOPS come from), so the Switch has a potential advantage in power in this regard.

Let's do a calculation: assuming the Tegra X1, we have 512 GFLOPS of NVIDIA FP32. Assuming (with no particular reason for assuming this, but some more technically-schooled GAFfers called it plausible, but it differs on a game-by-game basis) that 1/3 of the computations can be done in FP16 (which, as I mentioned, results in a 20% gain), we can compute the comparison between the two as follows:
Switch power = 512 * 1.20 * 4/3 = 819 GFLOPS. (the 4/3 is that NVIDIA to AMD ratio I mentioned before) This resultant number is how the Switch actually compares in power to PS4 and Xbox One (which are, respectively, 1.84 TFLOPS and 1.31(?) TFLOPS, remember 1 TFLOPS = 1000 GFLOPS).

If, on the other hand, the Switch uses a Tegra Parker, then the power will be (with the usual caveats that we are guessing a lot of numbers):
Switch power = 750 * 1.2 * 4/3 = 1200 GFLOPS = 1.20 TFLOPS. So, you see that using the latter setup, the Switch could be very close in power to the Xbox One. Remember, though, that the gain from FP16 computations is just a guess, as well as the ratio between NVIDIA and AMD FLOPS (the effect, though, is very real, just not numerically determined).

About the clock I mentioned: there is a standard clock value (Tegra X1 has it at 1 GHz), and the FLOPS rate scales linearly with this clock value (so, halving the clock value will half the FLOPS rate). In handheld mode, the clock value will go down as this saves heat production and ergo battery life. In dock mode, however, active cooling could possibly allow the chip to run at full clock speed and therefore allow the power values I determined above.

Disclaimer: The info I produce here is produced by someone who is not a computer engineer (yet), so there might be something wrong in my explanation. If someone spots an error in my explanation (remember, though, that this is purely a FLOPS determination: we simply don't know how RAM and other things will play into the equation), please correct me.

TL;DR/Conclusion: Depending on many factors, the Switch can possibly be very close to the Xbox One for at least a number of games, but that does assume lot of things we simply do not know, and things that often depend on a game-by-game basis. On the other end of the spectrum, though, the power could possibly be roughly half of the Xbox One, so you see there is a lot we do not know and a large margin for errors.

*: See Thraktor's post (#1551) to see how this gain can be calculated.

Edit: For those interested, I converted Thraktor's system of equations into a nice and simple calculation. Take a specific ratio (I will showcase the ratio FP32 : FP16 = 2 : 3). Do the following:
total = 2 * FP32 + FP16 = 2*2 + 3 = 7.
Now divide FP16 by the total:
gain = FP16 / total = 3/7 ~ 0.43. You can check by solving Thraktor's system that this result pans out every time. Here is the proof:
Assume FP32 : FP16 = n : m.
We must prove that FP32 + FP16 = FLOPS * (1 + m/(2n + m) ) (i.e. the new power is the original plus the fractional gain m/(2n +m) which is the fraction I described above).
Thraktor's system says:

FP32 + (FP16 / 2) = FLOPS
FP32 / FP16 = n/m. => FP16 = FP32 * m/n
FP32 + (FP32 * m/n) / 2 = FLOPS
FP32 (1 + m/(2n) ) = FLOPS
FP32 * ( (2n + m) / (2n) ) = FLOPS
FP32 =FLOPS * (2n) / (2n + m).
FP16 + FP32 = FLOPS * ((2n) / (2n + m)) * (1 + m/n). = FLOPS * ((2n) / (2n + m)) * ( (m + n) / n) = FLOPS * ( (2n * (m + n) / (n * (2n + m)) = FLOPS * (2*(m + n) / (2n + m)) = FLOPS * ( (2n + m) / (2n + m) + m / (2n + m) ) = FLOPS * ( 1 + m / (2n + m) ).
And that is what I had to prove (QED, as they say).
Click to expand...



This pits the question squarely on bandwidth. Extending shader throughput with half precision may be well and good, but even with Pascals advanced compression and sort of tile based rendering, 25GB/s is a hard stretch to even XBO levels on just its DDR3, let alone including the eSRAM.

Hopefully the custom Tegra chip includes eDRAM of its own to mitigate the bandwidth gulf, or else another solution.
 
C

Cuningas de Häme

Member
But if he's speaking about the chip in Switch and it really has some additional improvements in it?

Because, after all, it is a custom chip, like the Nvidia has said...
 
M

MikeHaggar

Member
so the division chief where i work just stopped by my desk to show me his latest purchase. it's an nvidia tablet he picked up for $199. i'm assuming this is essentially what the switch is going to be. the tablet is like 8" and 1080p though. i wasn't aware of this product. i'm guessing the switch will be $249.99 max considering the smaller screen size and resolution.
 
S

ShadowFox08

Member
MikeHaggar said:
so the division chief where i work just stopped by my desk to show me his latest purchase. it's an nvidia tablet he picked up for $199. i'm assuming this is essentially what the switch is going to be. the tablet is like 8" and 1080p though. i wasn't aware of this product. i'm guessing the switch will be $249.99 max considering the smaller screen size and resolution.
Click to expand...
249-300
 
You must log in or register to reply here.

Similar threads

  • Locked
Reuters: Switch 2 willfeature a custom Nvidia chip, coming later this year
News
45
3K
Marvel14 replied
(Super Duper Nintendo Switch 2 ?) Nvidia announces a new Tegra with Blackwell architecture to be available next year
48
3K
FireFly replied
Rumor: Nvidia might be working on a PC handheld rival to Steam Deck
Rumor 2
73
4K
CityHunter1981 replied
Rumor | Nintendo Switch 2 May Use Samsung 5th Generation V-NAND With up to 1.4 GB/S Read Speed
2 3
121
10K
Kataploom replied
Nvidia has a new graphics Control panel
Analysis 2 3
125
9K
Dirk Benedict replied
Top Bottom