Nintendo Switch Dev Kit Stats Leaked? Cortex A57, 4GB RAM, 32GB Storage, Multi-Touch.

Status
Not open for further replies.
A 39 watt supply seems about right to me for the alleged specs.

Ideally half of that power would be for charging whilst in use, you don't want the Switch to drain battery too quickly whilst playing even when powered, like many modern devices though you'll see some drain whilst plugged in, but not too much.

Assuming they use a typical (for late 2016) 5V/2.4A for charging, that's 12w (5 * 2.4) of the 39w accounted for.
(Edit: Specs say 5V / 2.6A so 13w, not 12w)

So:

26w: SoC & system, periperhals etc.
13w: Charging

If anything, I'd say the power supply specs confirm the 2SM / 256 Core count.

The AC Adapter is 5-15v, and 15v*2.6a=39w. So perhaps the theoretical maximum of the adapter is for when the unit is using some sort of quick charging.
 
I'm going by the maximum potential power needed. When you have it charging while undocked it will need to power the screen, and could potentially run at the docked clock rates. We don't really know enough at this point.

Also LKD suggested that the top of the screen would have the IR emitters when docked, so it would need a small amount of power at least.

True, though when you have it charging undocked it won't need to power any USB ports or anything else on the dock itself. Not sure about IR emitters (?).

I agree we don't know enough to come to any real conclusion yet.

Does anyone remember how much power the X1 GPU itself draws under full load at 1Ghz, I seem to remember there was a number out there but can't find it now.
 
I don't see why a lower clock speed would suggest 2SM,

You're right. It doesn't as far as a tech reason is concerned. It just seems odd for us to go from assuming it was 2SM with higher clocks to assuming 3 SM simply because the clocks are now lower. Would you need an audible fan at 3 SM at those clocks?

I mean it might be, but clock speed doesn't suggest that at all. Also being Maxwell based was obvious, its development will have started when Maxwell was the new top Tegra GPU.

There is a large portion of this thread that had discussion with the adsumption it would be pascal based. We aren't pretending that isnt the case are we?
 
And many other, if not all big third party releases that require future patches, installs, etc.

It will be the Wii, Wii U, GameCube, etc all over again.

Third parties aren't going to gimp the actual console versions to somehow fit them within the Switches lowest common dominator guidelines.

Sigh.
The Switch is portable
 
You're right. It doesn't as far as a tech reason is concerned. It just seems odd for us to go from assuming it was 2SM with higher clocks to assuming 3 SM simply because the clocks are now lower. Would you need an audible fan at 3 SM at those clocks?

There is a large portion of this thread that had discussion with the adsumption it would be pascal based. We aren't pretending that isnt the case are we?

Lower clock speeds with more SMs is a good amount more power efficient than higher clock speeds with less SMs, when targeting the same performance. That would be the primary reason they might opt for that configuration.

As for Pascal vs Maxwell, the DF article did say they believe some Pascal customizations have been made, so at that point it's really a matter of semantics whether you want to call the SoC Maxwell based or Pascal based. The real question is 20nm or 16nm.
 
Never, ever take input power ratings on AC adaptors at face value. They do not tell you what you think they tell you. They are anywhere between 50% and 3 times higher than the actual peak power consumption (as measured at the wall).

They include reactive power which is important for capacity planning in house installations, power strip daisy chains etc. But that's not what the adaptor "uses".

Nearest available example: standard 65W laptop charger. Rated 180W. What do we do with the 180W figure? We ignore it.
According to the Power Triangle, for a circuit's Apparent Power to be 3x its Active Power, the Reactive Power alone would need to be sqrt(8) = ~2.83x the Active Power. Is that feasible?
 
Lower clock speeds with more SMs is a good amount more power efficient than higher clock speeds with less SMs, when targeting the same performance. That would be the primary reason they might opt for that configuration.

As for Pascal vs Maxwell, the DF article did say they believe some Pascal customizations have been made, so at that point it's really a matter of semantics whether you want to call the SoC Maxwell based or Pascal based. The real question is 20nm or 16nm.
Maxwell 16nm makes very little sense, as that's basically what pascal is.
 
Never, ever take input power ratings on AC adaptors at face value. They do not tell you what you think they tell you. They are anywhere between 50% and 3 times higher than the actual peak power consumption (as measured at the wall).

They include reactive power which is important for capacity planning in house installations, power strip daisy chains etc. But that's not what the adaptor "uses".

Nearest available example: standard 65W laptop charger. Rated 180W. What do we do with the 180W figure? We ignore it.

Surely this only applied to input ratings on power supplies, not output ratings (which is what's being discussed here)?

For reference, here's what's listed in the FCC filing:

AC Adapter input: AC 100 – 240 V, 50 / 60 Hz, 1 A,
AC Adapter output: DC 5 V – DC 15 V, 2.6 A

The input is rated at 100W/240W (depending on voltage), whereas the output is rated at up to 39W. It's the latter that we're discussing.
 
You're right. It doesn't as far as a tech reason is concerned. It just seems odd for us to go from assuming it was 2SM with higher clocks to assuming 3 SM simply because the clocks are now lower. Would you need an audible fan at 3 SM at those clocks?

3SM with lower clocks would be better for performance per watt than 2SM at a higher clock. So there would be every advantage in increasing shader cores and dropping clock frequency rather than just keeping clocks higher.

For the record even with 2SM and rumoured clocks I'm ok with the performance, its about 30% under what I thought in pure processing power. But I still expect some good customisations that will increase real world performance. I just don't think more shader cores is completely out of the question like some do. Some things don't add up to me, especially potential handheld power draw vs the inclusion of a fan,

There is a large portion of this thread that had discussion with the adsumption it would be pascal based. We aren't pretending that isnt the case are we?

No I'm not at all. But there are far too many people misunderstanding what Pascal is vs Maxwell. Pascal is just a slightly customised Maxwell GPU on a 16nm process. I've been pointing that out well before the DF article.

What I'm trying to say is that describing the GPU as Maxwell based unfortunately doesn't tell us anything very specific. It could be telling us that the GPU isn't 16nm (one difference between Maxwell and Pascal) or it could be telling us that the GPU doesn't use any of the few minor changes that Nvidia added to Pascal. Or all of the above. But it doesn't tell us what features it does have, as it is a custom GPU.
 
I want it to be 666

6 cpu cores
6 GB RAM
6 SM

The number of the beast haha.

Spioler: no way.


democat.jpg
 
If the base model ends up being $250, and the bundle gets more storage and costs $300 with a game(lets say splatoon or botw), how much more storage space can we really predict getting? I'm guessing 64GB-120GB, with 64GB the most likely.
Not a formal guess, but for sake of discussion: the only other time there was such a disparity (more expensive model with game and more storage) was Wii U, and in that case space quadrupled from 8 to 32 GB.
 
Lower clock speeds with more SMs is a good amount more power efficient than higher clock speeds with less SMs, when targeting the same performance. That would be the primary reason they might opt for that configuration.

That is what I figured. I was wondering what the means about the fan in the system internals. Anyone have an idea what the point is? 3SMs is basically more expensive right due to material? The fan is throwing ne off because it doesnt seem necessary at all.

As for Pascal vs Maxwell, the DF article did say they believe some Pascal customizations have been made, so at that point it's really a matter of semantics whether you want to call the SoC Maxwell based or Pascal based. The real question is 20nm or 16nm.

I wish they clarified the process.

3SM with lower clocks would be better for performance per watt than 2SM at a higher clock. So there would be every advantage in increasing shader cores and dropping clock frequency rather than just keeping clocks higher.

For the record even with 2SM and rumoured clocks I'm ok with the performance, its about 30% under what I thought in pure processing power. But I still expect some good customisations that will increase real world performance. I just don't think more shader cores is completely out of the question like some do.

More cores I dont think is out of the question either. I just am not betting on it. Roughly how much would it add to the cost? After my initial "fuck nintendo gimping their shit" meltdown I pretty much cam back and realized the specs are mostly fine.


No I'm not at all. But there are far too many posts misunderstanding what Pascal is vs Maxwell as well. Pascal is just a slightly customised Maxwell GPU on a 16nm process. I've been pointing that out well before the DF article.

Yeah I understand this. Pascal is basically a 16nm process with some adjustments.

What I'm trying to say is that describing the GPU as Maxwell based unfortunately doesn't tell us anything very specific. It could be telling us that the GPU isn't 16nm (one difference between Maxwell and Pascal) or it could be telling us that the GPU doesn't use any of the few minor changes that Nvidia added to Pascal. Or all of the above. But it doesn't tell us what features it does have, as it is a custom GPU.

Okay I gotchu now.
 
That is what I figured. I was wondering what the means about the fan in the system internals. Anyone have an idea what the point is? 3SMs is basically more expensive right due to material? The fan is throwing ne off because it doesnt seem necessary at all.

I think the fan is throwing everyone off.
 
Do you have a source for that? I thought the console itself was supposed to run at around 32-35W max, but I didn't know that was accounting for the USB power transferred.

If true, nevermind!
Vague memories from the analysis thread. I don't have the energy to try to look it up.
 
Is it really so strange it has a fan for docked mode?

Yes, because according to people here it's questionable that the docked clock speeds would need any cooling* whatsoever.

“Nintendo Switch is a home gaming system first and foremost,” - Nintendo.

Interesting.

First of all, that's only NoA. Second, that's a marketing message which is specifically designed to ensure that people know this will get full console games, rather than "console-like" games that Vita got.

Third, a marketing message doesn't define what a product literally is.


*EDIT: Active cooling I should say
 
Not a formal guess, but for sake of discussion: the only other time there was such a disparity (more expensive model with game and more storage) was Wii U, and in that case space quadrupled from 8 to 32 GB.

what are you talking about? there were xbox 360 s with 250 GB HD and others without HD and having 4 GB internal memory
 
Because the fan works more efficiently that way. Makes sense to have the fan right next to what you are cooling.

Introducing a fan into the internals of a portable device is a pretty huge deal, from the perspective of design, cost and warranty. It's not something they would do if there was any other way around it, even if it might be more efficient.

It's clear that- if there is a fan in the final hardware- then they had no other choice, which would suggest that the fan wouldn't just be for docked mode. Now, if they have the ability for it to run at docked speeds when portable then that could explain it. But as it is it's a very strange inconsistency based on the specs we've heard.
 
Because the fan works more efficiently that way. Makes sense to have the fan right next to what you are cooling.

More efficient when it's on the chip producing heat. Wouldn't need to spin as hard as if it were external.

Introducing a fan in a portable system is a big deal. If the fan breaks in a dock its easy to replace the dock. If it breaks in the handheld or is defective it is much more work to fix. The clock rates in dock mode probably dont need a fan at all let alone right over the chip. Why would you complicate the design of your system like this? How many protable electronics do you own that have fans?
 
The most efficient solution is throttling. But that's not an option for a gaming device.
Then you limit the clock speeds of your main processing components in order to generate less heat just like every previous handheld has done before.
 
What if the final unit has no fan!?


:P

Everything would make more sense :P

Nintendo does odd design choices so maybe its just precautionary but it really strikes me as weird for a handheld to have a fan. They downclocked the shit out of the portable mode. Docked isnt crazy either. Why put a fan in there.

And I dont think there is some extra power thing to it either. I still think it is 2 SMs. That is why the fan just makes no sense.
 
What if the final unit has no fan!?


:P

I'm beginning to think this is a real possibility. Adding SMs and CPU cores seems like it might be unnecessary to get any third party support, but maybe its not. But if they have removed the fan from the final unit that would make a lot of sense with these clock speeds. Then again the final unit shown on Jimmy Fallon had vents, at the very least.

Going 16nm could've helped there too. It'll be very interesting to see a teardown eventually.
 
Sure have fun. I just seen a lot of people imply having a fan in the unit is a sign of bad engineering. I won't really go along with that one.
I don't think that's the point some people were trying to make, including me.
 
Sure have fun. I just seen a lot of people imply having a fan in the unit is a sign of bad engineering. I won't really go along with that one.

If you're referring to posts like mine, I'm in no way saying it's a sign of bad engineering. I'm saying that if there is an internal fan then they really had no other option than to put it there to get the type of performance and cooling that they targeted. Which would indicate higher performance than rumored by Digital Foundry, like potentially more SMs or CPU cores.
 
Is it really so strange it has a fan for docked mode?

No it wouldn't be strange if that was the case, as in a fan inside the dock. Probably still unnecessary but a relatively cheap precaution so why not. But it isn't, the fan is inside the hanheld itself. It runs in portable mode, it just runs at a higher RPM in docked mode.
 
Is there any electronic device where the fan and the chip it's cooling are separated? I can't think of one.

In the case of Switch, it seems like it would be difficult to force air through a tiny rectangular opening at the bottom of the console. Much easier if the fan is inside, pulling air in and moving it out the top.
 
Because the fan works more efficiently that way. Makes sense to have the fan right next to what you are cooling.

It would be quite a risk putting a fan in a handheld, if they want better cooling just run it faster in the dock. Anyway I'm pretty sure I remember the patents actually saying outright that the fan runs at one speed in handheld mode and then faster RPM when docked.
 
Is there any electronic device where the fan and the chip it's cooling are separated? I can't think of one.
I can't think of one either.

Forgetting for a minute that it would be crazy to risk putting a fan in a handheld just for very slightly more efficient cooling (just run it faster in the dock..). I'm pretty sure I remember the patents actually saying outright that the fan runs at one speed in handheld mode and then faster RPM when docked.
Are you sure it's just slightly more efficient to have it in the unit?

Running fans at higher speed = more noise. Nintendo hates noisy hardware.
 
I will just assume actual engineers at Nvidia / Nintendo are a hell of a lot more competent than armchair engineers at gaf :p

I'd agree Nintendo know what they're doing and don't do stupid things like including fans for basically no reason (they've never included a fan, ever in a handheld) and Nvidia absolutely know what they're doing with graphics hardware. That's why the fan clearly points to something being a miss with the assumption that Switch uses a standard down-clocked X1 (well that and the fact we know it doesn't anyway since its a custom chip..).

That doesn't mean it has to be 3SM or whatever, just that its more than just a down-clocked X1.
 
Sure have fun. I just seen a lot of people imply having a fan in the unit is a sign of bad engineering. I won't really go along with that one.

I dont really think it is bad engineering at all. It is more that we dont have enough info right now to come up with a good reason with the readily available sources. Which really just means we need more info.

Info could come out tomorrow that makes it all make sense. At least that is my take.
 
Status
Not open for further replies.
Top Bottom