It'll be whatever hits a good consumer price and profit for Nintendo with passable graphics.
Ahh but remember, this is Nintendo's next Home Console, not handheld.
Nintendo Switch Dev Kit Stats Leaked? Cortex A57, 4GB RAM, 32GB Storage, Multi-Touch.
- Thread starter Robert at Zeboyd Games
- Start date
- Status
It'll be whatever hits a good consumer price and profit for Nintendo with passable graphics.
Maybe some flights do. Most longer flights comes with USB ports though so if you can charge through USB it's a non issue.
So two months before the final hardware went into production developers were still using X1 dev kits? I'm starting to think NS is basically a handheld SHIELD Android TV.
For comparison, Orbis dev kits with the final Liverpool APU were sent to developers in January 2013, PS4 went into full production 7 months later in August.
500Gflops is 2.8x Wii U
And this comparison matters how?
IronicSonic
Member
Jose92
[Membe
Even if it is in docked mode the cooling system should be some super cutting edge technology that is not even used in best laptops, to cool a 1 Tf 32fp gpu in such a compact area. and from the video you can clearly see their is only one vent. When most new laptops with decent cooling system have at least 2 fans 1 for the cpu and 1 for the gpu and several vents to exhaust heat. And dont forget here we are taking about a device of the size of a tablet where heat will get trapped easily.
And i dont think their will be magic sauce inside the docking bay to magically remove accumulative heat from the unit. it is all about thermals, hence it is quite impossible to fit a 1tf 32bit gpu inside such a device unless they use some cutting edge technology not available to the public which is quite improbable because it will increase the price of the unit quite exponentially . I am just being realistic.
Skittzo0413
Member
From the Nvidia Switch thread:
So it's certainly technically possible. Just likely not cheap.
RoadHazard
Gold Member
Underpowered home console. How is that so difficult to understand? They've said the NS is first and foremost a home console (although in reality it is of course a (pretty large) mobile device which can be docked for TV-out).
Jose92
[Membe
That is true, however their is still a 2.4x power deficit compared to the Xbox one GPU.
What do you think? It's a fairly obvious comparison.
PS4 dev kits were on final hardware 7 months before production started.
If NS started production in September, dev kits with the final SoC should have been available months in advance. So why then do the July leaks point to the X1 still being used in dev kits?
Jose92
[Membe
Technically it is possible however, their are still other factors to cosider for the tdp of the whole device. and every component will produce heat, because R=V/I.
So as i said before unless they make it a premium item like the surface pro and charge a premium price, a high performance detachable device is quite improbable.
The NS most certainly is not "finalized" with a Jetson X1 board that I can, right now, go and buy on Amazon. In fact the very notion that a commonly available, non-custom board is somehow the final kit for a custom board is ludicrous. Logic aside, the grammar of the sentence doesn't even make any sense.
All this tells me is PS4 used well established (rather than bleeding edge) tech. Xbox 360 XDKs didn't get final hardware until July/August 2005 (launch in Nov 2005) for comparison.
calc84maniac
Member
The leaks also pointed to them having noisy active cooling, so it's possible they were overclocked to approximate the power of the final chip.
Could be Nvidia is really cutting it close with getting the architecture they're using into production. Or at least I'd like to think it's that cutting-edge.
Sorry, which part are you having issues with?
Skittzo0413
Member
Nvidia has confirmed that the Switch uses a custom SoC. The devkits we know about use stock TX1s. Therefore we have confirmation that the devkits we know about (which could have been sent out even earlier than July, when the report came out) are not using finalized hardware.
If you read Thraktor's post, that should be able to get 1.5TF of 32FP at ~10W, so excessive cooling wouldn't be needed.
The SoC itself would likely be far more expensive than Nintendo wants, but your initial post claimed that it was "quite impossible" to get a 1TF GPU into that form factor, which his post clearly says to be an incorrect claim.
There were also rumours about new kits arriving in August. Your insistence on finding some definitive proof where is none yet is charming, though.
PS4 was using old tech, of course they had them in the kits earlier. It's not that complicated to understand why there can be differences.
Not all systems follow an identical manufacturing roadmap. Again to use Xbox 360 as a counter example, XDKs got final hardware roughly a month before mass production started.
You keep pointing to PS4 to rationalize NS using a stock X1 like the early devkits, but PS4 is entirely irrelevant to this really. BTW where are you getting that NS production even started in September?
Louis Cyphre
Banned
And as a home console it's what?
I'll link you a little story to see how Nintendo works and why it's irrelevant to compare Nintendo's processes with the others'
http://www.eurogamer.net/articles/digitalfoundry-2014-secret-developers-wii-u-the-inside-story
It doesn't matter, the Switch concept is only possible with mobile hardware for obvious reasons. And apparently they're using the best there is.
tkscz
Member
Ah, but the WiiU uses AMD and supposedly Nvidia FLOPS are better than AMD FLOPS (I never could figure out why however. No one really explains it).
Skittzo0413
Member
Ah, after a little digging I think his post was referring to the power draw of the GPU cores alone, whereas the 75 watts for the GTX1050 includes the whole board, fan included.
But it's also obviously a prioritization of cores vs. clock speeds. More cores = more cost but more performance per watt.
I'll admit I'm no expert here, I'm just going by what people a lot smarter than me have been saying.
Standard Jetson X1 board.
Custom Tegra board.
These are non-equatable by the very simple word "custom" which is where the grammar of *my* sentence made no sense when trying to equate them.
FLOPs are FLOPs. What people mean when they make such a statement (which is both inaccurate and somewhat right at the same time) is that historically (and to this day) Nvidia GPU architectures with N FLOPs will perform better over a wide variety of games than AMD architectures with N FLOPs. AMD also requires more memory bandwidth than NV for similar performance.
Of course, this is mostly based on PC comparisons, so it's unclear how much of it is applicable in the console space -- it depends on how much of the difference is down to hardware or low-level software and how much of it is down to drivers. For example, regarding the bandwidth comparison we recently learned that NV have been using a far more efficient rasterization order since the Maxwell architecture.
ROTFLMAO!!! talk about revisionist history 😂
It's true, I got my 280X in a time portal from the future just months before the PS4 launch.
Proelite
Member
They used the latest GCN 1.1 features. Cutting edge is not necessarily most powerful.
WhiteWaterM00se
Member
I'm actually kind of worried that despite everyone's optimism that Switch will not in fact run all Xbone ports.
If true it would leave Nintendo in a tricky situation. Wii U, Vita, 3ds, and PS360 up ports are all most likely easily done. But I'm not sure that is enough to sustain the Switch. Valve games should be easily ported to Switch, if Nintendo can convince Valve.
Or Nintendo is creating a streaming service that streams more complex games. But I'm not sure that will be very popular.
I'm just not sure what to think.
If true it would leave Nintendo in a tricky situation. Wii U, Vita, 3ds, and PS360 up ports are all most likely easily done. But I'm not sure that is enough to sustain the Switch. Valve games should be easily ported to Switch, if Nintendo can convince Valve.
Or Nintendo is creating a streaming service that streams more complex games. But I'm not sure that will be very popular.
I'm just not sure what to think.
The quoted TDP of GTX1050/1050Ti aren't really comparable to the hypothetical Tegra chip I was describing for a number of reasons:
1. They're clocked much higher (~1.4GHz boost clocks)
2. That's a full-board TDP including GDDR5 (not the most power efficient RAM)
3. They're manufactured on Samsung's 14nm process, not TSMC's 16nm process which the rest of the Pascal line is manufactured on (although admittedly there's no particular reason to rule out Samsung's 14nm for Switch)
To illustrate the difference, all you need to do is to look at the 180W TDP for the GTX1080, which uses a GP104 at ~1.7GHz boost clock, and compare it to the mere 36W the exact same chip pulls once you clock it down to a little over 1GHz and remove the RAM.
This is pretty typical for CPUs, GPUs and any other manner of integrated circuits. The relationship between power consumption and clock speed is close to exponential, and if you clock down by only a relatively small amount you can save a surprisingly large amount of power. It's how Intel can sell the exact same die as either a 65W TDP desktop chip or a 4.5W TDP ultra-thin laptop chip.
Principate
Saint Titanfall
If Nintendo cares about that it's possible to get close enough. Operative word being if.
The biggest potential impediment to ports will be how the system sells.
Then the Wii U was cutting edge. It had many modern features.
ZombiePlatypus
Member
Doesn't apply. It's not a regular home console.
Old tech? I'm pretty sure no such APUs existed before they were developed specifically for the consoles.
Also TSMC's 16nmFF has been in volume production for low power parts for over a year, and Maxwell Tegra already had most of the low power optimizations and features (like double rate fp16) before they came to Pascal.
That specific AMD architecture was 1 year old. So the APU was most ready to be produced much earlier in the devkits life.
So you're saying that even if Nvidia didn't have yet some Parker ready to provide to Nintendo for devkits they could have used anything produced by TSMC on the 16nmFF node? Or what's your point here?
Jose92
[Membe
Power consumption is about linear with frequency. And the voltage is exponential in relation to the power.
$P=CV^2f+P_s where P_s is the zero frequency static power dissipation.
However their are many more things and constrains to consider which leads for the huge power increase when frequencies increases by a certain margins, such as node size of the transistor. And as the frequency and power increases the temperature produces by the chip is increased which would mean, the electrons inside the chip are relatively in a higher excited state which would cause the chip to have higher losses, hence more voltage margin is needed when increasing frequency above a certain threshold.
ouble post :SCow Mengde
Banned
So the Switch is cutting edge since it uses Nvidia's newest chip.
I linked that very same article in this thread a couple of pages back. What's notable about it is that throughout the entire development process the dev kits remained almost identical to the target hardware, aside from bug fixes.
If that's how Nintendo works, then we would have seen something long in advance of full production.
BTW a Tegra X1 in the NS is logically consistent with the custom SoC claim since the dev kit was never explicitly identified as a Jetson board. It could be a die shrunk X1 with a few alterations for power saving, and therefore custom in the same way RSX was a custom GPU. Would totally fit in with the way Nvidia works.
A die shrunk X1 wouldn't be a X1 anymore.
normally Nvidia flop for flop vs AMD is generally more effiecient is the better way of saying it. I think the best example someone told me is 2 people walking down a sidewalk from point A to point B. While one walks straight from point A to point B. The other person would walk and stop occasionally. They still get to the same point just one gets distracted and or trip along the way. Well thats how it was describe to me.
Schnozberry
Member
Pascal is essentially a die shrunken Maxwell that allows for higher clocks.
https://www.youtube.com/watch?v=nDaekpMBYUA
If Nintendo went with a "shrunken down TX1", they'd be getting Pascal. At that point it would just be a matter of determining whether they use the extra power overhead to increase efficiency or performance. My guess is they would increase the clocks in docked mode for better performance, and use the efficiency gains while mobile to increase battery life.
They would be fools to stick with TX1 and 20nm.
Cow Mengde
Banned
Can it really do that? I thought the chip was stuck with either one or the other. Like if you choose efficiency, then you're stuck with it, and not something that can be switched back and forth. Unless that is why the system is called Switch.
Skittzo0413
Member
The way I understand it, the "60% increased efficiency, 40% increased performance" line is just a way of saying that Pascal architecture allows for higher performance per watt. The GPU clocks can be changed dynamically depending on what Nintendo wants in any given scenario, and that still doesn't override the performance per watt gain.
My point is a 16nmFF Parker SoC could have been ready to sample much earlier than August, if such a design was in the works, in order to meet Nintendo's timeline.
Also, GCN in the PS4's APU had custom feature sets that weren't in GCN 1.0, not to mention that no AMD APUs at the time even featured GCN GPUs and certainly no APU had ever featured a 1152 shader core part.
Manufacturing larger, power hungry APUs is much more complicated than mobile chips, and attempting to downplay the effort it took to make the console APUs doesn't bolster your argument about the NS SoC being so difficult to sample it had to have come out much later.
- Status
Similar threads
- Featured