Nintendo Switch

[pulsemyne]

17 minutes ago, mushashi said:

Either I'm going blind or nobody has chosen to post the news from Dean Takahashi (who literally wrote the book on the behind the scenes development of the first Two Xboxes). He's likely correct if he has chosen to publish what his sources told him, very typical Nintendo move, we already knew the power level of the Switch as Nvidia released it already in their own box. Long lead times in development, same as why Sony and Microsoft chose their parts.

It's not too major as maxwell and Pascal are very similar, its just Pascal is a die shrink so can be clocked higher. Now we know that Dev kits used maxwell but we also know that the chip for nintendo is custom so it maybe based of maxwell but have some improvements or some unnecessary things chopped out. Basically until someone gets a die shot we won't really know anything.

[rgraves]

51 minutes ago, mushashi said:

Either I'm going blind or nobody has chosen to post the news from Dean Takahashi (who literally wrote the book on the behind the scenes development of the first Two Xboxes). He's likely correct if he has chosen to publish what his sources told him, very typical Nintendo move, we already knew the power level of the Switch as Nvidia released it already in their own box. Long lead times in development, same as why Sony and Microsoft chose their parts.

Mainly because the article is (was, as it appears to be being edited on-the-go in response to the holes being pointed out) just really terribly written. It says Nintendo went with Maxwell as Pascal would be running hotter for example - which is just incredibly wrong.

 

A poster on NeoGAF gave a pretty good response in to the article in fact:

 

Quote

Originally Posted by Thraktor 

I have a few immediate thoughts after reading through the article:

• Firstly, it's worth noting the difference between Maxwell and Pascal is almost entirely down to the manufacturing process. Maxwell was made on 28nm (and in the case of the TX1, 20nm) whereas Pascal is made on 16nm. The actual architectural difference between the two is minimal, and aside from improved color buffer compression, largely irrelevant for a device like the Switch.
• Despite that, the article never makes any mention of the manufacturing process. I find that extremely strange, as it's obviously the defining difference between the two sets of GPUs.
• In fact, the article gets the difference between the two completely the wrong way around, saying "Nintendo’s box is relatively small, and so it has to fit into the heat profile of a portable device, rather than a set-top box. That’s another reason that explains the older Maxwell technology, as opposed to the Pascal’s state-of-the-art tech." Pascal is literally a more power efficient version of Maxwell, so the incentive would be the other way around.
• The author says "we expect the Nintendo Switch to be more than 1 teraflop in performance", which is notably higher than even those of us who were expecting Pascal were considering (I literally posted earlier today with a 500-750 Gflop estimate). If this is a Maxwell chip, then that would mean at least 4 SMs (512 "CUDA cores") at 1GHz, as they're not going to be able to push much past that on 28/20nm. This is a much larger GPU than most people would have been expecting.

I see a few different scenarios here:

1. The Switch SoC uses Maxwell at 20nm, and simply has a much larger GPU than anticipated to account for the performance.
2. Nintendo looked at the feature-set planned for Pascal when design started, realised that the new features were largely irrelevant, and decided that they would save time and just use a straight-forward die shrink of Maxwell to 16nm instead. That would technically be a Maxwell GPU, but would be almost completely indistinguishable from Pascal in terms of performance.
3. The sources are wrong about Maxwell, the 1 Tflop performance, or both.

A Maxwell Tflop is identical to a Pascal Tflop, and it's largely irrelevant to us whether they achieved that by using a larger Maxwell GPU on 20nm/28nm at a lower clock or a smaller Pascal GPU on 16nm at a higher clock.

 

[Alex W.]

2 hours ago, deKay said:

DIRE WARNING TEAM!

 

Patents != products

 

These patents confirm nothing.

 

Yeah, it wouldn't be the first time Nintendo patented a bunch of possible variations and features that they never got around to. All sorts of things that they'd come up with during development and have no reason not to file patents for.

 

https://www.engadget.com/2012/05/04/nintendo-applies-for-wii-u-patents/

http://www.nintendolife.com/news/2014/02/feature_the_weird_and_wonderful_world_of_nintendo_patents

 

[Bojangle]

None of those are even comparable to these though. They're all batshit ideas. Everything here is well within the realm of possibility, and I'd wager will be stuff we'll see in the final hardware.

[Not Thread Owner]

1 minute ago, Bojangle said:

None of those are even comparable to these though. They're all batshit ideas. Everything here is well within the realm of possibility, and I'd wager will be stuff we'll see in the final hardware.

 

I'm sure many will.

 

But the patent is not proof. That's my point. There's no "confirmation" there.

[Bojangle]

[Alex W.]

10 minutes ago, Bojangle said:

None of those are even comparable to these though. They're all batshit ideas. Everything here is well within the realm of possibility, and I'd wager will be stuff we'll see in the final hardware.

 

Sure, a lot of it is going to be a lock and I'm sure it tells us a lot about the actual system, but some of those patents are for things like a palm recognition system and VR goggles. A bunch of that, like the WiiU Zapper, is going to be stuff they dabbled with in development and aren't necessarily going to go ahead with.

 

 

Weirdly, what people assumed was a Wii Remote style sensor is actually also an Eyetoy-but-infrared motion sensor camera:

 

Quote

 

The right controller 4 includes the infrared image-capturing section 123. The infrared image-capturing section 123 includes an infrared camera for capturing an image in a region around the right controller 4. In the present embodiment, the infrared image-capturing section 123 is used to capture an image of a hand of a user. The information processing device 1 identifies an input made with the hand (e.g., a gesture input, etc.) based on information (e.g., the position, the size, the shape, etc.) of the hand of which the image has been captured. The infrared image-capturing section 123 can rely on ambient infrared illumination, or in one example includes an illuminating section for outputting infrared light. The illuminating section outputs infrared light in synchronism with the timing at which the infrared camera captures an image, for example. The infrared light output from the illuminating section is reflected by an object (e.g., a hand of a user), and the reflected infrared light is received by the infrared camera, thus obtaining an image of the infrared light. Thus, it is possible to obtain a clearer infrared light image. Note that while the infrared image-capturing section 123 including an infrared camera is used in the present embodiment, a visible light camera (a camera using a visible light image sensor) or other image sensor may be used, instead of an infrared camera, as an image-capturing device, in other embodiments.

 

It also has a "share" button.

 

[Bojangle]

Oh yeah, I don't think the VR thing means they're coming out with a VR offering, I think that's a case of them future proofing themselves if VR does actually take off.

[Alex W.]

1 minute ago, Bojangle said:

Oh yeah, I don't think the VR thing means they're coming out with a VR offering, I think that's a case of them future proofing themselves if VR does actually take off.

 

Right, it'll doubtless entirely be based on things Nintendo has actually worked on, and some subset of it is going to be the final product, and some of it might show up later or maybe not pan out. I wouldn't get too excited about some of the less obvious things.

[mushashi]

On 15/12/2016 at 16:11, pulsemyne said:

It's not too major as maxwell and Pascal are very similar, its just Pascal is a die shrink so can be clocked higher. Now we know that Dev kits used maxwell but we also know that the chip for nintendo is custom so it maybe based of maxwell but have some improvements or some unnecessary things chopped out. Basically until someone gets a die shot we won't really know anything.

 

Hmmm, maybe, Maxwell and Pascal are related, but the reason everybody hoped for Pascal was precisely because it can do more work for the same power requirements (ie - it's more powerful as a mobile/console part where TDP matters, unlike on desktop PCs where climate change denial is a religion). You can already surmise the likely differential between a portable and non-thermally/power constrained variant of the X1 SoC when looking at the performance delta between the Google Pixel C and Nvidia's own SHIELD TV micro console.

 

Runtime is going to be highly dependent on how much Nintendo are willing to spend on a massive battery if they are planning to clock it as high as Google did. The PSV had the same basic chip as the equivalent Apple device of the time, but they had to clock it much lower for power usage reasons.

[rgraves]

40 minutes ago, mushashi said:

 

Hmmm, maybe, Maxwell and Pascal are related

There is no maybe, they are totally related. But that's a bit of a tangent, because we also know this is a custom Nintendo chip anyway - Maxwell/Pascal was just the base starting point. Who knows what they did to it, but it's certainly not a stock X1 regardless.

[pulsemyne]

5 hours ago, mushashi said:

 

Hmmm, maybe, Maxwell and Pascal are related, but the reason everybody hoped for Pascal was precisely because it can do more work for the same power requirements (ie - it's more powerful as a mobile/console part where TDP matters, unlike on desktop PCs where climate change denial is a religion). You can already surmise the likely differential between a portable and non-thermally/power constrained variant of the X1 SoC when looking at the performance delta between the Google Pixel C and Nvidia's own SHIELD TV micro console.

 

Runtime is going to be highly dependent on how much Nintendo are willing to spend on a massive battery if they are planning to clock it as high as Google did. The PSV had the same basic chip as the equivalent Apple device of the time, but they had to clock it much lower for power usage reasons.

The thing that gets me is that the article said it was "Based on maxwell". The same argument can be made for Pascal as it's based on maxwell. If the chip in switch is a die shrink of maxwell then it pretty much becomes Pascal.

Anyway as you said it all depends on the battery, although unlike Pixel there is an active cooling fan in the switch so it should keep it's speed up and not be so thermally throttled like Pixel.

[JohnC]

On 15/12/2016 at 15:16, pulsemyne said:

On the Jimmy Kimmel show you could hear the fan working away when docked and it then become silent when removed. It's basically overclocking when docked, probably like GPU's and CPU's "Boost" these days.

Or maybe the fan was put on because it's stuck in a largely enclosed space when docked. 

[bear]

57 minutes ago, JohnC said:

Or maybe the fan was put on because it's stuck in a largely enclosed space when docked. 

That'd mean the Switch is a terribly designed bit of hardware so I hope you're wrong. 

[pelago]

4 hours ago, JohnC said:

Or maybe the fan was put on because it's stuck in a largely enclosed space when docked. 

I still hope that the fan is in the dock itself, for size/weight reasons of the tablet, but it's probably in the tablet. 

[Lorfarius]

On 15/12/2016 at 17:02, Alex W. said:

 

Yeah, it wouldn't be the first time Nintendo patented a bunch of possible variations and features that they never got around to. All sorts of things that they'd come up with during development and have no reason not to file patents for.

 

https://www.engadget.com/2012/05/04/nintendo-applies-for-wii-u-patents/

http://www.nintendolife.com/news/2014/02/feature_the_weird_and_wonderful_world_of_nintendo_patents

 

 

Soon

 

[mushashi]

23 hours ago, rgraves said:

There is no maybe, they are totally related. But that's a bit of a tangent, because we also know this is a custom Nintendo chip anyway - Maxwell/Pascal was just the base starting point. Who knows what they did to it, but it's certainly not a stock X1 regardless.

 

18 hours ago, pulsemyne said:

The thing that gets me is that the article said it was "Based on maxwell". The same argument can be made for Pascal as it's based on maxwell. If the chip in switch is a die shrink of maxwell then it pretty much becomes Pascal.

Anyway as you said it all depends on the battery, although unlike Pixel there is an active cooling fan in the switch so it should keep it's speed up and not be so thermally throttled like Pixel.

 

Being a custom chip is fairly meaningless in terms of potential power, the original Xbox has custom variants of the Nvidia Geforce 3 and Intel's Pentium III (so consoles have been PCs since 2001, and by that logic, the Dreamcast was basically a very big mobile phone ), the PS3 had a custom variant of Nvidia's 7 series GPU and the PS4 Pro has a custom variant of the Polaris chip found in the RX480 desktop GPU, that doesn't mean any of them were as powerful as if they had chosen to go with the next desktop generational design. You can compare desktop Pascal and Maxwell V2 chips at the same TDP to see how big a jump there was in performance between the two (or the performance difference between the K1 and X1 SoCs if you insist on a mobile comparison), or how much less energy and how much smaller a physical chip you can get away with when using Pascal instead of Maxwell for the same performance.

 

If they have used a die-shrunk Maxwell as the base hardware, then the sources wouldn't refer to it as Maxwell, as Nvidia already spent a shitload of money optimising Maxwell for a die shrink/completely different manufacturing method and giving it a new name in the process, it would have been a waste of engineering effort to do that when Pascal has done that for them.

[RubberJohnny]

[wretcherd]

 

[rgraves]

On 12/17/2016 at 17:10, mushashi said:

it would have been a waste of engineering effort to do that when Pascal has done that for them.

I think that's kind of the point though isn't it - the development of the Switch chip, and the development of Pascal probably ran somewhat in parallel. I don't think Pascal was in any way final when Nintendo and Nvidia started with Switch. They both used Maxwell as a launching point as that was what they had at the time.

[rgraves]

2 hours ago, Spainkiller said:

if you are at the one in Frankfurt lets meet for a picture to taunt the rest of the forum

A pox on you both. 

[Adrock]

The closing of the competition was yesterday. When do we hear back?

 

It's highly unlikely I know but I'd love to go to one of the events.

[Captain Kelsten]

Eurogamer have done a piece on the clock speed of a docked and undocked switch and the difference is, according to them, "night and day". 

 

I've only skimmed it and can't say anything I read causes me to worry - I was never expecting power parity with ps4 or Xbox one, let alone the pro or Scorpio, and especially not when used as a handheld. 

 

http://www.eurogamer.net/articles/digitalfoundry-2016-nintendo-switch-spec-analysis

 

[pulsemyne]

30 minutes ago, Doctor Shark said:

Eurogamer have done a piece on the clock speed of a docked and undocked switch and the difference is, according to them, "night and day". 

 

I've only skimmed it and can't say anything I read causes me to worry - I was never expecting power parity with ps4 or Xbox one, let alone the pro or Scorpio, and especially not when used as a handheld. 

 

http://www.eurogamer.net/articles/digitalfoundry-2016-nintendo-switch-spec-analysis

 

A fair bit lower than what I expected but still better than the wiiu even in portable mode (although not by a huge amount). Battery life seems to have been a big concern which is understandable, however letting developers have full access to the system and it's power in mobile mode is very good on nintendos behalf.

[Captain Kelsten]

What concerns me most is the comment about developers essentially having to create two versions of each game - again, I only skimmed through so perhaps I've got the wrong end of the stick - but what developer is going to do that? Seems like a massive cost that will either push the price of switch games up, or result in massive losses for each game. 

[Nespresso]

1 minute ago, Doctor Shark said:

What concerns me most is the comment about developers essentially having to create two versions of each game - again, I only skimmed through so perhaps I've got the wrong end of the stick - but what developer is going to do that? Seems like a massive cost that will either push the price of switch games up, or result in massive losses for each game. 

You mean like PS4/Pro, 3DS/n3DS, and XB1/Scorpio? 

[MojoChin]

2 minutes ago, VN1X said:

You mean like PS4/Pro, 3DS/n3DS, and XB1/Scorpio? 

PC...

[Lorfarius]

17 minutes ago, *chin* said:

PC...

 

Get out!

[pulsemyne]

24 minutes ago, Doctor Shark said:

What concerns me most is the comment about developers essentially having to create two versions of each game - again, I only skimmed through so perhaps I've got the wrong end of the stick - but what developer is going to do that? Seems like a massive cost that will either push the price of switch games up, or result in massive losses for each game. 

They don't have too really. Most engines are very capable of having effects switched on and off on the fly. Also it being tegra based it really shouldn't prove a problem. What will be interesting though is when developers start using the full power in mobile mode. Wonder what it will do to battery life. Maybe the earlier report of people saying it lasted 3 hours in mobile mode was due to it running full speed and then later nintendo and Nvidia decided on the downclocks for the sake of battery life.

[JPL]

Maybe I'm misreading it, but according to that specs diagram, it seems that they can only run the GPU at 307.2MHz whilst in portable mode, but they can choose to either keep it locked to that or up it to 768MHz when docked. I don't understand why they'd keep the GPU crippled when docked though.