Nintendo Switch Technical discussion [SOC = Tegra X1]

[phoenix_chipset]

TX2 is larger and made on a more expensive process.
Plus nvidia is selling those for a premium on the auto space.

It's larger because of the denver cores (well and the larger bus which is frankly needed) which would either be disabled or non present in a switch revision. According to Rich Leadbetter it may be cost prohibitive to stay on 20nm in the not too distant future.
Yes, but when Xavier hits it'll be old news and won't command that premium.

 

[orangpelupa]

so ... the new soc hinted at the new FW, probably just tiiiiiny revision? Like those newer vanilla PS4 that have the exact same body as the launch model, but tiny bit more power eficient

 

[Entropy]

It's larger because of the denver cores (well and the larger bus which is frankly needed) which would either be disabled or non present in a switch revision. According to Rich Leadbetter it may be cost prohibitive to stay on 20nm in the not too distant future.

He is very vague when he talks about it though. With all due respect to Richard, I doubt he knows the minutiae around specific tsmc fab lines, their customers, associated costs of running them, what can be shared over different processes, pressure on alternative nodes...

I think the most relevant observation regarding the Switch and future SoCs is that it has proven itself in terms of volume. Ironically, that means that it is harder to predict what the next SoC could be. Nintendo is not compelled to minimize risk by utilizing an off-the-shelf nVidia product, but could elect to go with a bespoke variant on whatever process makes sense for them.
Depending on your view on Nintendo, this can be seen as both a positive and a negative.

 

[Deleted member 13524]

Yes, but when Xavier hits it'll be old news and won't command that premium.

Xavier is huge at 350mm^2. It's as big as a console SoC and it consumes almost as much. I'd say TX2 and Xavier (similarly Drive PX2 and Pegasus) will probably co-exist for different autonomous vehicle driver levels until nvidia decides to come up with a successor for both at 7nm.

 

[Picao84]

The plot thickens. Why would they need to double the memory to 8GB if the SoC is still pretty much the same? OS / Dashboard new functions?

https://wccftech.com/nintendo-switch-hardware-refresh/

 

[Karamazov]

Video recording ?

 

[tongue_of_colicab]

It can already record now and I suppose the 30 second limit has more to do with storage than ram.

One website mentioned the 8gb might be for dev units. Imo that is the more likely explanation as 8gb by itself isn't really going to change anything and I doubt Nintendo is going to introduce features that can't be used on the original model. It has been out for barely a year.

 

[phoenix_chipset]

8gb + tegra x2 would be amazing, but as someone mentioned just adding 8gb to the base unit would be rather pointless except for faster loading I suppose. Switch has some bottlenecks, but memory amount isn't one of them.

Perhaps we'll see a new switch model late this year? Or spring of 2019.

 

[CSI PC]

Although these things aren’t trivial, 20SoC probably wouldn’t be picked for a new design, something like 16FFC would probably be picked instead. However, for an existing design, staying on a (cheap) process can make a lot of sense. It saves on design expense, respins, wafer cost. Again, it’s not that a new process wouldn’t bring benefits technically, the question is whether a change in process can bring sufficient benefits in the marketplace to more than offset the costs involved, improving profitability over staying put.
I doubt it, until a process like 7nmFF+ (EUV reducing process steps required) has reached some maturity.

But worth remembering they also can get an existing design with the TX2 that is at 16FF.
Sort of why I think there is a case for 16nm and the TX2, along with the fact the gains still mean compatibility with current Switch but still a worthy upgrade for new customers.
It seems the Switch is basically a TX1 not really a new design, I would had thought same principle for Nintendo to adopt the TX2 (for same performance nearly halves power demand or provide better performance with TV), depending upon costs.
It has the same mechanical size as TX1; ignoring the developers kit that is rather cumbersome.

Still, I would not like to say what Nintendo is going to do, like you say it is difficult to predict.

 

[Silent_Buddha]

It can already record now and I suppose the 30 second limit has more to do with storage than ram.

One website mentioned the 8gb might be for dev units. Imo that is the more likely explanation as 8gb by itself isn't really going to change anything and I doubt Nintendo is going to introduce features that can't be used on the original model. It has been out for barely a year.

While the timeframe is short, it isn't like Nintendo are afraid of putting features in a refresh that can't be used on the base model. New 3DS compared to 3DS, for example, featured controller changes (additional analog nub) which was supported in some games that sold for both 3DS and New 3DS.

If they made games scalable, they could potentially use the extra memory for various things. Pre-caching for better load times, better textures (doubtful since I don't think NAND storage is going to increase on devices and higher capacity carts are going to be more expensive), etc.

That said, I agree, 8 GB for a development platform makes a lot more sense.

Regards,
SB

 

[Picao84]

Nintendo is apparently releasing a game engine to use with Switch. Interesting.

https://wccftech.com/nintendo-creates-bezel-engine/

 

[Shifty Geezer]

Can't see why any indie would use this over Unity/UE unless creating a Switch platform exclusive using features not available in the other engines. Business sense isn't there. Unless it's absolutely amazeballs fast and easy to use.

 

[Deleted member 11852]

Can't see why any indie would use this over Unity/UE unless creating a Switch platform exclusive using features not available in the other engines. Business sense isn't there. Unless it's absolutely amazeballs fast and easy to use.

The official description doesn’t say much more about the engine as well, simply saying that it’s part of the publisher’s wish to provide an environment that can allow developers to develop games in a short amount of time while keeping costs as low as possible.​

Rapid time-to-triangle and low cost will appeal to some Indies, where indie games are doing quite well. I'm sure UE4 still has that 5% licensing fee after $3,000 of sales and I was under the impression that Unity has still some way to go to be competitive on Switch in terms of performance and stability. Unity like to boast that 30% of games on Switch are made with Unity and Switch owners know what games they are, low resolution and janky frame rate? That's probably a Unity game. Unity had (still has?) issues on PS4 as well.

 

[Shifty Geezer]

I'm not convinced. MS has offered XNA etc. Sony created PhyrEngine* et al. Indies still go en masse with Unity. Unity isn't great at performance, but it works and has the fastest time-to-triangle possible, and importantly gives you access to so many more targets than just Switch. Potentially two orders of magnitude a larger audience. Plus if Unity's roadmap delivers as intended, performance won't be a concern either and it gives you an engine you can use your experience with elsewhere on other projects.

Nintendo would be better off helping Unity (or Unreal) improve their Switch performance.

* Switch support has been added to PhyrEngine.

 

[orangpelupa]

Probably it will be still as complicated as unity...
Cryengine is simpler but too buggy
Construct 2 is crazy simple and fast but limited to 2D

Hmm.. Probably Dreams will be the simple 3d game "engine" I've been looking for

 

[Deleted member 11852]

I'm not convinced. MS has offered XNA etc.

Until Nintendo show more it'd be silly to write it off. As a day one Switch owner, I love the device but indie titles are hugely inconsistent with their technical issues - and much of this devs themselves lay at Unity's door. I'll believe Unity will fix it when Unity actually fix it, I'm not big believer in jam tomorrow. Whether this will be a full engine and SDK or if it's something that makes it easier to port from another engine for better performance on Switch we'll have to wait and see.

 

[Shifty Geezer]

I can see the desire from Nintendo for better performing indie games. They need to sell it though. As I say, as an Indie, why pick a Switch exclusive game with zero portability? Nintendo would need to make it worth my while. Some titles set out as Switch exclusives would benefit, but it's mostly an option with little appeal from the Indie sidelines where I'm standing.

 

[Deleted member 11852]

They need to sell it though.

It sounds like it's a work in progress, presumably they'll push it when it's done. Maybe, as it was shown at GDC's Osaka's event, they're focussing on Japan's indie scene.

 

[AlNom]

The official description doesn’t say much more about the engine as well

Just following standard Nintendocumentation procedure, amirite.

 

[tuna]

Nintendo would be better off helping Unity (or Unreal) improve their Switch performance.

Unity uses garbage collecting as memory management, that is inherently crap for games because you cant predict performance within a frame.