Nintendo Switch Tech Speculation discussion

[Shifty Geezer]

Nothing disappointing at all, it's nice, and I think the goal was not to match desktop PC/consoles in a handheld, because that makes no sense, neither from heat dissipation nor battery life perspective.

There was plenty of hope that, when docked, the machine would be console competitive. Hence recent cooling-solution discussion. Not to mention how much processing could be cooled in a handheld (battery life be damned!).

In terms of relative performance, as far as I can find the original shield K1 tablet clocked its 192 Kepler CUDA cores at 750 MHz. So Switch will be about comparable in GPU power to that at best? (double efficiency, half the clock?) Seems very much PS360 (though with better shaders and some smarter tech solutions) is the graphical target, at least regards game complexity. Better lighting and shadowing should make them look prettier.

Also, what do we think about that recent game demo with the insanely clean AA? Is that likely?

 

[Goodtwin]

This is still a really nice piece of hardware for mobile gaming, but as a console there is no way this thing will be seeing ports of AAA games. Even if Nvidias API and tools are awesome, and help developers maximize FP16 shaders and perhaps even include some of the more advanced memory compression techniques found in Pascal, this is still a modest upgrade over the Wii U. In portable mode, it probably performs similar to a Wii in comparison to the Gamecube, and in docked it can run Wii U games in 1080p, but that's probably about the best one can hope for in terms of performance. I'm starting to think some of the theories here on a good deal for a 28nm HPC+ process could be true, actually I think its likely now. I must say though, for such low clocks in portable mode, I am surprised the fan needs to even run. Looks like this hardware is in the same boat as Nintendo portables, Nintendo had better be able to sell it based on its exclusives, because multi plat games are certainly unlikely.

 

[milk]

Makes you wonder what we, the Internet Gestalt, would actually accomplish if our design ideas were put into practice? Would our visions of amazing mobile cooling solutions and sophisticated docks take the world by storm, or would we quickly be bankrupted with over-engineered products no-one really wants?

Very well placed bonus bit of thought. Take the oportunity of when the dreamy guard of the optimists is already lowered by this rude awakening, and give them an extra bit of perspective from which to put their expectations through next time new technology ia about to come.

 

[AlNom]

Remember back when Nvidia said "Porting PS4, Xbox One, and PC games to Nintendo Switch is simple."

Yeah, simple my arse.

Indie titles?

 

[AlNom]

Seems very much PS360 (though with better shaders and some smarter tech solutions) is the graphical target, at least regards game complexity. Better lighting and shadowing should make them look prettier.

The mobile mode seems close enough to be PS360U-level. Tegra X1 does have double the ROPs, so if they kept that even at 307MHz, it's slightly ahead. Whether the bandwidth compression can make up for the seeming lack of shared system bandwidth, well... Then again, they downclock the RAM in mobile mode with an option to keep it at docked-level spec. So who knows if they rebalanced things.

Also, what do we think about that recent game demo with the insanely clean AA? Is that likely?

I think... we should wait for the full reveal trailer.

 

[Lalaland]

Remember back when Nvidia said "Porting PS4, Xbox One, and PC games to Nintendo Switch is simple."

Yeah, simple my arse.

Indie titles?

IIRC it was clarified that he meant that GPU code is common between PC/PS4/XB1 and now Switch so porting that part would be simple but either the reporting or comments were vague enough that it kicked off a lot of false hope that Switch would see AAA multi-plats. Given we're talking Jen-Hsun here I'm going to lean on him being deliberately confusing or vague in his original remarks.

I don't think there will be much simple in porting titles to Switch even for smaller indie titles, I mean even PS4 has framerate troubles with many Unity engine titles (that Stories PS+ game this month drops frames all over the shop).

 

[Shifty Geezer]

Indie titles?

Especially 'target platform == Switch' in Unity/UE.

 

[MrFox]

Makes you wonder what we, the Internet Gestalt, would actually accomplish if our design ideas were put into practice? Would our visions of amazing mobile cooling solutions and sophisticated docks take the world by storm, or would we quickly be bankrupted with over-engineered products no-one really wants?

 

[Cyan]

There was plenty of hope that, when docked, the machine would be console competitive. Hence recent cooling-solution discussion. Not to mention how much processing could be cooled in a handheld (battery life be damned!).

In terms of relative performance, as far as I can find the original shield K1 tablet clocked its 192 Kepler CUDA cores at 750 MHz. So Switch will be about comparable in GPU power to that at best? (double efficiency, half the clock?) Seems very much PS360 (though with better shaders and some smarter tech solutions) is the graphical target, at least regards game complexity. Better lighting and shadowing should make them look prettier.

Also, what do we think about that recent game demo with the insanely clean AA? Is that likely?

So...380 Gflops docked, 150 Gflops in handheld mode?

 

[Shifty Geezer]

Surface Pro rival designed by The Internet. Of course, it's not out yet to evaluate. https://www.indiegogo.com/projects/eve-v-the-first-ever-crowd-developed-computer-laptop-tablet#/

 

[antic604]

Switch is thus a mobile device with TV out, for sure, and nothing like a good console alternative. No sophisticated GPU-housing dock or anything of the sort; just the most primitive solution to the job of portable, dockable gaming device possible.

My thoughts exactly - they've built a mobile device, checked what sort of performance (clocks) they can get targeting specific battery life & thermals and then just increased the clocks accordingly for "docked" mode to facilitate rendering at 1080p as opposed to 720p when mobile. Sure, they could have gone above that for "docked" to enable better AA and AF (I don't think they'll prioritise those for small handheld screen) or even supersampling, but then they probably wanted to keep the process as simple as possible for the developers.

On my end I'm really excited for it - never had a Nintendo console, but if games are good & it is successful, then I'm getting one. I can play AAA 3rd party titles elsewhere.

 

[Goodtwin]

IIRC it was clarified that he meant that GPU code is common between PC/PS4/XB1 and now Switch so porting that part would be simple but either the reporting or comments were vague enough that it kicked off a lot of false hope that Switch would see AAA multi-plats. Given we're talking Jen-Hsun here I'm going to lean on him being deliberately confusing or vague in his original remarks.

I don't think there will be much simple in porting titles to Switch even for smaller indie titles, I mean even PS4 has framerate troubles with many Unity engine titles (that Stories PS+ game this month drops frames all over the shop).

I wouldn't put too much stock in the performance of Unity on PS4, because the game engine seems to underperform in a big way on PS4 compared to other platforms. Nintendo partnered with Unity with the Wii U, and I am sure the relationship is solid. Most Indies will be good to go with this level of performance.

 

[bunnybug]

So I been telling Nintendo fans for months it will based on the tx1 and probably down clocked in portable mode. Been telling telling them all of eurogamer sources say tx1 in final hardware. Nintendo doesn't care about power if they did, it would have been a console.

 

[Lalaland]

I wouldn't put too much stock in the performance of Unity on PS4, because the game engine seems to underperform in a big way on PS4 compared to other platforms. Nintendo partnered with Unity with the Wii U, and I am sure the relationship is solid. Most Indies will be good to go with this level of performance.

I thought the issue with Unity was that it tended to produce highly single threaded program code that assumed way more per core performance that the Jaguar CPU cores provided. Newer versions have apparently alleviated this but if the issue was single threaded perf then that bodes very poorly for ARM based stuff. Of course I'm not a game coder and have very little familiarity with Unity beyond throwing side eye at a lot of Unity titles on my PS4 so I look forward to being told I'm full of it!

 

[BRiT]

We had an actual technology thread before with the DevKit but too much polution by the diehard wishful thinking fanboys reverted it back into speculation mode.

Actual technology thread is now here: https://forum.beyond3d.com/threads/hardware-specifications-of-nintendo-switch-reveal.59806/

 

[Deleted member 13524]

I'll still wait for release to confirm final specs and final judgement, but if these are true then my only guess is: can Nintendo take another Wii U-like hit?


The console probably won't be widely available until mid-2017. IIRC their Q1 sales estimates pointed to 300 000 Switch consoles sold meaning in March it'll probably be available only to a single market (maybe Japan only).
In mid-2017 this will be piss-poorer hardware than the piss-poor it already is.
Vanilla Tegra X1 doesn't mean nvidia made a good deal to Nintendo. To me, it means nvidia yet again sold old tech for a new console design, with minimum effort to boot, and Nintendo should have learned the lesson previously taken by Sony before them. They didn't, probably out of ignorance and greed.

It's rather cutting edge tech really

Lol when the console gets out, the Tegra X1 will have been 2 years old.

All the talk of cooling and docking overdrives and fans, it's simply a case of battery life. Docked, limitless electricity available, clock at cool-and-low-powered.

You're suggesting a 2 year-old SoC for tablets made on 20nm is getting the same performance/power as what a fully custom gaming-oriented 16FF SoC with a recent architecture would get.
We both know this isn't true.




If these specs turn out to be true, then Nintendo learned nothing and the management will get what they deserve. I feel sorry for their software devs, but hopefully they'll be spreading their wings to more capable hardware within a couple of years.

I love Mario, but I'm not getting a whole console + peripherals just to play Mario. I sidestepped the Wii U and this won't be any different.
My guess is some OEM will simply pick a 4-10W Raven Ridge and make a similar "PC mobile console" out of it. It'll be excellent and I'll get one of those instead.




EDIT: Unless this is all coming at sub-$200, effectively making it a 3DS successor. I won't pay more than $180 for this.

 

[MrFox]

Surface Pro rival designed by The Internet. Of course, it's not out yet to evaluate.

Wow. This is actually very nice.

Anybody wants to design a console? I can help for the heatsink/fan ideas and practically nothing else.

 

[bunnybug]

We had an actual technology thread before with the DevKit but too much polution by the diehard wishful thinking fanboys reverted it back into speculation mode. We will be issuing reply-bans if this thread devolves into blind wishful thinking by those unwilling to accept reality.

it's been the same for 3 generation with nintendo fans, reality is very hard for them to accept. we had an article from eurogamer 6 months ago saying developers were pretty damn sure, it would be a TX1 in final hardware. yet nintendo fans chose to ignore it, and just wanted to hear wishful speculation. like i said before before custom could mean anything.

 

[Lalaland]

In fairness every pre-launch console thread attracts magical thinking because until we're told it's made of copper and sand it's all dreams and rainbows, the real question is will Nintendo fandom have it in them to sustain a MisterX level of denial with "dark silicon" and MS Paint fueled delusions?

 

[milk]

For a variety of reasons, consoles have to lock final hardware architecture much earlier than tablets and phones do. They need to be able to manufacture many dev units with equal performance to the final product several months before relese, and need a prety close and reliable estimate of the performance and architecture more than a year before that, something the mobile marker doesn't do because the performance and quality bar for mobile apps is not set as high.
This whole "this architecture is 2 years old" talk is typical of EVERY console release, ps4/xbone included. WiiU, Wii, DS and 3DS were indeed overly archaic, with 5+ year-old choices of architecture inside them, Switch on the other hand is much more within the industry's standards, and it is not so bad really. It's just some of the wilder speculations were too good...