Nintendo Switch Tech Speculation discussion

"With that in mind, we can't help but wonder whether X1 is the final hardware we'll see in the NX."
Eurogamer and their source said about the development kits in the article. It was July and no source had info about the final hardware.

 

that's pure speculation on there part, there sources were pretty damn sure that x1 will be in the final hardware, and they clearly state that in the article and the video, watch the video at 6.18 to the end.

 

They mentioned dev kits were running on X1, and several other sources have also mentioned this.

They also speculate that Pascal could be the reason as to why the NX/Switch was headed for a March 2017 schedule.

The only thing confirmed so far have been dev kits.

 

"It is worth stressing with a firm emphasis that everything we have heard so far points to Tegra X1 as the SoC of choice for Nintendo NX"

THEN WATCH THE VIDEO, HE'S NOT TALKING ABOUT DEV KITS BUT FINAL HARDWARE.

 

is eurogamer

 

As Nvidia stated, its a custom SoC, so we can discard vanilla X1. Of course, it may be based on X1 but it also may be based on Parker (without the automotive stuff and with regular ARM cores like A72). In any case, it will need lot of power/battery optimizations for the intended portable usage.

 

nvidia said it was based on their top performing graphics card. that would indicated pascal. i think thats what theyre going with. at this point what advantages would using an older part afford them?

 

That's kind of how I see things too. X1 performance is already fine for a handheld and as you say differences between Maxwell and Pascal seem to be small.

Process is more important as it'll impact on clocks for both CPU and GPU and also impact battery life.

X1 on 16nm FF should deliver some really nice results.

 

My concern is that 16nm is still, and will most likely remain, more expensive than earlier process nodes. Moore's law really is nearing its end.

Nintendo is very cost sensitive.

Sent from my Nexus 6P using Tapatalk

 

Maybe they were talking about their top performing card in sales?

 

Don't read into it too much. It doesn't necessarily mean top performing compared to their other products that they offer. Language like this typically means top performing relative to their competitors, which we know can be subjective depending on what is being measured.

Sent from my Nexus 6P using Tapatalk

 

Sorry, I missed the smiley...

Sent from my Nexus 6P using Tapatalk

 

So the new Xbox One S and Playstation 4 Slim went shrunk their chips with Finfet 16nm to make it more expensive? Come on people, the smaller process leads to cost savings. It makes no sense what so ever to use anything other than the industry standard 16nm Finfet process. This isn't the Wii U with its pools of edram that inhibited dye shrinks. Maxwell or Pascal I really think matters very little. The Tegra X1 was already based on 2nd generation Maxwell, which introduced many of the benefits that became standard with Pascal.

This "source" has no proven track record. This is likely click bait based on previous rumors surrounding the Tegra X1 being the chip in development kits. Unless its Eurogamer, Emily Rogers, or Laura K Dale, then I chalk it up as nothing more than speculation, even if its plausible speculation.

 

Most important fact is memory bandwidth. If they use tegra x1 64bit lpddr4 then RIP

 

As mentioned before, this IS Nintendo who aren't exactly known for using the latest hardware available...

 

I wouldn't be surprised if the SOCs themselves were more expensive, but the improved thermals and integration allowed them to cut costs elsewhere.

Sent from my Nexus 6P using Tapatalk

 

The XB1S SoC costs more despite being much smaller. It was estimated at $100 versus $75 for for the last 28nm version they made.

But for a portable device, power consumption is extremely important, I agree it makes little sense to not use 14/16nm.

 

I don't know how much I rate Venture Beat as a source for Switch being Maxwell based but it fits with my pre-existing biases towards the part given Nintendo historically being so cost focused and a need to tie down the design a year or more prior to launch. I think bigger issues for the device is, as other have said, process node and RAM setup other issues that come for free from working with Nvidia are production quality and poor battery life (look at all of those 'wins' they've enjoyed with Tegra thus far).

On battery life with Nvidia own tests showing a roughly 2 hour battery life on the Shield Tablet (https://shield.nvidia.com/blog/shieldbattery) are we looking at a 4-5 hour battery life for Switch? I'm assuming a cumulative 150% boost from dedicated O/S, process shrink, smaller screen and design improvements which seems aggressive to me but I'm open to be persuaded in either direction.

On the expansion base it seems a bit odd to me that a simple plastic sleeve can offer enough of a boost to get the unit to offer 1080p when it seems to be 720p native. Does the idea that an extra fan in the base (or an internal fan that only runs/revs up when docked) could allow an up- clock significant enough to drive games at 1080p ring true?

 

the fact that x1 and parker are going to perform similarly on the gpu side would just make the dev kits very similar to final hardware as well.

what im wondering though is to what degree they are customizing the soc, the automotive computer vision stuff is def useless, but the cpu(s) could be altered as well.

although nvidia has put alot of effort into what theyre calling Heterogeneous Multi-Processing nintendo could just cut the a57 cores out completely and go with just the denver cores, and if you consider that as a possibility then the option of a larger die and more sms becomes similarly possible!