• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

What we know so far about the Nintendo NX with sources

Status
Not open for further replies.

10k

Banned
just be region free

THATS ALL I WANT!!!!!!!
3vKJqc7.gif

I dont know why people were wondering about that. even the Wii is more powerful than the gamecube (probably by the least amount in console history, but still) , and the Wii U isn't exactly a high bar to beat. Thats like confirmation that the sky is blue.
Because someone posted a thread saying what if it was weaker?

Don't forgot there were morons who legit thought the Wii U was an add on for the Wii :p
 

AyanFaust

Banned
just be region free

THATS ALL I WANT!!!!!!!

3vKJqc7.gif


Because someone posted a thread saying what if it was weaker?

Don't forgot there were morons who legit thought the Wii U was an add on for the Wii :p

"My understanding is that it is not realistic on the existing video game systems because unlocking them after they are already on the market poses a number of hurdles. On the other hand, regarding NX, we understand that many consumers hold such opinions and such suggestions exist in the market, and although we have nothing concrete at the moment, we are internally analyzing what hurdles exist to lifting region locks. That is the situation right now. We acknowledge your request, and I personally want to give it positive consideration."

It's at least under consideration? Maybe it'll finally be region free ;-;
 
:)

So each one is aboot... 1.8% of the die, so ~11mm^2 per 1024-bit.

On Durango, each DDR3 128-bit interface is about 17.5mm^2.

GDDR5 interface on PS4 is about 47mm^2 total.

That's much smaller than I expected! I'm hoping for HBM but it would probably be cost-prohibitive. That is unless they went with 28nm for the SoC. I know Carrizo is reported to draw ~15w. Is this one of those binned laptop parts? Could they count on such a power draw in a mass produced console? If those numbers are legit, a Carrizo w/ some HBM and ARM cores instead of the Excavator modules may be just what the doctor ordered.

The other scenario I was envisioning was something like the Wii configuration, with a larger chunk of HBM and then just one or two external DDR4 chips being mainly there for OS functions. That would make for a very simple PCB, although again. the cost.
 

Deku89

Member
"My understanding is that it is not realistic on the existing video game systems because unlocking them after they are already on the market poses a number of hurdles. On the other hand, regarding NX, we understand that many consumers hold such opinions and such suggestions exist in the market, and although we have nothing concrete at the moment, we are internally analyzing what hurdles exist to lifting region locks. That is the situation right now. We acknowledge your request, and I personally want to give it positive consideration."

It's at least under consideration? Maybe it'll finally be region free ;-;

I think it'll be region free in name only. Meaning, people will have so many hoops to jump through to access other regions, it might as well still have region locks. Hopefully they do away with it though and make it easy to work with.
 

McHuj

Member
Have we discussed LPDDR4 yet? That's something that might be suitable for the NX handheld as well as the NX console (if they are going the very small/low power route, as I believe they are). These modules from Samsung are faster than the desktop variety currently available. Is the pricing prohibitively expensive?

I think LPDDR4 maybe a possibility. Yes, it maybe costlier now. However, it will be the RAM most of the mobile world moves too in the coming years so I think costs are going to plummet over the next couple of years. It's probably a question of whether Nintendo wants to eat the costs for a year or so.
 
Anyone care to speculate? Is this gonna have a weird gimick thats gonna make me go "oh man! lame!" is is it gonna be like xbox and playstation? (a box with a controller and 3rd party support). If nintendo made a console like playstation with mature 3rd party titles ect, AND Nintendo games, i think it would do cool.
 
Anyone care to speculate? Is this gonna have a weird gimick thats gonna make me go "oh man! lame!" is is it gonna be like xbox and playstation? (a box with a controller and 3rd party support). If nintendo made a console like playstation with mature 3rd party titles ect, AND Nintendo games, i think it would do cool.

The Gamecube didnt cool and had all of this...in the end they lost more and more 3rd-Party support but in the first 2 years all looked good.
 
I was imagining some kind of Apple TV type situation where the mobile device feeds content into the console, giving it expanded features/graphics/etc.

Though now that I read through that stuff my hope is that the "handheld" aspects of the NX ecosystem is just mobile. As in, the new platform is a home console but you can use your mobile device in conjunction with it.

I dunno, I just feel like incorporating some kind of standalone handheld element to the NX while ALSO offering some kind of support for iPhone/Android use is going to feel redundant. I'd much rather have a juiced up home console that extends access to games and accounts to mobile and PC and skipping handheld/DS hardware altogether.


EDIT: Actually my real wish is that Nintendo creates a network ala Steam and offers it to be licensed on other platforms, though I know they're not giving up on the hardware business.
 

AlStrong

Member
That's much smaller than I expected! I'm hoping for HBM but it would probably be cost-prohibitive. That is unless they went with 28nm for the SoC. I know Carrizo is reported to draw ~15w. Is this one of those binned laptop parts? Could they count on such a power draw in a mass produced console? If those numbers are legit, a Carrizo w/ some HBM and ARM cores instead of the Excavator modules may be just what the doctor ordered.

hm... yeah, I believe they are binned, but they can always change clocks/voltage last minute, plus 28nm is extremely mature, so yields should be quite good especially for a relatively modest die size (244mm^2).

I suppose they could swap out for a Jaguar/Puma to keep it as close to the current gen twins for inexpensive ports.

I'd hope they would at least try for a similar power consumption as WiiU at the minimum. lol

The other scenario I was envisioning was something like the Wii configuration, with a larger chunk of HBM and then just one or two external DDR4 chips being mainly there for OS functions. That would make for a very simple PCB, although again. the cost.

I get the feeling HBM would just not be cheap enough at least for another year while IHVs deal with their next wave of high-margin GPUs that are relatively low volume compared to console.

We know how costly the 32MB SRAM is on Durango, but if they were "smart", they'd make it a part of the cache hierarchy not unlike Crystalwell to avoid mandatory buffer management. Hard to say if AMD can design it well. ;)
 

QaaQer

Member
What also seems to be a huge departure is the use of a virtualized dev environment. Like Android and iOS, software development will be done in a common way, regardless of what's in the box.

What does a non-virtualized dev environment look like compared to a virtualized one? What are the costs and benefits?
 
Even at a basic level, they could do some pretty impressive things with it. And in one respect they've had the technology for years-

That target program looks way better than any 3D I've seen on a 3DTV or in theaters. That's some truly awesome tech, and something I would love to pay for if it worked properly.
 
At best it'll be a little tidbit like a major western dev having a dev kit or something. Still at least it'll be interesting to talk about while at work :)

Indeed!

Edit: Really hoping the 2016 GameStop survey mentioning NX means something is finally about to happen.
 
Indeed!

Edit: Really hoping the 2016 GameStop survey mentioning NX means something is finally about to happen.

Something is certainly about to happen, as Nintendo has repeatedly said that more details will be shared in 2016. What we don't know is whether or not they are going to wait till their FY17 (starting April 1st, 2016).

My guess is that there will be a big marketing splash for:
- Fire Emblem Fates
- Pokken
- Star Fox Zero
- Miitomo
& (maybe) a tease for NX
 
Something is certainly about to happen, as Nintendo has repeatedly said that more details will be shared in 2016. What we don't know is whether or not they are going to wait till their FY17 (starting April 1st, 2016).

My guess is that there will be a big marketing splash for:
- Fire Emblem Fates
- Pokken
- Star Fox Zero
- Miitomo
& (maybe) a tease for NX

Hopefully no later than March, as Jose said on the IGN chat, the reveal needs to happen in the next few months, Game Informer online said the same thing.

Edit: That marketing break down makes sense for what they might be willing to show in Feb/Mar IF it happens. What happened to our regular Directs? Too much radio silence ever since Iwata's passing (and no, as good as they are the Nintendo Minute segments don't make up for it).
 

Neoxon

Junior Member
Hopefully no later than March, as Jose said on the IGN chat, the reveal needs to happen in the next few months, Game Informer online said the same thing.

Edit: That marketing break down makes sense for what they might be willing to show in Feb/Mar IF it happens. What happened to our regular Directs? Too much radio silence ever since Iwata's passing (and no, as good as they are the Nintendo Minute segments don't make up for it).
We did get a general Nintendo Direct in November & a Smash Direct in December, so the Big N hasn't been completely silent.
 
Hopefully no later than March, as Jose said on the IGN chat, the reveal needs to happen in the next few months, Game Informer online said the same thing.

My issue with IGN and Game Informer is that these firms are generally not in the business of being insightful market analysts, nor do they have much stake in Nintendo's financial position. I'm not really about to take their 'needs to happen' recommendations too seriously.

For example, Nintendo is certainly going to be careful with their NX messaging because they don't want to completely lose their existing lines of business; the classic "why buy, when it's about to be different?". (Hence my comment of FY16 vs FY17 since their revenue/opportunity costs will be booked in different periods.) This is something that neither IGN nor Game Informer care about.

That being said, I do think that if NX is to drop (in some form) in 2016, Nintendo would be wise to build momentum ahead of E3. It's a framework that worked well with Sony, so perhaps it should be emulated. That being said, one should be careful to judge causation too quickly without considering that Sony and Microsoft were both performing their song/dance in parallel. Thus perhaps Sony's ability to pivot their messaging post-Microsoft's has greater correlation to success than just timing. Food for thought.
For what it's worth, I'd personally love to have heard about it like yesterday :)
 
R

Rösti

Unconfirmed Member
The wheels are in motion, though initiated by someone close to the source, not the source itself (at time of writing). Not sure about presence in office.
 

10k

Banned
Rösti;193253666 said:
The wheels are in motion, though initiated by someone close to the source, not the source itself (at time of writing). Not sure about presence in office.
All this work to confirm that the NX will in fact play games. Smh. :p
 

Broken Joystick

At least you can talk. Who are you?
Rösti;193253666 said:
The wheels are in motion, though initiated by someone close to the source, not the source itself (at time of writing). Not sure about presence in office.
Wheels?
Mario Kart confirmed.
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
So let me ponder a bit on my favourite (rumoured) NX aspect, slow monday and all.

SCD, aka supplements.. That thing can fly. Especially when configured in a local network with the main unit. I've been experimenting lately with such setups for a pet project of mine, using highly asymmetrical protocols (and bypassing TCP/IP altogether but that's another story) where the supplement has all assets in advance (in my case - procedural assets), and the main unit sends out a single beacon/sync message when ready to accept during a frame, while in the meantime the supplement has finished or is about to finish its work, and at said beacon just sends away all done computations for the frame, and then immediately starts with the next frame. Given a proper heuristics for the work split between main and supplement, the two nodes achieve perfect utilisation. And just to give you a clue how asymmetrical it all is (in packets) - it's kilobytes from the main-unit side vs gigabytes from the supplement side (over a GbE link). In my case the workload has seen perfect linear scaling thanks to the precise workload split between the nodes.

So, two things make for a great distributed system:
  • running a minimal overhead sync protocol over a low-latency network connection
  • optimal workload split between the nodes, so no idle time on either end.
Satisfying those yields perfect (read: linear) gains, aka profit : )
 

Thraktor

Member
Have we discussed LPDDR4 yet? That's something that might be suitable for the NX handheld as well as the NX console (if they are going the very small/low power route, as I believe they are). These modules from Samsung are faster than the desktop variety currently available. Is the pricing prohibitively expensive?

I've been pondering what type of technologies they might decide on. My tentative conclusions are that they could go with a 14nm Soc or HBM, but not both. As people have mentioned, if they go with 14nm, it will likely be a small chip. AMD showed off a Polaris GPU and it was estimated at ~120mm2. Maybe Nintendo could squeeze a bit more surface area on there, but not much. The other point which may be relevant is how large an area the PHY for HBM requires. If each block of HBM is a 1024-bit interface, I would imagine it would require a large area on the SOC. It's too bad we don't have any die photos, but we do know that all the chips so far w/ HBM have been monsters in size.

On the home console RAM front, I can't really see the reason for LPDDR4 in the home console. DDR4 is relatively power efficient as it is, and if LPDDR4 were competitive in price with DDR4 for the same capacity and bandwidth, there wouldn't really be a need for DDR4 to exist.

Handheld RAM may be interesting, though. They went with specialist FCRAM for the 3DS, and my thinking was that they may go a little esoteric again with something like Wide I/O 2 (which is on-chip like HBM). Whether they actually need the extra bandwidth over LPDDR3/4 is another thing, though, and like the home console, it depends on whether there's any dedicated VRAM alongside it.

I also wouldn't see any reason to expect 14nm for the home console, unless they were locked into some FinFET-dependent tech early on in development (i.e. Zen or Polaris). With the expected performance levels we're looking at, a 28nm design would still be relatively power-efficient in the scheme of things.

:)

So each one is aboot... 1.8% of the die, so ~11mm^2 per 1024-bit.

On Durango, each DDR3 128-bit interface is about 17.5mm^2.

GDDR5 interface on PS4 is about 47mm^2 total.

And, if we're comparing expense against on-die SRAM, then 32MB of SRAM would occupy about 68.5mm² at 28nm. Of course physical interfaces should be cheaper than SRAM even at the same size, but it comes into play when the calculation is:

Cost of SRAM vs Cost of HBM + on-die interface + packaging

The other thing is latency, though. Nintendo has historically gone with low-latency RAM solutions (1T-SRAM, FCRAM, eDRAM), and SRAM will obviously provide the best latency possible, but modern graphics architectures just aren't that latency dependent, so are we at the point where Nintendo would be willing to sacrifice latency for bandwidth and capacity with something like HBM? (That said, HBM should still provide lower latency than the DDR/GDDR options.)

That's much smaller than I expected! I'm hoping for HBM but it would probably be cost-prohibitive. That is unless they went with 28nm for the SoC. I know Carrizo is reported to draw ~15w. Is this one of those binned laptop parts? Could they count on such a power draw in a mass produced console? If those numbers are legit, a Carrizo w/ some HBM and ARM cores instead of the Excavator modules may be just what the doctor ordered.

The other scenario I was envisioning was something like the Wii configuration, with a larger chunk of HBM and then just one or two external DDR4 chips being mainly there for OS functions. That would make for a very simple PCB, although again. the cost.

I don't think Carrizo is really a decent point of comparison. Those Excavator cores are just going to chew up power and screw up any comparison. Besides, console APUs are going to devote a lot more of their die area to the GPU than any PC APU would, so to be honest you're better off looking at GPU dies and adding some ARM cores to them.
 

Neoxon

Junior Member
Rösti;193253666 said:
The wheels are in motion, though initiated by someone close to the source, not the source itself (at time of writing). Not sure about presence in office.
So basically no news yet, but we should expect it soon.
 

Thraktor

Member
Also:

I'm gonna go with a 14nm chip w/ 512 GCN ALUs, maybe a combo of 4 A57(or A72 we can only hope) + 4 A53, and the 32 MB pool of eSRAM.

It's worth noting that A72 cores actually take up 10% less die space than A57s on the same node, so unless the design was locked before A72s could be added it wouldn't make much sense to use the A57.
 
What does a non-virtualized dev environment look like compared to a virtualized one? What are the costs and benefits?

A non-virtualized dev environment = a physical machine that you run games on. What people normally think devkits to be.

A virtualized dev environment = think emulators. Now imagine that the emulator has multiple preset performance profiles - correlating to a different target hardware.

The benefits of using a non-virtualized dev environment is that a dedicated physical machine is better indicative of actual performance in real-world environments. The drawback is that you need a separate machine for each target hardware, and that only one person can be using a machine at a time.

The benefits of a virtualized dev environment is that it allows much easier setup and testing of different target hardware. The drawback is that the performance of the virtual environment might not accurately reflect real-world environments (although this can potentially be mitigated by having dedicated hardware).
 

KingSnake

The Birthday Skeleton
So let me ponder a bit on my favourite (rumoured) NX aspect, slow monday and all.

SCD, aka supplements.. That thing can fly. Especially when configured in a local network with the main unit. I've been experimenting lately with such setups for a pet project of mine, using highly asymmetrical protocols (and bypassing TCP/IP altogether but that's another story) where the supplement has all assets in advance (in my case - procedural assets), and the main unit sends out a single beacon/sync message when ready to accept during a frame, while in the meantime the supplement has finished or is about to finish its work, and at said beacon just sends away all done computations for the frame, and then immediately starts with the next frame. Given a proper heuristics for the work split between main and supplement, the two nodes achieve perfect utilisation. And just to give you a clue how asymmetrical it all is (in packets) - it's kilobytes from the main-unit side vs gigabytes from the supplement side (over a GbE link). In my case the workload has seen perfect linear scaling thanks to the precise workload split between the nodes.

So, two things make for a great distributed system:
  • running a minimal overhead sync protocol over a low-latency network connection
  • optimal workload split between the nodes, so no idle time on either end.
Satisfying those yields perfect (read: linear) gains, aka profit : )

That's an interesting experiment. This is the possible new thing that is the most interesting for me among all the patents and rumors.

Would this possible work decently in a peer-to-peer network?
 
Status
Not open for further replies.
Top Bottom