Nintendo announce: Nintendo NX

Status
Not open for further replies.
14nm SoC
PS4/X1 level performance
AMD IP
Global Foundries mfg
ARM 8 core CPU
30~40 watts power consumption

Iconic console shell, retro looking, not the sleek wedge look of recent consoles.

Smart Device, NextDS, NX software/app fusion synergy.

Traditional controller primary, NextDS as second screen controller

Cartridge based, NextDS cartridge slot

Amiibo support primary to console/controller design.

Is there a even an ARM CPU that exists with sufficient grunt to feed a 1.2+TFLOPS GPU? Even with 8 cores?
 
But I wonder what's the point on killing the 3DS, it is selling well.

It was about 6 years between the DS and 3DS, even if we assume they're releasing one next year it's about 5 years which is a reasonable amount of time. November next year would be about 5 and a half years. But as you say it's selling well, something like an average of 10m units a year. Maybe they're confident in their next product.
 
And so, the speculation has begun. :)
I will point out though that Iwata used singular form, not plural, when announcing that Nintendo is working on another generation of dedicated hardware. Also, in their graphic, the 3DS, WiiU and the NX had their own graphic, (blue dot). (I also find it interesting that they used a shade of blue for the NX that was in between the WiiU and the 3DS, but I can't say if it signifies diddly squat.)

For the WiiU, they used separate CPU and GPU dies, mounted on an MCM, and with a liberal amount of EDRAM. and that was a console that Iwata admitted was designed to a budget, a budget that had to allow for the tablet controller. If they make a stationary device, I wouldn't automatically assume that it would be relatively less costly or sophisticated in it's design.

When it comes to design ethos, I'm more inclined to believe they will keep adhering to making unobtrusive hardware, with moderate power and cooling demands. An AMD SoC with HBM at 50W or so seems reasonable in that case, which by the way it is absolutely certain that AMD is considering variations of anyway for the x86 Windows market, providing mutual R&D savings. A gaming targeted APU in that rough bracket could perform very decently indeed, even if it obviously won't set the performance fetishists on fire.

Again, I'm more inclined to believe that they will make a battery powered device that could connect to TVs with low latency, since they already have that part of the design problem solved. In which case, in 2017, I would ballpark a dedicated gaming SoC designed by AMD to perform somewhere between the Tegra K1 and X1, and probably closer to the X1 thanks to a better performing process node, better memory technology at low cost, and another couple of years of evolution. Incidentally, AMD has their own vested interest in similar ARM/HSA designs.

I can't see AMD being utterly incompetent, when designing for their own areas of interest.
 
Last edited:
Low power SoC is confirmed:

Around 18 months ago, during an informal chat with an extremely well-placed individual in the hardware manufacturing business, an interesting nugget of information dropped into the conversation - Nintendo was already accepting pitches from third parties on the hardware make-up of its successor for Wii U. Two names were mentioned: AMD and Imagination Technologies, creators of the PowerVR mobile graphics tech.

http://www.eurogamer.net/articles/d...-next-gen-hardware-and-the-strategy-behind-it

Nintendo SoC engineer listing:

Summary of Requirements
The ideal candidate will have had experience working directly in a GPU architecture and design team with significant responsibilities.
Low power and SoC design experience would be a plus.
The candidate is expected to have good architectural insights and the ability to apply that for setting future graphics direction for Nintendo.
A bachelors degree (graduate degree preferred) in computer science/engineering or electrical engineering.
5+ years of lead or architectural role experience are required.

Is there a even an ARM CPU that exists with sufficient grunt to feed a 1.2+TFLOPS GPU? Even with 8 cores?

Some AMD A57 core (which has superior perf/watt to Jaguar stock) tweaked with better SIMD support. There is already a new ARM core, A72, but it might be too early for a release next year.

With Nintendo making ARM coded games and apps for mobile and NextDS, there is no reasonable question that NX will be ARM as well, for the efficiency of a common unified code base.
 
Imagination is a good thing, at least they have their hybrid RT accelerator in there, and that would be an improved version in a couple years time. Even if it can't produce effective RT by then, it can definitely add to the stylized nature of Nintendo games.
 
Is there a even an ARM CPU that exists with sufficient grunt to feed a 1.2+TFLOPS GPU? Even with 8 cores?

Here's a better answer... there will be in 2016 - AMD K12, custom, high performance, ARM 64-bit for embedded and ultra low power:


Slides:
http://wccftech.com/wp-content/uploads/2014/05/AMD-2014-2016-Roadmap.png

http://cdn2.wccftech.com/wp-content/uploads/2014/05/AMD-K12-ARM-Core-SOC.png

http://wccftech.com/wp-content/uploads/2014/05/AMD-K12-ARM-Core-20nm.png

Article:
http://wccftech.com/amd-announces-20142016-roadmap-20nm-project-skybridge-k12-64bit-arm-cores-2016/

Bet that NX console and NXDS will be variants of AMD K12.
 
From the Eurogamer article:

Handheld and traditional console are now overseen by a single, integrated department, run by Nintendo veteran, Genyo Takeda.

From Genyo Takeda's wikipedia entry:
. He is known for disagreement with the contemporary model of adding ever technical and graphical improvements to create new console generations. He claims that such a model is subject to diminishing returns.
(...)
He claims to have figured out that consumers will not be satisfied with graphics at any point, that the new improvement effects will eventually wear off, and that “there is no end to the desire of those who just want more.
(...)
one of his major technical goals on the Wii was (being conscious of rising electricity bills and cost-cutting) to scale back the power necessary to operate the console while maintaining the same high performance.

So the Wii U was a flop in no small part due to having terrible hardware, and they put the guy who loves using terrible hardware in charge of developing their next console.
Not a good sign.
And this last quoted sentence reminded me of something: "low-power SoC" doesn't mean "mobile". It just means the power envelope is low. Again.


I'm with @BRiT on this: expect a home console with hardware weaker than the XBone. They'll just launch whatever AMD can accomplish under 30W using 28nm in late 2016.
Which as much of a technical prowess AMD can make out of it, it'll always be something under 30W using 28nm..

I'd say "hey, at least they won't use a decade(s)-old CPU-architecture again", but then I remembered how the New 3DS launched a couple of months ago still uses the ARM11MPCore from 2005.
So what did AMD have back in 2006? Wikipedia tells me it's the 65nm Athlon 64 X2.
 
I'm with @BRiT on this: expect a home console with hardware weaker than the XBone. They'll just launch whatever AMD can accomplish under 30W using 28nm in late 2016.
Which as much of a technical prowess AMD can make out of it, it'll always be something under 30W using 28nm..
.

The thing is there's pretty much every AMD APU right now that can be had for $60 is still much better than Wii U hardware, and it begged the question even if you're going to aim low on the hardware why not toss one of those in? Wii U went above and beyond to be awful, mostly because of backward compatibility imo.
 
Wii U went above and beyond to be awful, mostly because of backward compatibility imo.

Back in 2012 and under the same TDP, I don't know how much better the Wii U could have been. ARM had Cortex A9 and AMD had x86 Bobcat cores and that would've helped, but no GPU at the time could do wonders at 20W using 45nm.
Of course, choosing a GPU architecture that was reportedly 4 generations old by then didn't help, but they'd never be able to do more than a ~300GFLOPS GPU at that TDP.


Likewise, if they predictably use 28nm (or worse) for their 2016 console together with a ~30W total TDP limit for CPU+GPU+RAM+optical drive+storage+peripherals, I doubt they'll achieve a GPU doing more than 800 GFLOPS.
 
It will be 14/16nm ARM CPU based SoC.

It's not rational that it is anything else.

Console SoC's have run a year behind the process used in iPhone since 2012 with Wii U.

This year iPhone is using 14/16nm fab for its iPhone 7 A8 processor.

Console SoC's are fabbed on the same HPM (high performance mobile) process as mobile SoC's.

By late next year the 14/16nm process will be mature enough for a console SoC.
 
It will be 14/16nm ARM CPU based SoC.

It's not rational that it is anything else.

Console SoC's have run a year behind the process used in iPhone since 2012 with Wii U.

This year iPhone is using 14/16nm fab for its iPhone 7 A8 processor.

Console SoC's are fabbed on the same HPM (high performance mobile) process as mobile SoC's.

By late next year the 14/16nm process will be mature enough for a console SoC.
No way Nintendo is using unproven tech for hardware. AMD does not even have 20nm CPUs. And 14/16nm won't be cheap at first. Also availibility will be limited. 28nm was used since Radeon 7000 Q1 2012 and Nintendo did not use it Q4 that year.

I think at best it would be a Carrizo-L derived SoC with 48-64MB eDram. Mybe even less perfomant than xbone.
At worst it would be buffed up Wii U with new gimmick.

For home console.
NX could be a portable.

For portable 4x Cortex A7 at best and low end powervr or whatever (DMP, mali)
 
No way Nintendo is using unproven tech for hardware. AMD does not even have 20nm CPUs. And 14/16nm won't be cheap at first. Also availibility will be limited. 28nm was used since Radeon 7000 Q1 2012 and Nintendo did not use it Q4 that year.

I expect there will be 14/16nm X1/PS4 Slim by the very latest Gamescom '17, and perhaps earlier than that, like E3. So.. I don't see the tech being too immature for a Nintendo console that will launch either in Fall '16 or, especially, Spring. '17.

20nm is being 'skipped' by AMD and TSMC because of cost vs benefits over 28nm, making all the more sense they have done the same in designing NX. If this were not the case you would be seeing 20nm X1/PS4 SoC's launch this Fall alongside, and just one generation behind, 14nm/16nm Apple iPhone 7.

People lose perspective and forget that Wii U was nothing more than a transitional solution, so it's no surprise that Nintendo went with the bare minimum of tech investment.

Samsung will be shipping 14nm Exynos in Q2 '15 and Apple 14/16nm IPhone 7 A9 in the Fall. Probably 100 million of these mobile SoC's will ship at 14/16nm before NX even launches.

Business wise there is too big of an efficiency gain at 14/16nm for Nintendo to ignore (right at the cusp of it being commercially viable) for a 20nm console below X1/PS4 performance, like some expect, that would be barely distinguishable graphically from Wii U in a child's eye.
 
Whatever their hardware is I really hope they get rid of hardware-based backward compatibility completely.

I can't even fathom what kind of Frankenstein hardware it would be if it was a hypothetical Wii U 2 fully backward compatible with Wii U. And it would be a weak hardware as backward compatibility would eat specs like it did with Wii U.
 
If they needed to include the CPU, it would be a very tiny fraction of the die space. It's 33mm at 45nm. On a Finfet process it would probably be sub 8mm. Hopefully, they can go with an HBM memory and that allows the emulation of the 32MB eDRAM. If the bandwidth and latency are as good or better, it should work?

Or they could just make you re-buy everything.
 
Two names were mentioned: AMD and Imagination Technologies, creators of the PowerVR mobile graphics tech.
I will eat my hat if my wild guess (in the old 9th gen console thread) about IMG powering the next Nintendo console turns out to be true: Series 7 GPU (wider than in current SOCs) and two 6 core MIPS CPU clusters :)

I am rooting for this option. It has been too long time since IMG battled against the big boys with "desktop class" hardware.
 
It was about 6 years between the DS and 3DS, even if we assume they're releasing one next year it's about 5 years which is a reasonable amount of time. November next year would be about 5 and a half years. But as you say it's selling well, something like an average of 10m units a year. Maybe they're confident in their next product.
They have the luxury of both time and the lack of competition, so they may well be. Also Nintendo handhelds have been always great for the most part. My favourite is the Game Boy Advance.
From the Eurogamer article:



From Genyo Takeda's wikipedia entry:


So the Wii U was a flop in no small part due to having terrible hardware, and they put the guy who loves using terrible hardware in charge of developing their next console.
Not a good sign.
And this last quoted sentence reminded me of something: "low-power SoC" doesn't mean "mobile". It just means the power envelope is low. Again.


I'm with @BRiT on this: expect a home console with hardware weaker than the XBone. They'll just launch whatever AMD can accomplish under 30W using 28nm in late 2016.
Which as much of a technical prowess AMD can make out of it, it'll always be something under 30W using 28nm..

I'd say "hey, at least they won't use a decade(s)-old CPU-architecture again", but then I remembered how the New 3DS launched a couple of months ago still uses the ARM11MPCore from 2005.
So what did AMD have back in 2006? Wikipedia tells me it's the 65nm Athlon 64 X2.
I think the WiiU didn't fail 'cos of the lack of poer, but I wonder what changed in the company that created the SNES. Some extra grunt would allow their console to remain relevant for a longer period of time in a generation.

Cliff Bleszinski has attacked Nintendo:

cliffb_tweet.jpg

He ironizes about the smartphone market.
 
A very interesting Time interview with Nintendo's CEO Satoru Iwata:

http://time.com/3748920/nintendo-mobile-games/?xid=tcoshare

What prompted this move now, in the light of all you’ve said about Nintendo’s reluctance to craft games for smartphones?

In the digital world, content has the tendency to lose value, and especially on smart devices, we recognize that it is challenging to maintain the value of our content. It is because of this recognition that we have maintained our careful stance. However, we have been seriously and continuously considering how we should make use of smart devices. We made the announcement yesterday because we finally found solutions to the problems we identified. More specifically, we will not merely port games developed for our dedicated game systems to smart devices just as they are—we will develop brand new software which perfectly matches the play style and control mechanisms of smart devices.

We have come to the stage where we can say that we will be able to develop and operate software which, in the end, will not hurt the value of Nintendo IP but, rather, will become an opportunity for the great number of people around the world who own smart devices—but do not have interest in dedicated video game hardware—to be interested in Nintendo IP and eventually to become fans of our dedicated game systems. Yesterday, we finally came to the stage where we were able to announce the alliance with DeNA, which plays a key role in these solutions.
 
Status
Not open for further replies.
Back
Top