Nintendo DX GPU?

[Blitzvogel]

I'm gonna laugh so hard is Nintendo somehow justifies once again using a PowerPC 750 core, though I'd be very doubtful they'd be that stupid. It's just a question of whether AMD is making Zen or ARM based SoCs for them, since it's a "hybrid" platform.

2 out of 3 of AMD's announced SoC wins could be attributed to the NX system, one for the home console, and one for the mobile component. I do very much wonder if one or two Zen cores and a few Polaris compute units could really be crammed into a handheld form factor. ARM could be used for the mobile NX system, with games compiled for both architectures. The latter is possible given AMD's past couple years experience with ARM design and engineering.

 

[sm625]

If it releases in 11 months then the hardware has to be pretty much finalized by now. Since AMD has no 14nm products on the market right now it is foolish to assume that this will even be a 14nm SoC, let alone Zen or Polaris based.

 

[R0H1T]

If it releases in 11 months then the hardware has to be pretty much finalized by now. Since AMD has no 14nm products on the market right now it is foolish to assume that this will even be a 14nm SoC, let alone Zen or Polaris based.

In fact I'd say it's the opposite, the fact that NX will release a year from now makes it (almost) certain that it'd be based on some form of 14nm GPU, very likely from AMD, with the (remote) possibility of Zen cores being in there lest they want a repeat of Wii U.

 

[CuriousMike]

Its totally possible they're using "approximation" hardware to do current development.

Before the XBOX dev kits came out, you used a PowerPC Mac with an ATI video card in it.
It wasn't until about 9 months before shipping you got actual 360 dev kits.

 

[tenks]

In fact I'd say it's the opposite, the fact that NX will release a year from now makes it (almost) certain that it'd be based on some form of 14nm GPU, very likely from AMD, with the (remote) possibility of Zen cores being in there lest they want a repeat of Wii U.

this, 100%. AMD already announced it won a contract for nintendo.

Its totally possible they're using "approximation" hardware to do current development.

Before the XBOX dev kits came out, you used a PowerPC Mac with an ATI video card in it.
It wasn't until about 9 months before shipping you got actual 360 dev kits.

and this

 

[FatherMurphy]

Nintendo has never been interested in cutting edge technology (specifically, the costs inherent therein). Innovative technology, but not expensive technology. I think @sm625 makes a good point. There are no 14nm Polaris APUs out right now, and there won't be for a while. There certainly are no Zen cores even near mass production stage (sampling this quarter, possible mass production late Q4). And the Zen cores that are sampling and that are targeted for introduction later this year (if nothing goes wrong) are high-performance server cores, not the kind of cores that go in a console.

So, a 14nm Polaris APUGPU? Perhaps, but it does not fit the pattern of Nintendo relying on tried-and-true, cheaper technology.

A 14nm AMD Zen CPU? Not a chance.

EDIT: I believe AMD announced that they were developing custom ARM cores in step with Zen? Perhaps that could be a possibility, as those are smaller cores and use less power. But the point remains that AMD has yet to ship a 14nm product, and its first 14nm product will be a discrete GPU. It's only 14nm CPU core, Zen, has not shipped, is designed for Servers, and is not integrated with a GPU. Nintendo is an incredibly unlikely candidate to assume the risk and cost of being AMD's first 14nm APU customer.

 

[raghu78]

Nintendo has never been interested in cutting edge technology (specifically, the costs inherent therein).

The beauty of AMD's semi-custom technology is it takes existing PC tech developed for CPUs, GPUs and APUs and assembles it into a semi-custom design tuned to the customer's requirements.

Innovative technology, but not expensive technology. I think @sm625 makes a good point. There are no 14nm Polaris APUs out right now, and there won't be for a while. There certainly are no Zen cores even near mass production stage (sampling this quarter, possible mass production late Q4). And the Zen cores that are sampling and that are targeted for introduction later this year (if nothing goes wrong) are high-performance server cores, not the kind of cores that go in a console.

Zen is sampling this quarter and will be available in high performance desktops (8c,16t) in Q4 2016 and in servers (32c,64t) in Q1 2017. This means the CPU core will be available for use within a quarter or so assuming they need another spin to iron out bugs. It means semi-custom APU and PC consumer APUs can start using Zen for launch in late Q1/Q2 2017.

So, a 14nm Polaris APUGPU? Perhaps, but it does not fit the pattern of Nintendo relying on tried-and-true, cheaper technology.

A 14nm AMD Zen CPU? Not a chance.

EDIT: I believe AMD announced that they were developing custom ARM cores in step with Zen? Perhaps that could be a possibility, as those are smaller cores and use less power. But the point remains that AMD has yet to ship a 14nm product, and its first 14nm product will be a discrete GPU. It's only 14nm CPU core, Zen, has not shipped, is designed for Servers, and is not integrated with a GPU. Nintendo is an incredibly unlikely candidate to assume the risk and cost of being AMD's first 14nm APU customer.

There is no risk and cost with being AMD's first 14nm APU customer. AMD Zen APUs for consumer PCs are scheduled for a Q2 2017 launch. btw Zen R&D has been paid for by AMD over a period of 4+ years. So there is no risk/cost for Nintendo.

 

[renderstate]

this, 100%. AMD already announced it won a contract for nintendo.

First time I hear this. Source?

 

[Mopetar]

So, a 14nm Polaris APUGPU? Perhaps, but it does not fit the pattern of Nintendo relying on tried-and-true, cheaper technology.

Which is a strategy that failed them horribly with the WiiU. They've got a dedicated fan base, but it's not enough to sustain another failed console.

A 14nm AMD Zen CPU? Not a chance.

EDIT: I believe AMD announced that they were developing custom ARM cores in step with Zen?

ARM is very unlikely, one of the big reasons being that Nintendo needs third party support, which means that they're likely to aim for a system that makes ports easy. There was one rumor supposedly from a developer supporting that, and although not concrete, it makes a good deal of sense.

Zen could be a possibility, far more likely though is that they'd use the same Jaguar cores that are in the Sony and Microsoft consoles rather than ARM cores.

 

[IllogicalGlory]

All the leaks I'm seeing point to x86. We have a few sources directly saying x86, and then there are the rumors that porting to it from PS4/Xbone will be easy. Can't see it being ARM at this point. My guess is it's using Polaris 11, with a 2GHz+ Jaguar. Somewhat faster than the PS4.

Nintendo wants/needs to make a capable machine this time around. The well seems to have run dry with the Wii U.

 

[Sonikku]

Think they could use Polaris 11 for NX?

 

[Blitzvogel]

Think they could use Polaris 11 for NX?

I don't see why not, it's a sensible size at 14nm, low TDP and power draw, while offering what is likely much more power than either the Xbone or even PS4. I'm expecting the NX to be right between the PS4 and PS4 Neo, so about 3.0 TFLOPS. I have no idea how it's gonna pan out if the NX really is a hybrid system with both base and portable consoles available that run the same software at different "settings". If that's the case, ARM would make more sense for portability, unless 14nm Zen is as miraculous at low clocks as AMD and so much of the public wants it to be.

 

[Magee_MC]

Nintendo's CEO said "he company was “not building the next version of Wii or Wii U” and that the device will be something “unique and different.” What if that something different is a game platform heavily focused on VR. In that case, using AMD Polaris chips with or without Zen would make a lot of sense. It would also dovetail nicely with AMD's focus on VR with Polaris.

 

[Ancalagon44]

My prediction would be that they won't use Zen in the NX console. What they will use instead will be ARM cores of some kind. Either an off the shelf design, or very close to off the shelf. 4-8 ARM cores packaged into an SoC with a Polaris class GPU.

They will use ARM to keep the cost down, and to make sure they can hit the time to market that they want. It would be difficult to launch with Zen next year. The other problem is that the SoC in the handheld unit is much more likely to be ARM than x86. AMD doesn't currently have any x86 CPU designs that could work in a handheld device, and Nintendo isn't going to pay them to design one.

Polaris 10 or Polaris 11? Really not sure at this stage. Polaris 10 isn't really all that big as far as die size goes, at least not compared to something like Fury. All depends on yields I guess.

 

[R0H1T]

My prediction would be that they won't use Zen in the NX console. What they will use instead will be ARM cores of some kind. Either an off the shelf design, or very close to off the shelf. 4-8 ARM cores packaged into an SoC with a Polaris class GPU.

They will use ARM to keep the cost down, and to make sure they can hit the time to market that they want. It would be difficult to launch with Zen next year. The other problem is that the SoC in the handheld unit is much more likely to be ARM than x86. AMD doesn't currently have any x86 CPU designs that could work in a handheld device, and Nintendo isn't going to pay them to design one.

Polaris 10 or Polaris 11? Really not sure at this stage. Polaris 10 isn't really all that big as far as die size goes, at least not compared to something like Fury. All depends on yields I guess.

Well AMD was working on their custom ARM cores, for servers IIRC, so it's not like they can't make the NX or the handheld console, based on their own custom ARM cores. Secondly two cat (Puma+) cores on 14nm can easily fit inside a small tablet sized device, therefore even handhelds could be based on x86. The only reason why Nintendo would prefer ARM in the portable console, even the NX, is to make use of the vast library of Android games out there otherwise x86 is still the way to go for either, or both, of them IMO.

 

[sm625]

I do not see nintendo using a SoC that costs them more than $70. And that is assuming they go bigger than they've ever gone in the past. I think the Wii and Wii U SoC + RAM were both $50? For $70 they could get a SoC very similar to the XBox One and that should include the cost of the memory, which is not insignificant. If they go 14nm, I just do not see how they get 8GB of RAM along with a new top of the line process node SoC for under $100. And there is zero point in trying to build a powerful console if you're not going to have 8GB RAM. I just dont see it. Even if they could, I dont see what business sense it would make to for a foundry to commit such a huge portion of its brand new node's capacity towards such a low margin part. It would be financial suicide. Until I see absolute proof that NX is on 14nm I am 100% sure it will be a 28nm part, or an extremely small 14nm, like 78 sq mm.

 

[Ancalagon44]

Well AMD was working on their custom ARM cores, for servers IIRC, so it's not like they can't make the NX or the handheld console, based on their own custom ARM cores. Secondly two cat (Puma+) cores on 14nm can easily fit inside a small tablet sized device, therefore even handhelds could be based on x86. The only reason why Nintendo would prefer ARM in the portable console, even the NX, is to make use of the vast library of Android games out there otherwise x86 is still the way to go for either, or both, of them IMO.

The problem is that even die shrunk Cat cores would only be useful in a device the size of a tablet (at smallest) or preferably a laptop. Given that the NX handheld must be usable by kids as well as adults, I'm thinking it is on the smaller end of the form factor. Because of this, I'd be surprised if they use 14nm Cat cores in it.

I don't know how far AMD got with their custom ARM cores. Also remember that their custom ARM cores were supposed to target the server market, not the phone/handheld market. They might be too large for a handheld.

 

[railven]

Not sure if it was discussed here, but else where it was believed that the Nintendo NX would sell at a loss, like other consoles:

http://www.eurogamer.net/articles/2016-05-02-the-nx-wont-be-sold-at-a-loss

Seems that's not going to be happening. Wondering if Nintendo is going to aim for $300 again, or more.

 

[RussianSensation]

Not sure if it was discussed here, but else where it was believed that the Nintendo NX would sell at a loss, like other consoles:

http://www.eurogamer.net/articles/2016-05-02-the-nx-wont-be-sold-at-a-loss

Seems that's not going to be happening. Wondering if Nintendo is going to aim for $300 again, or more.

I think Nintendo is in a very tough spot, regardless of the pricing or hardware strategy they have up their sleeve. Simply said, this is yet another case of if you don't show up on time, you lose/forfeit a generation.

By the time the NX launches in Spring 2017, the mainstream gamer will consider it a 4-year-old console. Currently, Xbox One and PS4 have an aggregate userbase of approximately 60 million gamers. It would not be surprising if this reaches 75-80 million by the time the NX even shows up. This is important because every day the NX is not on the market, someone is buying Xbox One/PS4 consoles and that means with every lost sale, there is a shrinking market of remaining customers still waiting to jump onto this generation of consoles.

What exacerbates the situation is if someone hasn't purchased an XB1/PS4 yet, perhaps waiting for their slimmer/updated versions or the Neo, chances are they may still go for XB1/PS4 consoles in 2017 simply because their friends/relatives already own those consoles. Just imagine, even if NX was 3-4X faster than PS4, it would essentially pull the Dreamcast. If the NX barely matches or beats a PS4, even for $299, that's 4 years too late.

Then we have another obstacle. How do you convince the general market and ex-Wii/Wii U owners to get the NX given how Nintendo has failed miserably with the Wii U? If I owned a Wii U, I'd wait on the side-lines to see if Nintendo actually delivers the games unlike just believing their word. This creates a catch-22 for developers who may abandon the NX before it has a chance to take off if 2017 NX's sales aren't meeting expectations.

Since Nintendo isn't even willing to take a loss on their console, the chance that this console will even match PS4 Neo's specs is slim to none. It just seems they are launching WAY too late to be a contender for XB1/PS4 generation but at the same time WAY too early to have the NX carry the torch during the 2nd half of this generation and the 1st half of XB2/PS5 gen.

Basically, the NX is launching in the no man's land time-frame in this console gen. The timing itself could be this console's greatest obstacle, regardless of its hardware specs or price.

 

[Ancalagon44]

I think Nintendo is in a very tough spot, regardless of the pricing or hardware strategy they have up their sleeve. Simply said, this is yet another case of if you don't show up on time, you lose/forfeit a generation.

By the time the NX launches in Spring 2017, the mainstream gamer will consider it a 4-year-old console. Currently, Xbox One and PS4 have an aggregate userbase of approximately 60 million gamers. It would not be surprising if this reaches 75-80 million by the time the NX even shows up. This is important because every day the NX is not on the market, someone is buying Xbox One/PS4 consoles and that means with every lost sale, there is a shrinking market of remaining customers still waiting to jump onto this generation of consoles.

What exacerbates the situation is if someone hasn't purchased an XB1/PS4 yet, perhaps waiting for their slimmer/updated versions or the Neo, chances are they may still go for XB1/PS4 consoles in 2017 simply because their friends/relatives already own those consoles. Just imagine, even if NX was 3-4X faster than PS4, it would essentially pull the Dreamcast. If the NX barely matches or beats a PS4, even for $299, that's 4 years too late.

Then we have another obstacle. How do you convince the general market and ex-Wii/Wii U owners to get the NX given how Nintendo has failed miserably with the Wii U? If I owned a Wii U, I'd wait on the side-lines to see if Nintendo actually delivers the games unlike just believing their word. This creates a catch-22 for developers who may abandon the NX before it has a chance to take off if 2017 NX's sales aren't meeting expectations.

Since Nintendo isn't even willing to take a loss on their console, the chance that this console will even match PS4 Neo's specs is slim to none. It just seems they are launching WAY too late to be a contender for XB1/PS4 generation but at the same time WAY too early to have the NX carry the torch during the 2nd half of this generation and the 1st half of XB2/PS5 gen.

Basically, the NX is launching in the no man's land time-frame in this console gen. The timing itself could be this console's greatest obstacle, regardless of its hardware specs or price.

All valid points and I'm interested to see how Nintendo markets the console.

By far I think the most important thing is going to be the games on offer at launch. At minimum, they need 1-2 strong first party exclusives to even stand a chance. Something like a new Metroid, Mario Kart or Smash Bros. As well as the usual 3rd party games, preferably a few exclusives if they can convince the publishers to do so.

I think the launch games are the factor that will make or break this console. It took ages for decent games to come out on the Wii U, and that is part of the reason it didn't sell well. Nobody will buy the console if there are no games to play on it.

Given that this console supposedly represents a big shift in Nintendo's strategy, I'm really not sure how they will price and market it. However, I'd be surprised if they launch at $299 or above - it's too risky. More likely, they will target a $199 to $249 price point. At such prices, they likely won't have the hardware to match a PS4 Neo, you're right.

Will that matter? If they can get "close enough" with the hardware for cheaper, then the software remains the unique selling point.

 

[nanaki333]

here's to hoping nintendo goes for a console that can do 4k gaming at least 30fps, otherwise, like others say, it's going to be considered a 4 year old console if performance is similar to ps4.

 

[Raduque]

Frankly, it makes more sense for them to release a handheld and a $50 micro-console. Any actual console they release is a guaranteed failure unless Apple buys them or something.

I'd buy both. Be great to play my handheld games on the big screen.

 

[RussianSensation]

here's to hoping nintendo goes for a console that can do 4k gaming at least 30fps, otherwise, like others say, it's going to be considered a 4 year old console if performance is similar to ps4.

There is a 0% chance this is happening. You cannot design a sub-$399 console that can do 4K 30 FPS modern gaming at high settings. More importantly, the market for console gamers who also own 4K TVs is too small. You would be wasting hardware resources to target a tiny fraction of 4K TV owners. It is far better to design a console that can do 1920x1080 60 FPS instead. This way, PS4/XB1 multi-plays look and run best on NX and yet it requires little to no work to port those games from those consoles. For that reason X86 APU + GCN GPU is the way to go.

The issue Nintendo has is this strategy could have worked in 2012-2014 but not in 2017. Even if they manage to increase PS4's hardware 2X, that won't sell the console. Their timing is the worst it can possibly be.

If they want to level PS4/XB1, they need to offer something out of this world, such as you but a Nintendo NX game and you get a free PC version of the same game. Then, offer online game sharing with your friends. I buy Zelda and can allow my friend to play it by streaming it over the cloud using my game key for instance. They can limit it to 1-2 family/friend member accounts.

Essentially, they need to do something entirely unique that Sony and MS do not offer because faster hardware won't sell the NX 4 years out.

I think the traditional console market is just too small now to support 3x 80-100 million consoles such as PS360/Wii era. The casuals left, hardcores are getting into PC gaming. 80 million or so will own PS4/XB1 by Spring 2017, leaving Nintendo with scraps. They will be lucky to sell 30-40 million NX consoles before PS5 drops in 2019-2020. When that happens, the NX is again outdated just 2-3 years out.

Like I said, the timing could not be worse.

 

[exar333]

I think Nintendo is in a very tough spot, regardless of the pricing or hardware strategy they have up their sleeve. Simply said, this is yet another case of if you don't show up on time, you lose/forfeit a generation.

By the time the NX launches in Spring 2017, the mainstream gamer will consider it a 4-year-old console. Currently, Xbox One and PS4 have an aggregate userbase of approximately 60 million gamers. It would not be surprising if this reaches 75-80 million by the time the NX even shows up. This is important because every day the NX is not on the market, someone is buying Xbox One/PS4 consoles and that means with every lost sale, there is a shrinking market of remaining customers still waiting to jump onto this generation of consoles.

What exacerbates the situation is if someone hasn't purchased an XB1/PS4 yet, perhaps waiting for their slimmer/updated versions or the Neo, chances are they may still go for XB1/PS4 consoles in 2017 simply because their friends/relatives already own those consoles. Just imagine, even if NX was 3-4X faster than PS4, it would essentially pull the Dreamcast. If the NX barely matches or beats a PS4, even for $299, that's 4 years too late.

Then we have another obstacle. How do you convince the general market and ex-Wii/Wii U owners to get the NX given how Nintendo has failed miserably with the Wii U? If I owned a Wii U, I'd wait on the side-lines to see if Nintendo actually delivers the games unlike just believing their word. This creates a catch-22 for developers who may abandon the NX before it has a chance to take off if 2017 NX's sales aren't meeting expectations.

Since Nintendo isn't even willing to take a loss on their console, the chance that this console will even match PS4 Neo's specs is slim to none. It just seems they are launching WAY too late to be a contender for XB1/PS4 generation but at the same time WAY too early to have the NX carry the torch during the 2nd half of this generation and the 1st half of XB2/PS5 gen.

Basically, the NX is launching in the no man's land time-frame in this console gen. The timing itself could be this console's greatest obstacle, regardless of its hardware specs or price.

Great point.

I would add that it is not only the Wii and Wii-U users burned. This has been Nintendo behavior everywhere recently. While I really like my 'New' 3DS-XL, just look at how confusing that hw release was. DS to 3DS to 2DS (kind of backward/side product) to 3DS XL - 'new' 3DS XL. That constant churn speaks to a lack of vision and planning. Nintendo has done a great job at confusing their customers in the past 5 years and the Wii-U debacle was just icing on the cake. The Wii-U isn't terrible by any means, but was poorly executed, marketed and has a lot of re-hashes. No wonder 3rd parties left in droves.

 

[JDG1980]

There is a 0% chance this is happening. You cannot design a sub-$399 console that can do 4K 30 FPS modern gaming at high settings.

Consoles don't do high settings, they do medium settings. PC gamers wind up wasting 2x-4x the GPU power for tiny, almost unnoticeable improvements in rendering quality. And Nintendo's first-party titles (which are about the only reason anyone even buys a Nintendo console in the first place) are far more efficient than the newest bloated "AAA" FPS.