Support NeoGAF

[Thraktor]

This is true. They could also go the unconventional route of using a multi-chip module again rather than an APU, which would let them decouple the CPU from the GPU and use whatever combination they want.

While this is true, in this case it would almost certainly end up more expensive than just using a single 14nm die. They were sort of forced into it on Wii U due to the IBM CPU and AMD GPU, but if they're using AMD tech for both, then it makes a lot more sense to just go with a single die, even on an immature node like 14nm.

 

[thefro]

If I remember right it was just dropped on everyone out of nowhere. Don't believe there was any conference.

Yeah, GAF melted down hardcore over that one.

Keep in mind Iwata talked about the Wii at E3 2004, GDC 2005, E3 2005, and the conference before TGS 2005 showing the controller before the E3 2006 final reveal. Then there was still conferences in September in Japan, US, and Europe.

We've gotten way less info from Nintendo on NX at this point.

 

[Doctre81]

Well looky here.

http://www.hardwareluxx.com/index.p...ings-and-10-nm-design-for-games-consoles.html

 

[James Scott]

I wonder, would it be best to delay Mario? I think Zelda has holiday 2016 covered and they'll need something big for 2017.
There's also the new Sonic which Nintendo likely wouldn't delay anything for, but if good could cover that niche while they keep working on the game

 

[ultrazilla]

Well looky here.

http://www.hardwareluxx.com/index.p...ings-and-10-nm-design-for-games-consoles.html

Article says that chip doesn't go into mass production til 2017

 

[Dark Cloud]

So this Investors Meeting they're definitely going to talk about Miitomo numbers and My Nintendo along with Nintendo Accounts. Maybe they'll announce or give hints to their next mobile game? I just hope they give a lot of time to discuss NX instead of saying general comments because this is going to be a long 10 days.

 

[Doctre81]

Article says that chip doesn't go into mass production til 2017

And I'm team 2017 since the get go lol. Could be ps4k or something too.

 

[Malus]

Well looky here.

http://www.hardwareluxx.com/index.p...ings-and-10-nm-design-for-games-consoles.html

Xbox?

I highly doubt that's Nintendo.

 

[Josh5890]

Yeah, GAF melted down hardcore over that one.

Keep in mind Iwata talked about the Wii at E3 2004, GDC 2005, E3 2005, and the conference before TGS 2005 showing the controller before the E3 2006 final reveal. Then there was still conferences in September in Japan, US, and Europe.

We've gotten way less info from Nintendo on NX at this point.

Yea. However at this point in 2013 we had about the same amount of info about the Xbox One

I wonder, would it be best to delay Mario? I think Zelda has holiday 2016 covered and they'll need something big for 2017.
There's also the new Sonic which Nintendo likely wouldn't delay anything for, but if good could cover that niche while they keep working on the game

Galaxy and Sunshine came out a year after their respective consoles launched so that wouldn't be unheard of.

 

[Schnozberry]

While this is true, in this case it would almost certainly end up more expensive than just using a single 14nm die. They were sort of forced into it on Wii U due to the IBM CPU and AMD GPU, but if they're using AMD tech for both, then it makes a lot more sense to just go with a single die, even on an immature node like 14nm.

Yeah, it seems way less likely than 14nm. I just remember their enthusiasm about it during the Iwata Asks for the Wii U hardware.

 

[BY2K]

Well looky here.

http://www.hardwareluxx.com/index.p...ings-and-10-nm-design-for-games-consoles.html

That can't be for the NX, mass production only starts in 2017.

 

[Doctre81]

Wait hold up. Doesn't TSMC manufacture memory for samsung? I had a theory that the NX may use samsung 10nm ddr4. Maybe this is what that could be?

 

[Thraktor]

Article says that chip doesn't go into mass production til 2017

They say the first 10nm chip will go into production in 2017, they didn't say the games console was that chip. The console chip could be 2018 or even later.

 

[Pyrokai]

That can't be for the NX, mass production only starts in 2017.

It's for the iterative console future and the NXK releasing in 3 years.

 

[Schnozberry]

I wonder, would it be best to delay Mario? I think Zelda has holiday 2016 covered and they'll need something big for 2017.
There's also the new Sonic which Nintendo likely wouldn't delay anything for, but if good could cover that niche while they keep working on the game

Mario would be a great title to debut on both the handheld and the console simultaneously. So maybe whenever the handheld goes out the door would be a good time for a major cross buy release.

 

[Malus]

That can't be for the NX, mass production only starts in 2017.

Maybe it's a handheld thing.

edit: I don't know shit.

Games industry y u so mysterious? I just wanna Mario: https://www.youtube.com/watch?v=65uNCLBTje0

 

[10k]

Just wanted to respond to a few posts on the topic of the Primitive Discard Accelerator, and how it relates to Polaris and 14nm.



Just to clarify here, the quote in 10k's post describing the Primitive Discard Accelerator isn't from his source, it's from me (and I'm definitely not any kind of insider). His source had told him that the GPU featured new features like the PDA (which AMD has publicly listed as one of the new features in Polaris), and asked me what the PDA does, and whether it could be used in a 28nm GPU.

I should also note that although AMD have listed PDA as one of the new features in their Polaris presentation, they haven't given any details on what it does or how it does it, so my description of its use is purely my own assumption based on the name of the feature, although I can't really see how the words "primitive discard accelerator" could mean anything very different from what I described.

The "developer" also gave a description of what the PDA does and why it's useful, which ties in pretty much exactly with my understanding of how it would work. Which doesn't necessarily mean that the rumour is true, but it does mean that the person making the claim at least has a decent understanding of real-time 3D graphics.

In any case, the reason for discussing the PDA isn't that it's necessarily a revolutionary feature (although it could certainly be useful), but that if the source is real, then the GPU has at least some Polaris features, and is therefore very likely to be manufactured on 14nm.



If true, then it would be reasonable to assume that it has at least some other Polaris features, but it doesn't necessarily guarantee that it has all of them. For example, let's say the improved Command Processor, Geometry Processor and PDA for Polaris had all been all-but finished in late 2014 when work on NX started, but the improved CUs weren't finished until much later. It possible that Nintendo would have been able to make use of several of the new features of Polaris, but would still use GCN 1.2-era CUs. Both PS4 and XBO's APUs sort of sit between generations like this, so it wouldn't be all that out of the ordinary. On the other hand, there are reports that Polaris has been pretty much finished for a long time now, and was delayed first by the abandonment of 20nm and secondly by the slow growth in 14nm yields. In that case Nintendo may have had almost the entire suite of Polaris features to use when building the NX chip.

The other thing I'd like to mention here is that wccftech's description of the Primitive Discard Accelerator here is actually wrong, at least as far as my understanding goes. They claim it will be used to implement conservative rasterisation (a new DX12 technique which currently only works on Nvidia cards), although there's nothing about conservative rasterisation which I would describe as "discarding primitives". In fact, one of the main use cases of conservative rasterisation (ray-traced shadows) uses it for precisely the opposite effect, in preventing sub-pixel triangles from being thrown away.

The reason I bring this up is that wccftech is a tech website that's heavily focussed on graphics technology. If they can get something like this wrong, then it's all the more impressive that someone claiming to have insider information would get it right. That doesn't mean it's any more than a rumour, but there's a relatively high threshold of knowledge on the part of anyone trying to fake this.



As mentioned above, I wouldn't consider a Primitive Discard Accelerator a major feature (although it could be a useful one), just that the source claimed that NX has it, which would indicate that the NX APU is on 14nm and using a Polaris-based GPU.

That said, although you're right in that GPUs do implement small-triangle culling, orientation culling, etc. all in hardware, they only do it at the rasterisation stage (and afaik this is true for both AMD and Nvidia). What this means is that the triangle has to run through the geometry front-end, the vertex shaders and the rasteriser before it actually gets thrown out. Even if we're just talking about orientation culling, regardless of what kind of game it is or how well optimised your engine is, that means that you're wasting about half of the throughput of your geometry front-end on unnecessary triangles and you're doubling your vertex shading workload without benefit. If you bring that culling right to the very start of the pipeline then you stand to make some fairly nice efficiency gains, and small-triangle, frustum and high-Z culling could also stand to benefit (although these will be more dependent on how well optimised games already are in each regard).

I'd recommend having a read through the slides of a recent GDC talk given by one of Frostbite's senior rendering engineers (PDF link here). It does a good job of running through the benefits of early culling on GCN-based hardware, and the performance gains from Frostbite's compute shader solution. They also implement occlusion culling in their solution, but I wouldn't expect to see acceleration for that at the front of the pipeline, due to it requiring pre-Z. That said, I could certainly see developers being happy about a solution which brings efficient orientation culling and small-triangle culling to the start of the pipeline. If you're working on a game where geometry throughput is your main bottleneck (say an open-world game with sub-optimal LOD) then it could make your life quite a bit easier.

I didn't know if you wanted the attention or not so I called you a tech gaf source

 

[MuchoMalo]

Well looky here.

http://www.hardwareluxx.com/index.p...ings-and-10-nm-design-for-games-consoles.html

You are not going to see 10nm TSMC chips on the market before 2019, and that's being optimistic.

EDIT: Okay, maybe they pushed it up a bit, but either way it wouldn't happen in a console earlier than 2018 if the first chip enters mass production in 2017.

 

[Schnozberry]

That 10nm stuff would make sense for a portable.

 

[double0cinco]

While this is true, in this case it would almost certainly end up more expensive than just using a single 14nm die. They were sort of forced into it on Wii U due to the IBM CPU and AMD GPU, but if they're using AMD tech for both, then it makes a lot more sense to just go with a single die, even on an immature node like 14nm.

So would it make any sense to shrink puma+, or would it be more practical/efficient to go with an existing 14nm design, i.e. Zen or A72. Do FinFETs complicate a potential die shrink of a 28nm chip like puma?

 

[James Scott]

Mario would be a great title to debut on both the handheld and the console simultaneously. So maybe whenever the handheld goes out the door would be a good time for a major cross buy release.

Yeah, that sounds about right.
In the original speculation thread I think most agree that the handheld would come out first and the console afterwards after a few months.
I wonder how it would play out.
If it releases in March they can have a steady stream of titles for the handheld and console after that.
Here's how I can see it going down:
November: NX launch with Zelda and some 1st party game we haven't guessed yet.
December: Smash complete edition.
January/Feburary: one of the Wii U ports each month
March: NX handheld launch with Mario which is playable on the console and DQXI in japan
Add in Luigi's Mansion 3 and Pikmin in there, I guess.
If they can get some solid 3rd party ports, this could end up pretty well, I'd think

 

[Schnozberry]

So would it make any sense to shrink puma+, or would it be more practical/efficient to go with an existing 14nm design, i.e. Zen or A72. Do FinFETs complicate a potential die shrink of a 28nm chip like puma?

If you're going 14nm and want to ease your development burden between supporting home console and portable devices, what is your motivation for using x86 at that point? x86 isn't that difficult to port over to 64-bit ARM now that most modern engines support both.

 

[double0cinco]

If you're going 14nm and want to ease your development burden between supporting home console and portable devices, what is your motivation for using x86 at that point? x86 isn't that difficult to port over to 64-bit ARM now that most modern engines support both.

I would definitely agree, but I'm thinking about that x86 rumor from a bit ago. To me, 14nm Polaris would seem to rule out x86, as the only likely candidate would be Zen, which would not mesh with LCGeek's info of only marginally better than the xbone Jaguar.

I'm thinking 14nm Polaris would strongly imply 14nm A72 cores.

 

[Doctre81]

You are not going to see 10nm TSMC chips on the market before 2019, and that's being optimistic.

EDIT: Okay, maybe they pushed it up a bit, but either way it wouldn't happen in a console earlier than 2018 if the first chip enters mass production in 2017.

Wait but everyone believes the nx just now hit mass production and is supposed to come out this year.

Wii entered mass production in september of 2006 btw.

 

[Malus]

Wait but everyone believes the nx just now hit mass production and is supposed to come out this year.

This doesn't sound right but I don't know enough about console production vs chip production to dispute it.

 

[maxcriden]

I wonder, would it be best to delay Mario? I think Zelda has holiday 2016 covered and they'll need something big for 2017.
There's also the new Sonic which Nintendo likely wouldn't delay anything for, but if good could cover that niche while they keep working on the game

I'm not convinced Sonic covers a Mario niche for fans like he did in the 90s. I'm definitely mixed on whether we'll see Mario at launch, though. I could certainly see Zelda being the big game as you said, provided there are some significant FP exclusives as well, like a showcase game and Pikmin 4. So, similar to the Wii launch really.

 

[Terrell]

I don't think 20,000 extra sales on a Nintendo console is make or break. At the moment the Wii U and Nintendo are basically sunk as far as developer's attitude towards them. Sure there are 1st party and some choice third party Japanese titles but the rest are non existent. Some titles sold in the 10's of thousands. That's appalling.

Why would a developer risk putting a "core" title on a Nintendo console anymore? Seriously? It costs them about a mill to do so, and they might sell 10k copies. Sure, things can change and Nintendo can pay to have those games etc etc, but really I think Nintendo are basically the bottom of the heap when it comes to places to put games. Maybe just above ouya.

I'm glad that you think that anything with Nintendo on the box automatically puts a glass ceiling on any aspirations for sales that even the most garbage indie title could achieve.

And people have been saying they're "bottom of the heap" since the Gamecube era ended. And yet there the Wii U was with 3rd-party games at launch. Clearly that went wrong even before the hardware launched, but yet they greenlit the games being made at all, so it really flies against this notion that 3rd-parties somehow don't want to be on a Nintendo console, that they will never entertain the risk, despite every entertainment industry being a risk-based industry.

What titles?

Tomb Raider, Resident Evil 6, Sleeping Dogs, Hitman Absolution... these are the 4 that come immediately to mind. I'm sure there are more, but those "failures" are the most extensively well documented and most publicly discussed.

They could cut costs there, although it sounds to me like we could be looking at a controller with a BoM that isn't cheap. Also, a cartridge system would almost fly in the face of all this talk about Nintendo courting 3rd parties. Cartridges are part of the reason 3rd parties gave them the middle finger two decades ago.

And a lot has changed since the times when we were using batteries to hold save files in volatile RAM chips. I think there should be a moratorium on discussing Nintendo's past without an actual acknowledgement of how the situation has changed since that particular event occurred, instead of acting like it happened yesterday or that it's some grudge that will never die.

There are also more cost-cutting methods, like using the same touch screens that are likely to be in the handheld and get a higher volume price break on production of that component and no 2nd production line for a different size. Among other options.

 

[Doctre81]

Couldn't the 10nm they are talking about just be the fab process of the RAM?

 

[10k]

We're almost there.

This counting down to the date of the earnings release/news conference.

I thought it would read 10k for a second lol.

 

[BY2K]

We're almost there.

This counting down to the date of the earnings release/news conference.

I'm gonna laugh/cry so hard if we don't get ANYTHING about the NX.

 

[wtsfox]

Could the 10nm be for the new handheld?

 

[Doctre81]

Could the 10nm be for the new handheld?

Possibly.

 

[Azalean]

I'm gonna laugh/cry so hard if we don't get ANYTHING about the NX.

At this point I'm not expecting anything before E3. If we get an announcement before then, that's wonderful.

 

[James Scott]

I'm gonna laugh/cry so hard if we don't get ANYTHING about the NX.

I'm hoNestly just eXpecting that we'll get some news on one of the 4 upcoming mobile games. Don't recall ever experiencing an investor's meeting that was worthy getting hyped about

 

[Malus]

I'm gonna laugh/cry so hard if we don't get ANYTHING about the NX.

Well, it'd be an informative "nothing." No NX until 2017

D:<

 

[Doctre81]

April 27th.

" We will walk about the NX at E3. See you there!"

 

[10k]

I would definitely agree, but I'm thinking about that x86 rumor from a bit ago. To me, 14nm Polaris would seem to rule out x86, as the only likely candidate would be Zen, which would not mesh with LCGeek's info of only marginally better than the xbone Jaguar.

I'm thinking 14nm Polaris would strongly imply 14nm A72 cores.

Yeah, the more 14nm seems likely, the more 14nm A72 seems to be the CPU that's going to be in the NX (which is 40% better in performance compared to the jaguar in PS4 and Xbox). The handheld will likely use the highly efficient A53.

14nm 8-Core A72 2.2Ghz
8GB GDDR5
Custom Polaris 14nm 2.0 Tflops

When reading up Polaris 10 and 11 info, I learned that it only comes in HBM and gddr5. HBM being expensive is likely for the high end Polaris 10 while the gddr5 is for the Polaris 11, which is aimed for notebooks and mid range desktops. Nintendo could theoretically still ask for DDR4 or DDR3 and add some eDRAM to counteract the latency, but that seems unnecessary.

 

[Doctre81]

Yeah, the more 14nm seems likely, the more 14nm A72 seems to be the CPU that's going to be in the NX (which is 40% better in performance compared to the jaguar in PS4 and Xbox). The handheld will likely use the highly efficient A53.

14nm 8-Core A72 2.2Ghz
8GB GDDR5
Custom Polaris 14nm 2.0 Tflops

When reading up Polaris 10 and 11 info, I learned that it only comes in HBM and gddr5. HBM being expensive is likely for the high end Polaris 10 while the gddr5 is for the Polaris 11, which is aimed for notebooks and mid range desktops. Nintendo could theoretically still ask for DDR4 or DDR3 and add some eDRAM to counteract the latency, but that seems unnecessary.

Nintendo is not agianst spending a large sum on memory if they feel it will drastically increase system performance. That much edram for the wiiu was not cheap and they have used expensive and/or new memory in their system in the past.

 

[ffdgh]

April 27th.

" We will walk about the NX at E3. See you there!"

More or less what I'm expecting. The official name could be cool though.

 

[Discomurf]

Ok so the Wii name was also announced on april 27th 2006.
http://www.neogaf.com/forum/showthread.php?t=97127

Was there also a conference meeting on that day? I can't remember. Mariokart wii also launched on april 27th.

Yep - came out of nowhere it seemed... they launched a very simple website with this video:
https://www.youtube.com/watch?v=NDz6SXdlTHQ

Most of GAF was freaking out about the name ( I always thought it was great)

 

[James Scott]

Yep - came out of nowhere it seemed... they launched a very simple website with this video:
https://www.youtube.com/watch?v=NDz6SXdlTHQ

Most of GAF was freaking out about the name ( I always thought it was great)

After 10 years I just realized the "ii"s that bow in the logo are the Wii would like to play guys...or at least I think that's what they're supposed to be

 

[Doctre81]

Yep - came out of nowhere it seemed... they launched a very simple website with this video:
https://www.youtube.com/watch?v=NDz6SXdlTHQ

Most of GAF was freaking out about the name ( I always thought it was great)

LOL watching that video and seeing how they "promoted" WiiU is kinda sickening. I hope they step their advertsing/hype game up for NX.

 

[MK_768]

I wonder, would it be best to delay Mario? I think Zelda has holiday 2016 covered and they'll need something big for 2017.
There's also the new Sonic which Nintendo likely wouldn't delay anything for, but if good could cover that niche while they keep working on the game

I dont see why they won't push Mario to late 2017. Retro's game, Luigi's Mansion 3(if it's happening), Pikmin 4 can handle the launch window.

Plus they have surprises coming.

 

[Tom Nook Sawyer]

After 10 years I just realized the "ii"s that bow in the logo are the Wii would like to play guys...or at least I think that's what they're supposed to be

Pretty sure they just represent people. If anything, the "Wii would like to play guys" are based on the logo, seeing as they came considerably later

 

[Pokemaniac]

And I'm team 2017 since the get go lol. Could be ps4k or something too.

Do you think it's gonna get delayed or something? We have a credible source (trev's leak) that it is currently scheduled for this year.

 

[bachikarn]

I dont see why they won't push Mario to late 2017. Retro's game, Luigi's Mansion 3(if it's happening), Pikmin 4 can handle the launch window.

Plus they have surprises coming.

I also they don't need a lot of help at launch. I think Nintendo system typically sell out at launch. The issue is after launch, and having a big seller like Mario might be better.

 

[Doctre81]

Do you think it's gonna get delayed or something? We have a credible source (trev's leak) that it is currently scheduled for this year.

If I remember correctly even Trevs source said the system is "planned" to launch this year. Not confirmed..

 

[MK_768]

I also they don't need a lot of help at launch. I think Nintendo system typically sell out at launch. The issue is after launch, and having a big seller like Mario might be better.

I think they just need to show the games that will be coming throughout the first year. That is very crucial. They should also have a decent launch of course and make sure there isn't a significant drought.

 

[Anth0ny]

I dont see why they won't push Mario to late 2017. Retro's game, Luigi's Mansion 3(if it's happening), Pikmin 4 can handle the launch window.

Plus they have surprises coming.

smash, luigi's mansion, pikmin...

what is this the fucking gamecube?

















because that would be great

 

[Doctre81]

smash, luigi's mansion, pikmin...

what is this the fucking gamecube?

















because that would be great

Gamecube also made the mistake of not launching with a mario title. I doubt they will make that mistake again.