Nintendo GOing Forward.

[Fourth Storm]

How much are carts in terms of GB/$ now? I can't find anything ...

I can't find anything solid either, except someone who says they heard it was up to a third of the retail price. Sounds insane, but it's Nintendo so might be true. Either way, it's definitely not optimal. I'd imagine with Nintendo putting more effort into their digital store, as well as making it available to purchase from a PC/smart device, they will gradually shift to that method. Still, they'll probably want to have some Game Cards on the shelves for visibility purposes. Maybe they'll just stock cards w/ download codes. Put some decent art on there and make them more collectible...

Nintendo might also make up for their lack of physical games at retail via a notification system on smart phones and email, since Iwata has talked about expanding the concept of a platform and building stronger relationships with consumers.

 

[Newguy]

Assuming AMD are making their next handheld, what's the likelihood it'll be their ARM CPU? I know they're trying ARM server stuff and I think their ARM CPU is K12, supposedly in 2016 on 16nm/14nm FF. That seems like a good fit. Then on the GPU side, 1-2 CUs - do we have any examples of the power consumption of really low CU counts on their own (i.e. excluding APUs)?

 

[Fourth Storm]

It'll probably be a vanilla Cortex A57 core, since K12 is aimed at servers and will probably be a bigger/hotter core. As for TDP, right now Mullins is at ~2w with 2 CUs and that's with x86 CPU cores, so with a die shrink and ARM CPU I think they'll be good to go.

 

[Newguy]

In that case then it'll do nicely.

 

[eastmen]

can't rule out x86 for a handheld either. They just released a 3ds update which wont even be out till next year in the states. So a 2017 launch for the next new handheld is my thoughts.

Nintendo said something to the effect of wanting to merge thier handheld and home consoles together.

jaguar in the consoles is already very energy efficent on 28nm. I believe two cores are sub 1 watt at 28nm.

Mullins is a 4 puma core apu with gcn with a 128 shaders at 2w . Shrink this to 16nm (which will be fine for a 2017 launch at almost 2 years old) throw in more shaders at that point and higher clocks you'd still most likely hit sub 1watt.

For the console use the same type of APU and just scale it up. 8 core for the console many more shaders for the gpu portion and alot more ram.

Then you can run all your hand held software on your home console. You can even remove the need to pack in a tablet and just let a person use the handheld for the games.

 

[function]

I can't find anything solid either, except someone who says they heard it was up to a third of the retail price. Sounds insane, but it's Nintendo so might be true. Either way, it's definitely not optimal. I'd imagine with Nintendo putting more effort into their digital store, as well as making it available to purchase from a PC/smart device, they will gradually shift to that method. Still, they'll probably want to have some Game Cards on the shelves for visibility purposes. Maybe they'll just stock cards w/ download codes. Put some decent art on there and make them more collectible...

A third of retail would be a pretty large chunk! Do Nintendo insist on producing carts for third parties? Could be a good way to may extra profits (above licensing) and drive volumes which would ultimately benefit them...

Nintendo seem to be the only major software developer that still routinely make big games that are worth collecting physically. That Drive Club or H:MCC disk isn't going to be worth a great deal without the patches that make the games function properly.

Disks now seem to be a simple token of ownership that gives you a different set of licensing options for the software you've paid for - the disk is basically there to allow access to the software without a downloaded digital certificate. I suppose a simple cart of a couple of Megabytes could serve the same purpose cheaply. I.E. plug it in and you're allowed to download and run the game for as long as the key in the ignition ...

Perhaps that's going to be the future of game "ownership". You own a physical copy of the 'key' that you buy at retail, and that allows you to run the downloaded product offline, on any device. Ideal for collectors (run the game when the servers go down, so long as you have a backup of the game) and you get to maintain retail presence.

Nintendo might also make up for their lack of physical games at retail via a notification system on smart phones and email, since Iwata has talked about expanding the concept of a platform and building stronger relationships with consumers.

This could be a good idea, and would seem to fit well with a move into mobile.

Could a Nintendo phone work? Key partnerships with big phone vendors to integrate into flagship phones?

 

[Fourth Storm]

can't rule out x86 for a handheld either. They just released a 3ds update which wont even be out till next year in the states. So a 2017 launch for the next new handheld is my thoughts.

Nintendo said something to the effect of wanting to merge thier handheld and home consoles together.

jaguar in the consoles is already very energy efficent on 28nm. I believe two cores are sub 1 watt at 28nm.

Jaguar/Puma are low power, true, but ARM cores are likely still more efficient. I'd say Nintendo have more experience with the ISA, but apparantly ARMv8 is completely different than the 32-bit variety. In my non-expert opinion, it is probably easier for them to port their middleware and OS over to another RISC architecture than x86.

For the home console, I'm thinking:

8 ARM Cortex A57 cores @ ~ 1.6 Ghz
10-14 GCN2 CUs @ ~800 Mhz
128MB-2GB HBM @ 128-256 GB/s on package w/ the main SoC
8 GB of DDR4 RAM @ ~50 GB/s (8 chips on a 128-bit bus)

On 14nm/16nm, they could pull this off and still keep the console about the size of Wii U. Even better if they ditch the optical drive for an HDD.

Could a Nintendo phone work? Key partnerships with big phone vendors to integrate into flagship phones?

I agree with you on the topic of ownership and Game Cards acting as a key. Afaik, yes, publishers must buy them from Nintendo, so that is a plus and minus depending on what perspective you take.

Nintendo must have thought of a phone at some point. Perhaps we will see such a thing since Iwata has suggested even more hardware form factors under the unified architecture. I am not sure that it would fare well vs Android and iOS. Likely, this would be a fallback if their next generation fails. Go Android, but lock their games to their own phones/tablets with specialized controls and whatnot.

 

[east of eastside]

I expect a small, low power console, 1.5x the performance of 8th gen, all-digital (no disc). It makes little business sense for Nintendo to make a much more powerful console.

 

[Esrever]

Nintendo didn't use 28nm mainly because they were working with Renesas as a manufacturer and they didn't have a 28nm process or 28nm eDRAM. While cost per transistor is not going down with each shrink anymore, iirc Gamecube and Wii were on the most current nodes available at the time when they launched. If AMD make a quick switch over to finFETs in 2016, it's not out of the realm of possibility that Nintendo use a 14nm/16nm process --especially since the energy savings of 20nm over 28nm are not all that drastic and it appears to be a troubled node. With their handheld, if they are using AMD (as I believe they are), they will want to shrink down the SoC as much as possible in order to improve battery life.

Another area of discussion is storage medium. I've been thinking lately, and with the unified architecture and shrinking shelf space, Nintendo might want to just use gamecards in both handheld and console. Providing both an optical drive and HDD is likely not an option, so given the choice, I'd take the latter. This way, 3rd parties can choose to release a home version digital-only or make a game card that then installs the game and/or downloads additional assets once you insert it. It's true that discs are cheaper to produce, but ditching the optical drive would also make the system quieter and smaller. I expect most games for Nintendo OS will be designed around the handheld with only a few blockbusters a year exclusive to the home console, so including an optical drive just for those seems wasteful.

After nintendo only put 160 shaders in the Wii-u when it would have been easily doable to put 400+, I personally don't trust them to do anything that is not at least outdated by 5 years. They didn't need the Renesas edram, they choose it because it was energy efficient and I assume cheaper than alternatives. The cpu was also fabbed at 45nm when IBM was already doing 22nm work and had 32nm in mass production for a while. I doubt HBM will be cheap in low quantities, the bus to integrate it will likely be costly, it is not all that energy efficient at the low end either as far as I know, it is mainly able to cut down power that high performance gddr5 uses.

I doubt storage would matter for integration unless you want the ability to play the cart on the console, which I doubt is what nintendo is looking to do. Optical drives are mandatory unless they want to go full digital, which nintendo is unlikely to do given their target market (families with children). They'd want their system to be as plug and play as possible, not something that takes a hour or 2 of download before the kids are able to play mario. Size and noise doesn't matter that much to kids anyways, it just needs to be within reason. They will definitely need to release with a HDD if they want to get in the digital age for some online games and DLC compatibility, maybe they will release 2 versions one with HDD and one without like the 360 did.

 

[Fourth Storm]

After nintendo only put 160 shaders in the Wii-u when it would have been easily doable to put 400+, I personally don't trust them to do anything that is not at least outdated by 5 years. They didn't need the Renesas edram, they choose it because it was energy efficient and I assume cheaper than alternatives. The cpu was also fabbed at 45nm when IBM was already doing 22nm work and had 32nm in mass production for a while. I doubt HBM will be cheap in low quantities, the bus to integrate it will likely be costly, it is not all that energy efficient at the low end either as far as I know, it is mainly able to cut down power that high performance gddr5 uses.

I doubt storage would matter for integration unless you want the ability to play the cart on the console, which I doubt is what nintendo is looking to do. Optical drives are mandatory unless they want to go full digital, which nintendo is unlikely to do given their target market (families with children). They'd want their system to be as plug and play as possible, not something that takes a hour or 2 of download before the kids are able to play mario. Size and noise doesn't matter that much to kids anyways, it just needs to be within reason. They will definitely need to release with a HDD if they want to get in the digital age for some online games and DLC compatibility, maybe they will release 2 versions one with HDD and one without like the 360 did.

To address the last part of your post first, I do think size and noise matter to Nintendo if not kids/parents. This is especially since it will likely end up being a second (or third) console for most people. The way I am picturing the relationship between the handheld and home consoles is that the home console exclusives will be basically like new 3DS exclusives: a rarity. Seeing as that they will be entering the generation with next to no third party support outside of indies, printing two skus for each of their first party titles seems incredibly wasteful. Kids are quite used to waiting for software to download these days with iOS/Android, and I am sure there will be an option to automatically and remotely download purchased software in a low-power state similar to current gen. Most of Nintendo's software has a low footprint anyway, so they could possibly fit many games on a Game Card with compression/install.

Nintendo are conservative with technology, but the Renesas eDRAM was quite necessary in order to power Wii BC. They actually spent more money integrating that on die in order to improve performance rather than just keeping it separate like the 1t-SRAM on Wii. As for IBM, they only introduced the 32nm POWER7+ in August of 2012. Granted, Nintendo needed the chip complete much earlier as they then had to work on integrating it on the MCM with Renesas. This time around, the SoC will be handled by one company (AMD), so the process node will likely be determined by whatever is the base design for this semicustom processor. If it's an ARM Skybridge variant, that's 20nm, but we don't know much about AMD's 2016 roadmap except that they have plans to quickly move to finFET and are working with both Samsung and GF, whose 14nm tech is now already in production.

 

[eastmen]

Jaguar/Puma are low power, true, but ARM cores are likely still more efficient. I'd say Nintendo have more experience with the ISA, but apparantly ARMv8 is completely different than the 32-bit variety. In my non-expert opinion, it is probably easier for them to port their middleware and OS over to another RISC architecture than x86.

For the home console, I'm thinking:

8 ARM Cortex A57 cores @ ~ 1.6 Ghz
10-14 GCN2 CUs @ ~800 Mhz
128MB-2GB HBM @ 128-256 GB/s on package w/ the main SoC
8 GB of DDR4 RAM @ ~50 GB/s (8 chips on a 128-bit bus)

On 14nm/16nm, they could pull this off and still keep the console about the size of Wii U. Even better if they ditch the optical drive for an HDD.



I agree with you on the topic of ownership and Game Cards acting as a key. Afaik, yes, publishers must buy them from Nintendo, so that is a plus and minus depending on what perspective you take.

Nintendo must have thought of a phone at some point. Perhaps we will see such a thing since Iwata has suggested even more hardware form factors under the unified architecture. I am not sure that it would fare well vs Android and iOS. Likely, this would be a fallback if their next generation fails. Go Android, but lock their games to their own phones/tablets with specialized controls and whatnot.

ARM on a home console ? They might as well continue using the same cpu they are now.

Nintendo has horrible third party support they should be looking at x86 to fix that not going arm and hindering it. If they launch in 2016 they will have 3 years of engines being optimised for amd x86 and gcn gpus . Why would they want to buck that and go to arm ? Makes little sense .

I also don't see nintendo being massively behind again. The nes , super nes , n64 , gamecube were all power houses for the time.

The wii was not of course but it has the motion controls as a gimmick the wii u tried to do the same and has failed horribly. I think nintendo sees that and will go with a more powerful system

Remember also , if nintendo reboots quickly like in 2016 they can be in a spot where tech changed enough that they can easily out perform the 8th gen systems while staying close in price. Going AMD would allow them to do that while also allowing third parties to easily adjust to take advantage of the more powerful hardware

 

[function]

There are some pretty fast arm chips on the way that will offer much better performance/watt than Jaguar and that will be massively beyond Nintendo's current CPU.

And AMD will be able to integrate them into an APU.

 

[Deleted member 86764]

I also don't see nintendo being massively behind again. The nes , super nes , n64 , gamecube were all power houses for the time.

Their handheld machines were also a generation or two behind the home consoles; GBC - NES, GBA - SNES, DS - N64. It only really all changed around when the Wii released. It's entirely possible that their next handheld could be aiming for Wii/WiiU specs, since the same games can be sold all over again.

What they need to capture is all of the different consumers, which they haven't yet hit on. The Wii did very, very well with non-traditional gamers, but they'd lost the support of a lot of core gamers in the process. I see the main sets of consumers are something like these:

Children
Casuals
Core

Children are pretty well covered at the moment and they lost the casuals with the WiiU since Sports were seemingly dropped. Core were lost around the time of the Wii.

If they want to please everyone, I'd suggested more powerful hardware than PS4/Xbox One, release one SKU with a Sports equivalent and one with a core-type game. Maybe also another with a new Mario. Keep the wands and don't try and innovate with further external hardware.

 

[Shifty Geezer]

Nintendo has horrible third party support they should be looking at x86 to fix that not going arm and hindering it.

Given prevalence of ARM development on mobile/handheld and support for it in middleware, I don't think think devs will have trouble using ARM. In other discussions about going x86, the focus was ease of use and performance, excluding PPC and MIPS (esoteric) and ARM (lack of power). Given that Jaguar is fairly feeble, aiming for that level of performance in a future ARM might be possible, with the advantage of hardware compatibility in handhelds. x86 would be the most convenient to support PS4/XB1 ports, though the differences in libraries might make that still a significant burden, but I don't think ARM would be particularly problematic, and certainly less so than the current niche Espresso architecture.

 

[east of eastside]

ARM fits with Nintendo's handheld/console integration strategy and ARMv8 has ISA similarities with PPC. There is no precedent of Nintendo doing X86, why would it do that now? Plus if Nintendo ever goes mobile, then ARM would facilitate an integrated console/handheld/mobile strategy. Undoubtedly, Nintendo is going with ARM/AMD APU.

 

[Shifty Geezer]

There was no precedent for Sony going x86 either, so that's not really an argument.

 

[eastmen]

Given prevalence of ARM development on mobile/handheld and support for it in middleware, I don't think think devs will have trouble using ARM. In other discussions about going x86, the focus was ease of use and performance, excluding PPC and MIPS (esoteric) and ARM (lack of power). Given that Jaguar is fairly feeble, aiming for that level of performance in a future ARM might be possible, with the advantage of hardware compatibility in handhelds. x86 would be the most convenient to support PS4/XB1 ports, though the differences in libraries might make that still a significant burden, but I don't think ARM would be particularly problematic, and certainly less so than the current niche Espresso architecture.

I would think for nintendo the quickest way would be x86 arm might have a higher level of performance but then again so would the x86 chips vs whats out today. Plus you get the convinece of porting which will be whats needed to convince 3rd party developers to port to your platform.

I don't think the handhelds really factor into compaitbility.

 

[Deleted member 11852]

but I don't think ARM would be particularly problematic, and certainly less so than the current niche Espresso architecture.

Yes, of all things in a console the CPU should be the least misunderstood, but Nintendo still have to put together a cohesive SDK. It's astonishing to think that Sony's PS4 SDK is better well received that Microsoft's Xbox One SDK - and both are built around Microsoft's own VisualStudio. Tools are a big reason why and I gather this has never been a Nintendo strong point.

 

[function]

I would think for nintendo the quickest way would be x86 arm might have a higher level of performance but then again so would the x86 chips vs whats out today. Plus you get the convinece of porting which will be whats needed to convince 3rd party developers to port to your platform.

I don't think the handhelds really factor into compaitbility.

If porting from x86 to arm is going to be an issue for 3rd parties, then porting from x86 (home) to arm (handheld) would be an even bigger issue for Nintendo. And Nintendo rely more heavily on their own software than on 3rd part software.

In a unified strategy between home and handheld, Nintendo's only real option appears to be ARM. They can get AMD to make a scalable ARM APU platform: a low power handheld unit and a larger, faster home based system (more CUs, higher CPU clocks). Develop one game and deploy to both devices, depending on where the user is playing at the time.

 

[Deleted member 13524]

AMD has had SoCs for fanless tablets in the market for a while, using a less-than-stellar process from GF.
I think AMD would be able to build a custom SoC with 4-8 Puma cores and 4-8 GCN 2.0 CUs using 16/20nm+HBM for a Nintendo handheld in 2016.
Looking at 2016, this should probably be comparable to whatever nVidia, Qualcomm and apple will have launching that year for their higher-end ARM SoCs, at least in GPU performance.
It would also run circles around the Wii U's pitiful hardware.

But what AMD could do != what Nintendo will ask for.