[NintendoLife]AMD in new Nintendo console?

[Headfoot]

My money is on ARM to unify the 3DS successor and the home console. They already let you play Smash using your DS as a controller on Wii U. They've tried something along these lines for almost every console going back to the SNES. Hell, the Wii U is literally a big game controller tablet based system.

ARM a57 derivative (4 core?) with tons of fixed-function accelerators built by Nintendo in the console box, and ARM a53 or a7 derivative (4 core?) with fixed function accelerators for the Handheld. I anticipate them making some sort of system where they can run the same code and so you can add lots of Handhelds to your console without it increasing the rendering load on console. So you could have 8 players playing a game (and they've already moved this way for Wii U Smash). Cross platform online play too. The handheld will just render at much lower resolution.

Nintendo loves fixed function logic. Because they do so many 1st party games they get surprisingly good graphics from otherwise weak hardware because they make use of all these weird dedicated blocks. And for the same reason 3rd party developers dont like developing on it.

 

[Snafuh]

ARM a57 derivative (4 core?) with tons of fixed-function accelerators built by Nintendo in the console box, and ARM a53 or a7 derivative (4 core?) with fixed function accelerators for the Handheld. I anticipate them making some sort of system where they can run the same code and so you can add lots of Handhelds to your console without it increasing the rendering load on console. So you could have 8 players playing a game (and they've already moved this way for Wii U Smash). Cross platform online play too. The handheld will just render at much lower resolution.

I think Nintendo will use ARM, too. This way they have easy backwards compatibility on their mobile platform. Instead of going after the big multiplatform titles Nintendo might try to increase their own games output by porting games between platforms. A Mario Kart can work on a mobile platform with minor UI changes and lower resolution.
I think Nintedo will launch a mobile console first and the stationary NX platform 2017.

 

[AnandThenMan]

HBM is extremely expensive, extremely scarce and extremely complex technology to include in a console. No sane person in these companies would go for HBM that is just absurd!

Also 8-12GB HBM doesn't even exist, only 4GB HBM exists today and will continue existing until 2016 where 8GB HBM is to be introduced.

Also your console would be a power hog with those specs, it would never pass testing since it will be too hot with conventional cooling and in consoles you don't want expensive cooling solutions that add up costs.

I have a feeling this will be another "no APU in the PS4/Xbone" type of post. And yes please give sources to your "extremely" everything surrounding HBM.

 

[poofyhairguy]

I really hope the NX is more powerful than the PS4. I dying to see some of those classic IPs (Mario, Zelda, Metroid, Smash, Mario Kart, etc.) with next gen graphics. I've always found the graphics for Nintendo's game to be gorgeous despite its limited hardware. A powerful console will bring those IP to life so to speak.

Actually that has been part of the problem with the Wii U- Nintendo has had trouble maximizing THAT amount of power in first party titles. The learning curve was too great for them at first, hence the slow release schedule for the big games.

I think people need to look in the mirror and realize Nintendo will NEVER complete in the "mainstream" gaming market ever again. The kind of budgets and risks it takes to do a GTA V type game is COMPLETELY out of the question for them, they simply can't keep up with what has become a big budget blockbuster industry like Hollywood. When faced with all the things they would need to compete with a Sony or MS head on (big budget games, an online system, a console powerful enough it doesn't make money at first, etc.) it is obvious they can't and they won't. It would be much more natural for Nintendo to slide into a world that competes for the casuals on iOS than the hardcores on PS4s.

It doesn't matter if they "know how". Its a matter of them not wanting to spend the time and money to convert x86 games to ARM. Nintendo would have to be stupid to go with ARM, they may as well stick with PPC if that were the case so they can at least keep backwards compatibility.

If they want to be a success, they need good 3rd party support, which means they need the same hardware that the other consoles have.

If they go ARM then its easier to port mobile games to the platform. You say they need "third party support" but what you are implying is "they need the third party companies I care about." I am sure you are talking about EA and not a company like Roxio, even though it is the Roxios of the world that are taking over gaming revenue.

The real threat to Nintendo during the Wii U era was never the Xbone or PS4, it was the rise of the iPad and people giving ten year olds their own iPhone. The competition Nintendo needs to overcome is already on ARM, so maybe that is where they need to be.

 

[Sabrewings]

^ Maybe you mean Rovio, and not Roxio?

 

[poofyhairguy]

^ Maybe you mean Rovio, and not Roxio?

Yes, you are right. I mean Rovio, the Angry Birds company.

 

[Stuka87]

If they go ARM then its easier to port mobile games to the platform. You say they need "third party support" but what you are implying is "they need the third party companies I care about." I am sure you are talking about EA and not a company like Roxio, even though it is the Roxios of the world that are taking over gaming revenue.

The real threat to Nintendo during the Wii U era was never the Xbone or PS4, it was the rise of the iPad and people giving ten year olds their own iPhone. The competition Nintendo needs to overcome is already on ARM, so maybe that is where they need to be.

I feel it would be unwise for them to cater to making it easy for mobile games to be moved over. The majority of mobile games are F2P, and that model does not work for a console experience in most cases unless the console maker gets part of that revenue.

Nintendo is better off at making higher quality games to set itself apart from mobile games. Plus most mobile games are designed to be played minutes at a time, not hours at a time.

 

[NTMBK]

If they go ARM then its easier to port mobile games to the platform. You say they need "third party support" but what you are implying is "they need the third party companies I care about." I am sure you are talking about EA and not a company like Roxio, even though it is the Roxios of the world that are taking over gaming revenue.

Stuff the third parties, it will make ports for first parties easier. Unify the fundamental architectures and development tools for the two Nintendo platforms, and make their limited development resources go further.

 

[Blitzvogel]

From an engineering standpoint HBM (well really HBM2) makes sense from the perspective that it creates a huge reduction in circuit board traces if it's the only memory type the NX goes with. This is especially more if the APU is an off the shelf type AMD is using in the PC world. As long as the whole APU + HBM2 doesn't get too hot, having all the heat localized to one spot will require a less substantial cooling unit as far as total volume of that unit goes.

Pretty much the smartest thing Nintendo could do when going to AMD was asking what AMD had in the pipeline and if it would be a fit for the NX, without having to create new silicon.

 

[sm625]

They need to wait until AMD can produce a reasonably priced APU with 16GB of HBM, or maybe 12GB and some amount of flash.

 

[Stuka87]

They need to wait until AMD can produce a reasonably priced APU with 16GB of HBM, or maybe 12GB and some amount of flash.

Why on earth would they need that much memory?

 

[DustinBrowder]

I'll take a gimmick that sells 100 million units with a profit per unit any day!

Compared to a powerful but expensive console that sells, what, 30 million?

WOW. It was a gimmick and its time passed. What other gimmick can Nintendo develop? They tried with the Wii U controller which is basically a handheld and it failed.

That is like once in 20 years gimmick that worked, they are not going to repeat it anytime soon.

Second both the PS3 and Xboc360 sold over 100 million units, with the current console market estimated at at least 300 million users conservatively.

 

[RussianSensation]

dp

 

[RussianSensation]

I have a feeling this will be another "no APU in the PS4/Xbone" type of post. And yes please give sources to your "extremely" everything surrounding HBM.

HBM isn't logical for the next Nintendo console considering what kind of GPU do you need to pair with a 500GB/sec+ memory bandwidth? By today's standards, a flagship one with 600mm2 die size.

XB1 only has 68GB/sec of native GDDR5 bandwidth, 768 shaders. PS4 has 176GB/sec native GDDR5 and 1152 shaders but we aren't seeing games on PS4 that look 50% better than on XB1. That's because the bottleneck isn't just the GPU performance but also CPU performance. For Nintendo to make a significant leap over PS4 then, it'll need to have a CPU 50%+ faster and GPU 50%+ faster, but the problem with that if 50-60 million XB1/PS4 consoles are weaker than NX, developers won't spend much money and effort taking advantage of NX's superior hardware. In essence, it would be akin to Nintendo making a powerful console only for themselves which they would not do.

Even if NX had hardware 3X more powerful than PS4, it would end up like Dreamcast which is way more powerful than old gen consoles but not powerful enough for a true next gen console. Nintendo just needs something from 2016-2021 and there isn't much point in making a console more powerful than PS4 since PS5 is probably not going to launch until 2019-2020.

If they can get backwards compatibility in play, that would be a huge plus. All those gamers who never owned Wii or Wii U would be able to play all the gems on the new console. The 2 biggest obstacles Nintendo has are 3rd party support and online gaming.

Also, right now we are just speculating that NX will launch in 2016 but it could be announced in 2016 and launched Q1 2017 which means 14nm/16nm design is still a possibility.

From an engineering standpoint HBM (well really HBM2) makes sense from the perspective that it creates a huge reduction in circuit board traces if it's the only memory type the NX goes with. This is especially more if the APU is an off the shelf type AMD is using in the PC world. As long as the whole APU + HBM2 doesn't get too hot, having all the heat localized to one spot will require a less substantial cooling unit as far as total volume of that unit goes.

Pretty much the smartest thing Nintendo could do when going to AMD was asking what AMD had in the pipeline and if it would be a fit for the NX, without having to create new silicon.

But you are not looking at the big picture:

1) Modern consoles are not PCB constrained. In other words, there is little incentive to use HBM to cut down on the PCB since the cost of HBM and its lower yields offset any advantages you get in reducing the PCB complexity/size.

2) HBM1 is limited to 4GB of memory which is less than PS4/XB1. Looking at how horribly unoptimized modern games are when it comes to their usage of VRAM, a 4GB HBM1 console going against 8GB GDDR5 would lose. HBM2 is too far out unless Nintendo gets 1st dibs even over AMD's own GPUs.

3) HBM1 and especially HBM2 provide so much bandwidth that it would be almost wasteful to pair them with a low end GPU like an HD7950/GTX960 at this time given how expensive and scarce it is. In the GPU space where say 30-50W extra power headroom and memory controller size reduction allowed AMD to go wider (600mm2), this benefit doesn't exist for Nintendo since they wouldn't use a 600mm2 GPU inside the NX.

4) You aren't addressing the CPU vs. GPU balance. Even if hypothetically Nintendo used a 4000 shader 600mm2 HBM2 GPU, what kind of a CPU would they need to pair it with? It won't work. XB1/PS4 are already CPU bottlenecked and pairing an even faster GPU with slow Jaguar cores is just a waste of GPU horsepower.

Nintendo may want to incorporate Wii/Wii U hardware for backwards compatibility and/or add a custom ARM core so that NX's games can play their handheld games too. All of that will just keep eating into their APU size leaving less room for a very powerful GPU that benefits from HBM.

Also, it's probably likely Nintendo is going APU route since it's the most cost effective solution for consoles today.

Finally, the most important aspect is that even if NX were 3 or 4X more powerful than PS4, who would make next generation games for it when 99% of the market is comprised of XB1 and PS4? No one! It would be a colossal waste of $ to make a powerful NX console.

 

[Blitzvogel]

The XB1 uses DDR3 memory, not GDDR5, and still has to use a rather expensive implementation of small but fast ESRAM to alleviate main memory bandwidth woes. Also, HBM isn't just one stack, it's currently four in the AMD Fury products, each providing 1024 bit data paths, each ~128 GB/s bandwidth, and 1 GB of VRAM. There is no rule that 4 stacks have to be used (AFAIK)

Expect HBM2 to at least double the memory density per chip at the same bandwidth next year, while offering the ability to stack more chips per stack.

I'm expecting Zen APUs to be interposer capable for 1 or 2 small stacks of HBM VRAM with the addition of a large DDR3/4 pool. It's completely reasonable that Nintendo might go with this approach to create a more PC-like memory environment, with ample VRAM and system memory, maybe 2 x 2GB stacks of HBM with 4 GB of 64 bit DDR3. The DDR3 could even be put on the interposer (might as well be fully HBM then). It would avoid the headaches of limited sizes of EDRAM/ESRAM, while offering large memory amounts and the same kind of bandwidths.

What I think it really comes down to is what AMD has in the works for Zen APUs. If they are to be HBM capable, it's all the more likely Nintendo will make use of that capability to have a supply chain and processor that is cheaper because of scale of production, since it's for PCs as well.

 

[BigDaveX]

Second both the PS3 and Xboc360 sold over 100 million units, with the current console market estimated at at least 300 million users conservatively.

What?! Combined sales of the Xbox 360 and PS3 were only about 175 million. And that's not counting the people who owned both consoles and/or brought another 360 because of the RRoD problem.

 

[Madpacket]

Also, it's probably likely Nintendo is going APU route since it's the most cost effective solution for consoles today.

This is what they'll most likely do.

I'm guessing a newer revision but still low power/speed APU with at least 16-20 compute units and 4-6GB of DDR4. This should put them on par with XBOX One and PS4 speeds. They'll make up for the lack of bandwidth with colour compression and maybe a few other tricks. They need to go X86 to gain back 3rd party support, they're dying with just first party exclusives even though they're some of the best games being released. Developers will want to be able to very easily port third party games to it so if Nintendo can get the dev environment / tools nailed they may have a chance.

 

[Headfoot]

2) HBM1 is limited to 4GB of memory which is less than PS4/XB1. Looking at how horribly unoptimized modern games are when it comes to their usage of VRAM, a 4GB HBM1 console going against 8GB GDDR5 would lose. HBM2 is too far out unless Nintendo gets 1st dibs even over AMD's own GPUs.

I haven't seen any technical evidence that this is true. Its limited to 4 GiB per stack, but I have seen no literature which says there is a limit on the amount of stacks beyond the ability to make a sufficiently large interposer.

Fiji could only do 4 because Fiji is 600mm2, but that might not hold true for other products.

A single stack of HBM1 could be a very nice cache and could then take the transistor budget out of the eSRAM/eDRAM kind of caching. Probably not worth the interposer cost though.

 

[Tuna-Fish]

The only way I can see HBM happening is one-stack HBM2 solution. That can have 8GB of ram with reasonable enough bandwidth. Many stacks get expensive...

But again, while I can see why AMD would like to try and upsell HBM to Nintendo, I really don't think they are likely to go for it.

 

[Ancalagon44]

Yeah after reading RussianSensation's posts, I can't see HBM happening either. 4GB of HBM1 memory is not enough, and 8GB of HBM2 would be enough but would probably be either unavailable in volume or prohibitively expensive. Besides which, it is more bandwidth than they would need.

If HBM1 were to be used - and I doubt it will - I imagine the only possible use case for it is 2GB of high speed GPU memory. They could use that instead of having ESRAM. 2GB is large enough for the sub 1080p resolutions they would be targeting, and they could then have 4-6GB of DDR3 for system memory. However, such arrangements (separate GPU and CPU memory) are rare these days in consoles, because of the loss of flexibility.

My guess would be a small ESRAM cache and DDR3. Cost is VERY important to Nintendo, and that would be a good way to get costs down.

Other prediction - the console will maintain backwards compatibility by including the entire PowerPC chip from the previous generation. The PowerPC chip is actually pretty tiny - fabbing it won't be expensive. They don't need to fab the GPU because they are still using an AMD GPU.

 

[NTMBK]

Other prediction - the console will maintain backwards compatibility by including the entire PowerPC chip from the previous generation. The PowerPC chip is actually pretty tiny - fabbing it won't be expensive. They don't need to fab the GPU because they are still using an AMD GPU.

It may still be an AMD GPU, but they would be jumping from VLIW5 to GCN. It's going to be very different.

 

[Ancalagon44]

It may still be an AMD GPU, but they would be jumping from VLIW5 to GCN. It's going to be very different.

They could solve that for most cases via drivers.

I mean, if you think about it, the effect of having a faster GPU is that scenes will render in less time. Most programs shouldn't be affected.

 

[NTMBK]

They could solve that for most cases via drivers.

I mean, if you think about it, the effect of having a faster GPU is that scenes will render in less time. Most programs shouldn't be affected.

Yeah, I guess it depends on how low-level Nintendo got with their GPU programming. If they're hand crafting GPU assembly to squeeze out maximum performance, that's not going to run on GCN, but traditional compiled shader code should be alright in most cases. Don't know enough about Nintendo development practices *shrug*

 

[Snafuh]

Other prediction - the console will maintain backwards compatibility by including the entire PowerPC chip from the previous generation. The PowerPC chip is actually pretty tiny - fabbing it won't be expensive. They don't need to fab the GPU because they are still using an AMD GPU.

Maybe Nintendo can geht backwards compatibility via software. Microsoft did it with although the Xbox360 CPU was more powerful than the WiiU CPU.

 

[therealnickdanger]

Maybe Nintendo can geht backwards compatibility via software. Microsoft did it with although the Xbox360 CPU was more powerful than the WiiU CPU.

Nintendo already has software compatibility for NES, SNES, GB, GBC, N64, Sega Master System, Genesis, TurboGrafx, Neo-Geo, MSX, C64 and Wii games in the Virtual Console. Considering all of those platforms can be easily emulated today with the WiiU, as well as with any entry-level CPU (some Atoms and nearly all Celeron/Pentiums) today, I don't think that their next console will have trouble whether it's ARM, PPC, or X86. Virtual Console is a nice money-maker for them. They wouldn't abandon it.