Are there any technical reasons why Nintendo couldn't or shouldn't use ARM for Wii2?

[Squilliam]

From a technical perspective we have ARM based chips already being scaled up to the point where they can (tablets) and will (Windows 8 ARM) support a home computing style environment with the chips. Unlike say 7-8 years ago as Nintendo was finalising the Wii, we have powerful ARM chips and supporting graphics and media decoding hardware which is more than suitable for installing in a home console which is targetting <25W overall power consumption.

So given the known fact that ARM chips are small, compact and the supporting components are likewise small and compact whilst by default they include nothing that isn't useful for a home console and exclude nothing which is useful, how would that work out from a technical and software eco-system perspective? It already appears on the surface that the cost of console development and the royalty costs as well as the overall performance levels already point to their using ARM as their chipset as it seems that ARM systems are the closest to supporting their Wii philosophy of small, simple and low power systems with high per unit margins. With so much development in this area they may be able to effectively take a design off the shelf and implement it without too much hassle.

Can they produce a system which can take modern day console games, reproduce them at a slightly higher level of overall performance and offer backwards compatibility in a system which is more cost efficient and power efficient than AMD's Fusion based processors on both 40nm and 28nm or anything IBM has? We've already seen Sony talk up how easy it is to port from the PS3 to ARM though im sure they spent a lot of time developing tools for just that. So it may indeed be possible to emulate ancient PowerPC hardware on the fly given an 11 year time period between releases.

 

[tongue_of_colicab]

Would a ARM cpu really be fast enough? Even NGP isn't (nearly) as powerfull as a ps3/x360 so unless Wii2 will be not much more faster than x360/ps3 would a ARM cpu really be fast enough? And what about the gpu? Do they have to stick with one of the gpu's they now use in combination with ARM cpu's or could they also use a AMD chip? In that case maybe ARM could make sense if they go with a beefier gpu that would also do things like physics so a relative slow cpu won't matter? As for off the shelve ARM systems, AMD can offer that too right? Even if you take a standard bobcat you can (somewhat) play modern games. Even if you take the cheapest stand alone videocard gpu you can from AMD and add it, it will be much faster than any current ARM system.

 

[RudeCurve]

Backwards compatibility would be the main obstacle. Other than that i could see Wii2 with a quad core ARM cpu and 8 core PowerVR chip fitting easily into the same volume as the current Wii.

 

[AlNom]

Backwards compatibility would be the main obstacle.

hm... I'm not so sure. Graphics ought to be a cakewalk for translating the API calls, and the CPU isn't exactly a beast by any standard in the last half decade. How does that emulator on PC fare with hardware requirements? i.e. does setting it to 480p make a difference or is it particularly CPU limited?

You'd hope/think that Nintendo would come up with a better emulator as well.

 

[eastmen]

If they want low power why not just go with lanio or whatever its called from amd . That should provide good cpu/gpu power and low power draw.

 

[Shifty Geezer]

Would a ARM cpu really be fast enough? Even NGP isn't (nearly) as powerfull as a ps3/x360 so unless Wii2 will be not much more faster than x360/ps3 would a ARM cpu really be fast enough?

Squilliam's point is that NGP is having to fit ARM in a single-digit watt handheld, but how would it cope without those power constraints? One could easily just stack 4 NGPs on each other for a low power, moderate performance box if Nintendo want to pull another Wii. That'd be quite a lot of cores though!

I have no idea how high ARM can scale. Performance wise it probably doesn't have to be substantial, as lots of the heavy work of a CPU can be shifted to GPUs. Failing that, Nintendo could add a vector unit of their own choosing, like the extended SIMD units on Xenon.

If they want low power why not just go with lanio or whatever its called from amd . That should provide good cpu/gpu power and low power draw.

I suppose the advantage with ARM is handheld synergy. Same code pool between console and handheld, same games running on 3DS and Wii2. It'd allow Nintendo to offer NGP<>PS3/4 type game sharing (play your PSS game on NGP and then on PS3 and then back onto NGP) helping to build a Nintendo ecosystem that extends beyond just the box with its games.

 

[fehu]

3DS showed that in nintendo's opinion if you have a differentiating gimmik and sufficent graphic you can go for a moderate hardware
if this keep up, the wii2 will not be so worried about being only ~360 level
they can easily go for dual A9, or by the time single A15, any modern slow gpu and (think to) win again

 

[Rodéric]

Not contributing to the thread directly but:

Why call it "Wii 2" when there's a (IMO) better name : NES6 ?
(I believe this is the standard codename Nintendo uses for its home consoles.)

 

[Shifty Geezer]

I concur! I was thinking when posting that Wii2 is a bad name as Nintendo haven't ever named sequentially.

 

[MarkoIt]

I think that at higher power budget, AMD Fusion solutions might have more performance/mm^2. For example Llano successor should be plenty enough for a generation step up from Xbox360/PS3.

Or they maybe can go with 3x times this chip:
http://forum.beyond3d.com/showthread.php?t=59574

 

[corduroygt]

If they want low power why not just go with lanio or whatever its called from amd . That should provide good cpu/gpu power and low power draw.

ARM costs much less since you just license the IP and produce it wherever you want, and draws even less power.

 

[Arwin]

I wouldn't dare speculate on what the next home console is going to be for Nintendo. It all depends on how much value they put on being able to play Wii games on it. If they do want to support that, then I expect that will be a big reason why it won't be ARM. Otherwise, no idea.

 

[corduroygt]

How about a quad core ARM based AppleTV with some sort of camera input system, or using the magic trackpad. Play the same games on iphone, ipod, ipad, or new appleTV.

 

[Shifty Geezer]

I think that at higher power budget, AMD Fusion solutions might have more performance/mm^2. For example Llano successor should be plenty enough for a generation step up from Xbox360/PS3.

Let's just be clear, this isn't the next-gen tech thread, but specifically what's the reason not to choose ARM? The actual choice of processors and licensing deals belongs in the next-gen thread (where Squilliam asked this, but it got lost in the noise so he made a specific thread for a specific question).

 

[Squilliam]

Squilliam's point is that NGP is having to fit ARM in a single-digit watt handheld, but how would it cope without those power constraints? One could easily just stack 4 NGPs on each other for a low power, moderate performance box if Nintendo want to pull another Wii. That'd be quite a lot of cores though!

ARM probably can already give higher performance than the Wii already I believe. It can also probably clock a lot higher than people give it credit for such as from here: Where it says roughly speaking that a dual core A15 which can clock at up to 2.5Ghz with a Power VR 'Rogue' (VR6?) producing 210GF in ALU performance. Now obviously those numbers are peak but in saying that I would seriously suspect that the latest ARM cores are more optimised at the moment for embedded or large mobile devices given the fact they likely had a large lead time from Microsoft telling them that Windows ARM computer systems are on the way for 2012.

I have no idea how high ARM can scale. Performance wise it probably doesn't have to be substantial, as lots of the heavy work of a CPU can be shifted to GPUs. Failing that, Nintendo could add a vector unit of their own choosing, like the extended SIMD units on Xenon.

I suppose the advantage with ARM is handheld synergy. Same code pool between console and handheld, same games running on 3DS and Wii2. It'd allow Nintendo to offer NGP<>PS3/4 type game sharing (play your PSS game on NGP and then on PS3 and then back onto NGP) helping to build a Nintendo ecosystem that extends beyond just the box with its games.

I doubt that the Power VR series of GPUs does any form of GPGPU work although that doesn't exclude an SOC based on Tegra if they wanted to go down that route. Since the ARM cores are highly customiseable I wouldn't put it past them to add a vector unit or two to each core.

Overall theres a handheld synergy, they could almost go as far as to use the NES6 architecture as the basis for their next handheld when the time comes which would simplify their lives from a product development perspective. It would mean they could optimise their engines for just the one type of architecture and it would mean that they could take advantage of the exploding development of applications and games targetting the ARM ecosystem of processors. I have always thought they should implement an application store ALA Apple because the non gaming possibilities of the Wii were never really explored in that respect and they certainly didn't make great use of Wii Connect 24.

I think that at higher power budget, AMD Fusion solutions might have more performance/mm^2. For example Llano successor should be plenty enough for a generation step up from Xbox360/PS3.

Or they maybe can go with 3x times this chip:
http://forum.beyond3d.com/showthread.php?t=59574

Performance per mm^2 on just the CPU is one thing but when you can produce your chip on bulk silicon with lower licencing costs per unit and higher performance per watt you would be foolish to overlook that possibility as licencing costs may indeed outweigh any silicon costing advantages, not to mention overall packaging costs with cooling inclusive compared to the more integrated ARM offerings.

That chip does look promising, though knowing them a customised quad core is probably more likely at this point. They'd need to clock it higher to ensure backwards compatibility with Wii games.

 

[eastmen]

ARM costs much less since you just license the IP and produce it wherever you want, and draws even less power.

Are current arm chips going to be enough. Even if they put the same tech as the NGP it will still lag behind lanio. If they scale up they will quickly hit lanio's power budget and i'd wager it be harder to program for and be less powerful.


There isn't a reason they couldn't go with arm that isn't technical

 

[corduroygt]

Are current arm chips going to be enough. Even if they put the same tech as the NGP it will still lag behind lanio. If they scale up they will quickly hit lanio's power budget and i'd wager it be harder to program for and be less powerful.

How do you know how much an ARM cpu's power will be affected when you increase the performance and how is it going to be more than AMD's? x86 CPU's are only efficient when made by Intel using their superior process, AMD lags a lot in this regard.

Also, Nintendo always makes do with less powerful and cheaper hardware than you'd think.

 

[eastmen]

How do you know how much an ARM cpu's power will be affected when you increase the performance and how is it going to be more than AMD's? x86 CPU's are only efficient when made by Intel using their superior process, AMD lags a lot in this regard.

Also, Nintendo always makes do with less powerful and cheaper hardware than you'd think.

I'd put a current zacate chip against the quad core/gpu set up in any day of the week. The arm set up may win on power drain but the amd chip will walk over it in terms of performance. Llano will do even more.

They are quad core with each core only 9.69mm2 at 32nm and come in at 35M tranistors . Each core has 1MB L2 cache and are expected to run at 3GHZ. Its expected at 25W or lower TDP . For its gpu it will use something out of the evergreen family.

It will be a power house at such low power usage.

 

[corduroygt]

I'd put a current zacate chip against the quad core/gpu set up in any day of the week. The arm set up may win on power drain but the amd chip will walk over it in terms of performance.

Current ARM-Cortex A9 is faster and has better performance per watt than Atom running at the same clock frequency. CPU wise a Bobcat core is only 10%-15% faster than an Atom core running at the same clock frequency. Not to mention we're talking about A15, not A9, which will be even faster.

They are quad core with each core only 9.69mm2 at 32nm and come in at 35M tranistors.

Let's see AMD make one 32nm chip before jumping to conclusions. Also Llano is said to have power dissipation between 25W-59W, you bet the quad core 3 Ghz version is going to be closer to 59W, if AMD can even manufacture it within its timeline and power budget in the first place.

 

[3dcgi]

ARM costs much less since you just license the IP and produce it wherever you want, and draws even less power.

You mean like ATI licensed the graphics IP to Microsoft for the Xbox 360? The business model isn't a reason to choose or not choose ARM.