-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
SemiAccurate: Nintendo NX handheld to use Nvidia Tegra-based Soc
- Thread starter Fourth Storm
- Start date
Schnozberry
Member
If we presume that Nvidia was able to give them a decent deal on SOC's, perhaps it's because they'll have two of them bundled within a single SKU. At a price of $349 like the Wii U launched at, it might be possible to get a console and a handheld profitable inside of that price range with modest SOCs. Having the controller for the NX console double as a streaming unit capable of playing independent games on the go and running it's own apps is actually a pretty attractive set of features and would be unique in the industry at this point in time.
Davey Cakes
Member
NX will clearly be a family of products, I agree with you there. Or a joined ecosystem between different hardware and software, and form factors. PC, mobile, QOL will be there but I definitely see a semi-traditional handheld and console being part of the equation too.
Team Klimt
Member
10th Nintendo console?
NES
Super NES
GameBoy
Virtual Boy
64
Gamecube
DS
Wii
3DS
WiiU
NX
Dang. That's 11. Maybe they are pretending WiiU never happened...
SneakyStephan
Banned
I can answer this one
Yes it's the same as response time
Few things to keep in mind:
the number the manufacturer reports is just a marketing number, they only count grey to grey response time which is always the lowest number.
For other transitions it can be 20 or 30 ms (that means on a 60 hz monitor it can almost two full frames for a pixel to change to the right color and back again, so it's simply the wrong color for a full frame, that's why fast panning scenes on LCD panels just turn into an unintelligible smear)
On top of that the number is only reported with the highest Overdrive setting , which you'll never use because it makes things look like this:
http://www.tftcentral.co.uk/images/asus_rog_swift_pg348q/response_4.jpg
You get these annoying halos on high contrast edges that are very noticable and very distracting and annoying.
I have a "1 ms response time" lcd monitor and it's still a blurry piece of shit in motion (much of it thanks to sample and hold) and has these awful overshoot halos at high contrast edges.
Last thing to keep in mind is that even this marketing bullshit number is still hundreds of times higher than the switching time of an OLED, CRT or plasma panel
Or more accurately: response time isn't really a thing on those types of displays as their pixels work much differently.
And lcd pixel is just a glorified color filter that sits in front of a backlight, and needs to be physically reoriented to filter a different shade or color. This makes it garbage at displaying motion.
So mobile is as powerful as the Wii U as we speak and has a massively massive install base and yet no big western publisher will even port a 360 game to mobile?
They ported quite a few PS2-era games.
And btw they do spend millions developing mobile games nowadays. It's not all Flappy Bird.
I realize the state of the mobile market. That's why I'm looking for just one example. Also it is only part of a larger argument that says we aren't going to see Zelda Wii U on a Nintendo handheld next year.
Prisoner KSC2-303
Member
We don't know when the handheld is coming. Some (WSJ) even suggested that one may come with the console.
The point is not what the flagship title is, but what the most iconic titles represent. Zelda, Mario etc all need to incorporate any major selling point Nintendo put forward for the NX. Having caveats that may preclude use of one of the most popular titles with it is not going to help them market it. The system sellers all need to be 100% in line with the console's feature set.
If they wait for a shared library - assuming a handheld doesn't arrive until late 2017 at the earliest - when do they announce it? It's not a feature for the console until there's a handheld, so they risk once again seeing a home console lose momentum as a USP isn't shown working at launch.
A miserable little pile of secrets.
Bioshock and X-com are on mobile platforms.
Skittzo0413 already replied and I agree.
But, thinking on the NVIDIA connection, who's to say there might not be some cross-branding going on with each others' hardware? Surely this is all about just getting the chips, but it's kind of fun to wonder if they actually are working on something more! Maybe NVIDIA could put out the l33t version of the NX and basically be Nintendo's Alienware?
It certainly is hard with semantics, translations, etc.! Plus, being grilled on-the-spot and trying to choose your words appropriately doesn't help! Who also knows how these things are edited before they get put out there?! Lots of variables, but it makes the discussion lots of fun!
It's funny and crazy now how I'm imaging this thing. Back in the day, a system was a system. I had one "bingo card" with my guesses. Now, I am picturing three different scenarios, each with different types of systems, controller styles, etc. I really can't say I've had this happen before. But, it'll be fun playing three bingo cards this year for the reveal!
I'm actually hoping for a new input method too, or at least a substantial advancement of something they've done previously! You can do some cool things with screens with eye-tracking and IR! You can do neat things with buttons + touch, motion, etc. It's also fun to try and see what technology is being used more commonly in devices, entertainment, etc. now and how they may apply it to gaming! Lots of options there!
LOL!!! Nice and appropriate!
IT's prohibitive because of cost.
They wouldn't design a docking feature just to make use of additional memory from a handheld. No point. Just add more memory to the console.
IN both cases it would make more sense just to make the console more powerful. It's a much simpler idea. It wouldn't be any more expensive than adding the capability for the console to work in parallel with the handheld. And it wouldn't create a 3rd device for Nintendo to support.
It isn't that the "docking" idea isn't doable. It just isn't practical.
Emily is very odd when it comes to her writing style, which is why this is considered debatable in the first place. I'm 99% sure that the point of that paragraph is that it's weaker than Xbone and that she was saying that even calling it close to Xbone is a stretch.
wonderful103
Member
What she meant was that the ouya blows it out of the water
James Scott
Banned
She said it's weaker than ps4/closer to the X1.
"Even that's a stretch" is weirdly worded, though.
I can answer this one
Yes it's the same as response time
Few things to keep in mind:
the number the manufacturer reports is just a marketing number, they only count grey to grey response time which is always the lowest number.
For other transitions it can be 20 or 30 ms (that means on a 60 hz monitor it can almost two full frames for a pixel to change to the right color and back again, so it's simply the wrong color for a full frame, that's why fast panning scenes on LCD panels just turn into an unintelligible smear)
On top of that the number is only reported with the highest Overdrive setting , which you'll never use because it makes things look like this:
http://www.tftcentral.co.uk/images/asus_rog_swift_pg348q/response_4.jpg
You get these annoying halos on high contrast edges that are very noticable and very distracting and annoying.
I have a "1 ms response time" lcd monitor and it's still a blurry piece of shit in motion (much of it thanks to sample and hold) and has these awful overshoot halos at high contrast edges.
Last thing to keep in mind is that even this marketing bullshit number is still hundreds of times higher than the switching time of an OLED, CRT or plasma panel
Or more accurately: response time isn't really a thing on those types of displays as their pixels work much differently.
And lcd pixel is just a glorified color filter that sits in front of a backlight, and needs to be physically reoriented to filter a different shade or color. This makes it garbage at displaying motion.
Never personally noticed any of that but thanks for the explanation.
Nvidia Shield is at $199.
One of those sentences is presented poorly. We don't know which one. I think it's best to just expect the worst, so I'm sticking with noticeably below Xbone. Like, 4 SMs (512 CUDA cores) from Pascal at 900-1050MHz for the console, 2 SMs (256 CUDA cores) at 300-400MHz for the handheld. It would be nice if she could just give a straight answer, but she seems almost afraid to do so.
With a 540p screen, even a $150 2017 handheld could run it.
This discussion will never end, let's assume that it's around Xbone's power ± 5% and leave it at that.
As I said in the previous thread, that's how the information was presented to her. She can't clarify it unless she is given new information to share.
Graphics Horse
Member
tbh we have a thread for this specific quote so people can discuss it there if they want.
Skittzo0413
Member
Again, this is also referring to "raw power" which could also be interpreted as "on paper specs."
It's probably best to assume the worst anyway, but considering it could very well be an Nvidia GPU now we don't know how real world performance will compare.
I personally don't care about specs- as long as it supports UE4, Unity, Frostbite, etc. it can get a good amount of third party games, especially around launch and especially from Japan.
Edit:
Yeah this thread is for Nvidia/handheld discussion, lets leave Emily's rumor the other thread.
I know, but I can't help feeling that her paraphrasing was off or that she was combining two different statements. She probably should have just quoted what they said.
The discussion is somewhat related though, since the choices for the handheld and console go hand-in-hand. We'd need to find a performance level where scaling works well between the two. We also have an issue in that there hasn't been a mobile device with the Tegra X1 up until now. Basically, to find out where the handheld might be, we'd have to look at where the console might be and work backwards.
TSMC says that 16nmFF+ can either be 40% faster or consume 60% less power. Shield TV uses up to 20W. 60% less than that is 8W, which is still way too high. That's whole system though. I'd say that it would need to be cut down to a third of that in the best case, since we have to account for screens. That's 170 GFLOPS, or 2 SMMs running at 332MHz. We can't realistically expect the console to be more than 5-6x as fast as that. Why is all of this important? It all assumes that we're looking at Pascal. Knowing where the console is provides context on that as well as telling us where a Pascal Tegra could end up.
On that point, there's another thing I should mention. Previously, I criticized Emily's information on where the chips are being fabricated because it doesn't seem likely that she would know that. 16nmFF+, however, would explain how she was able to obtain such information. She wouldn't be able to figure that out for herself, of course, but the point is that 20nm and 16nm are the only possibilities what she said is true, and a Tegra X1 cut down to consume under 3W would be a bit of a joke. It would also be dead-end, since shrinking from a planar process to a FinFET one might be cost prohibitive. Thus, we are certainly looking at Wii U levels of performance with the handheld. The Shield Tablet is currently $200, so Nintendo should be able to hit that if they're willing to take a lower profit margin than usual. That's what's worrying though: would Nintendo be willing to take a lower margin?
Schnozberry
Member
Considering their margins on software, accessories and Amiibo, taking a slim margin on hardware at launch is not outside the realm of possibility.
But I don't see why/how this would be more costly? The idea I was running with is that the console would be near the same as the handheld in parts but cheaper because there is no screen and other extras, as the entry point into gaming on NX. When you add in extra memory and things like you're suggesting to make the console more powerful, THAT will drive up the price more than a simple port providing a connection to the handheld.
Now, the above hinges on the idea that Nintendo will want to provide different entry points. They've done this with the 3DS line, but I'm just conjecturing if/how they could do this with NX as a platform instead, both console and handheld. But, why wouldn't they? If you could provide a cheap console capable of running all the engines/middleware of the other consoles, but with Nintendo games and all on it, that's a great sell in and of itself. There's nothing stopping them from a premium super unit later, just that between the handheld and console, you have the iPad and the Mac Mini scenario. Letting them interact is a given and there will definitely be connectivity between them. (Lots of things come to mind, like a plug to use headphones but play on the TV screen, another controller input, providing a microphone, etc.) But, why not provide extra incentives to those who own both? The console could charge the handheld's battery, get an extra boost from the handheld, provide a sturdy place for the camera (only on the handheld) for doing special AR things at home on the big screen or even act like a Kinect, provide ambient lighting under your TV (I don't know?! The gaming version of the Phillips Hue!), provide a physical Navi to shout at you and give you hints while you're playing a game, and lots of other things. There's lots of practical things the connection could do. And none of this again is creating some third device for Nintendo to support!
They definitely could make the console much stronger than the handheld...but why when they can get fairly close to the others in the handheld itself? (It's possible the console could naturally be better just by being able to draw more power since there's no concern over a battery, but I'm definitely not the one to prove that!) There's no incentive for just trying to blow the other guys out of the water with the console, especially when their own games and development haven't reached some amazing technical heights or anything anyways. It also helps a ton with scalability in their library. They could just try to make both fairly close and use the savings from the things removed from the handheld to provide a cheaper alternative for those wanting just the console and (hopefully) make bank on volume.
Thats what I was waiting for
You got it pretty quick too!
It is worth remembering that in order to make money from the mobile market you have to target devices from at least two generations of processor/GPU ago. Because that is by far the bulk of the market.
This issue is irrelevant for dedicated games machines (although comes into play somewhat when you are doing iterative hardware revisions, but won't be as severe as it is in the mobile market).
KojiKnight
Member
Let's be honest... the real reason they don't port console games to mobile is a lot simpler... Economically, they don't work on the platform. People aren't willing to pay even $15-20 for a decent port, and the companies could be making so much more money on something new designed to wring money out of people via microtransactions and mandatory energy wait timers.
Graphics Horse
Member
I don't think there are necessarily two systems so I don't need to do such a thing
I think Nvidia will be able to provide something that can outperform the Wii U in a portable form. That might require a handheld chunkier than we're used to, say a large 2 inch+ deep box, like...um... the Shield Portable!
My only doubt is fast RAM performance, idk if they can get something on a par with the 32mb the Wii U had, but it's outside my brainzone.
I don't think there are necessarily two systems so I don't need to do such a thing
I think Nvidia will be able to provide something that can outperform the Wii U in a portable form. That might require a handheld chunkier than we're used to, say a large 2 inch+ deep box, like...um... the Shield Portable!
My only doubt is fast RAM performance, idk if they can get something on a par with the 32mb the Wii U had, but it's outside my brainzone.
I can't see such a chunky handheld even being remotely within the realm of possibility, especially since there's no real reason for them to do so. They can match Wii U in a 3DS-sized casing for $200 or less simply by using Pascal. If that's too new, they'll end up just not having something as powerful as Wii U. I'm guessing that you're talking about a hybrid here?
Schnozberry
Member
I don't think there are necessarily two systems so I don't need to do such a thing
I think Nvidia will be able to provide something that can outperform the Wii U in a portable form. That might require a handheld chunkier than we're used to, say a large 2 inch+ deep box, like...um... the Shield Portable!
My only doubt is fast RAM performance, idk if they can get something on a par with the 32mb the Wii U had, but it's outside my brainzone.
They wouldn't need to replicate the EDRAM if they use LPDDR4 with a wide enough bus. Framebuffer is smaller at what we're assuming the resolution will be (540p)
Thanks for clarifying, it's easy to get bogged down in arguing over the semantics of what Iwata or Kimishima has said while completely forgetting that we're arguing over the wrong semantics (i.e. the translated English statements rather than the original Japanese).
While there are some technical and cost issues with such a plan, the main reason I wouldn't see this happening is that it's approaching the challenge of satisfying both target markets from the opposite direction I would expect from Nintendo. That is, it's an attempt to take a home console and make it appealing to the portable market, rather than the other way around. As Nintendo has historically had a much larger market with portables, and that gap has only grown bigger over the past generation, if they were to attempt to design one device to sell to both sets of customers, I would imagine they would place much more emphasis on the segment which has been the primary source of their business. Although I don't believe the mythical "hybrid" handheld which you place in a dock to play on your TV is at all likely, I do think it's more likely than a home console that you can take on the road.
The 2x Denver and 4x A57 are actually some of the only specs Nvidia have confirmed about Parker. Which is the primary reason I wouldn't see Nintendo using an off-the-shelf version of the chip. Why Nvidia is using A57s as opposed to A72s is anyone's guess, perhaps with A72s on board the proportion of tasks where Denver has the performance advantage is just too small. Otherwise it might be the case that the chip was taped-out quite a while ago, and they've simply been waiting for 16FF+ to mature and customers (or even a single customer) to come along. If I were designing the NX SoC (for a home console) I'd just stick with A72s and a couple of A35s for those background OS duties. Keep the CPU arch homogenous for developers and avoid the optimisation headaches which would come with Denver.
Regarding allocating threads between Denver and A57, I'm sure it would be doable in software with some kind of heuristics to determine which threads to send to which cores (similar to HMP-based big.little ARM arrangements). My guess is that they could run a wide variety of real-world code in a laboratory setting on both Denver and A57, and use the results of that to develop heuristics which could be applied in the task scheduler, either proactively (i.e. anticipating threads which perform poorly on Denver before even running any code) or reactively (i.e. knowing when a performance drop is due to code optimisation overhead and migrating to an A57) or more likely a combination of both. Of course the latter presents another problem; is Denver register-compatible with ARMv8? If not then thread migration just presents another headache to deal with.
I absolutely agree that Nintendo won't consider performance, in absolute terms, to be essential to their vision with NX. I do, however, think that there is
- A reason to believe that Nintendo would target ballpark competitive performance with a home console to attract better third party support after the failure of the Wii U
- A reason to believe that they will try to bring their home console and handheld as close together as possible to make cross-device development as easy as possible
I don't think a shared library is, in a direct sense, a selling point for the home console, though. I don't think Nintendo would attempt to sell people an NX home console primarily on the basis that they'll also be able to buy a handheld which plays the same games. I think the benefit of the shared library is indirect, that is it massively increases the number of first-party releases which will be available for each device, and I don't believe that is particularly diminished by separating the release of the two devices by even as much as a year. More games means more games, even if the technological justification behind it isn't immediately obvious.
Graphics Horse
Member
Maybe at 540p, but I've ditched that assumption.
IT's not a handheld.
And the portable version needs a pc. It only plays streamed games afaik.
Graphics Horse
Member
It does both. And what's not a hh? What is this thread about?
Schnozberry
Member
I think the Denver + A57 variant was built for release in January as a part of the Drive PX2 design. Perhaps Nvidia didn't have enough experience with the finalized A72 design to be able to use it for that particular chip. They are an ARM licensee, but unlike MediaTek, Rockchip, and Qualcomm, they didn't help design the A72 and probably didn't have early access.
As far as a I know, Denver is not register compatible with ARMv8, and instead uses a hardware ARM decoder prior to code execution. I am in total agreement that a A72 + A35 design would be superior.
Prisoner KSC2-303
Member
I think we'll have to agree to disagree. I think the shared library is sounding like the main selling point for both systems. A traditional home console, even with more first party titles thanks to the reduced development time, still only sells to the same crowd as those who buy Nintendo home consoles anyway (i.e. people like me who always buy them). We're at the point where that pool is shrinking, so they need to draw in the handheld market they so dominate. A shared library does that.
Opening up the variation and greater third party support their handhelds have received would be significant. Delay that too long, and you enter the period when people start looking towards the PS5/XBwhatever, and then Nintendo face the same problem of their home console looking dated and not feature complete before it's proven itself.
I still feel that we may see the handheld sooner rather than later.
Schnozberry
Member
I wouldn't assume the native resolution of the display will be 540p either. What resolution games are rendered at is another variable altogether. 540p upscales cleanly on 1080p displays, and on a mobile screen 5 inches or smaller it would probably look fine.
You're brushing off a pretty huge issue in your theory then, especially if you expect a hybrid.
It is a handheld running Android. There is also a tablet. Both at $199. Check them:
https://shield.nvidia.com/store/portable
https://shield.nvidia.com/tablet/k1
There's different levels of dedicated machines. We are talking about Nintendo. The company that had to lower the price of their 3ds to $170 a few months after it launched at $250.
They aren't going to release a $500 handheld.
I'v seen them. What's your point? That it can play Flappy Bird?
Nintendo makes a $200 handheld called the 3ds XL.
Wait, each SOC in Denver matches Parker's specs? That's odd... Super odd, since that makes it less efficient than TX1 on paper.
It can play quite a bit more than Flappy Bird. It's more capable than 3DS and Vita combined and tripled. The tablet beats Wii U handily.
I never said that they would release a $500 handheld.
My point is that developers won't be targeting a huge range of devices with a variety of capabilities. They'll be targeting a relatively small number of devices (probably two, at least initially).
As such they'll be able to push these devices further, because they won't have the baggage inherent in mobile development.
It's not what you can do with it, it's what it is inside. That's the point. A slightly better version of what's inside those could do wonders for Nintendo.
Do you even know what thread you are in? You discuss about $500 handhelds, why?
Look up the price of an external gpu slot for a pc and get back to me. Parallel graphics processing requires more than a simple hdmi or usb type port.
Again I'm not saying it isn't doable. It just doesn't make any practical sense. It's a clumsy solution too. Putting in a faster sister chip makes more sense if they are going to add cost to the console anyway.
Never did she said that it was weaker. The stretch means that it could be on par.
Still this doesn't mean the games will look like ps4 visually, we just don't know.