• 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.

Nvidia CEO: Relationship with Nintendo "will likely last two decades"

Status
Not open for further replies.

AmyS

Member
The 20 years number likely isn't a coincidence either, the previous console GPU partnership lasted nearly that long (Silicon Graphics with N64, ArtX with GC, ATi with Wii and AMD with Wii U were the continuity of one team first splitting off and then being bought multiple times).

nawtqwB.jpg
 

joesiv

Member
Nintendo's decision to stick with PPC750 for so long wasn't just a desire to use precisely the same tech for as long as possible, it was because there literally weren't any other options for them while retaining full backwards-compatibility.
I don't have a source, because this was many years ago, but I recall that Nintendo's original deal with IBM, was actually for 3 generations of chips. The WiiU would just be the fulfilment of that deal. It was probably to avoid having to work out viable deals each generation.
 
That's true, it may not be quite the same.

I also wonder how low level or interoperable the new NVN API will be. I seem to remember a dev saying with either ease of use or low level programming flexibility, the Wii Us API had neither while the Switch's has at least one.

Sony for instance is sticking to Jaguar cores and clocking them the same as the base PS4 in unpatched games for perfect compatibility, while even the XBO S uses higher clocks for a bit better performance, so perhaps there are differences in how low level vs portable the XBO version of DX and the PS4s GNM are.
NVN API is probably a proprietary, closed, Nvidia-licensed API. Kinda like PhysX or GameWorks. Doesn't sound good to me, but I hope I'm wrong...
 

NeOak

Member
The GPU is pretty much the least likely part of the console design to break compatibility if the manufacturer changes.

Not really.

Consoles give low-level access to the GPU and this complicates BC... PCs are different, because they utilize high-level API abstractions.

There's a reason OG XBOX emulation is still not possible (even on XB1) and Sony had to emulate GCN 1.1 shaders with base mode/downclocking.

Its not trivial, but the idea people are writing assembly and coding to the metal to maximise hardware specific fixed functions is less true today than it ever has been.

OG Xbox emulation was possible on 360, presumably it would be on One as well.

Wrong. It is a matter of patents. The original Xbox had chips designed by their respective companies, so they keep the design under copyright law.

This is what is going to fuck up Backwards compatibility of the switch later on.

NVIDIA has to give a license to MS in order to emulate the NV2A inside the OGXB on the XBO. They gave a license for the 360.

This is why pretty much with the 360/PS3/Wii, companies order the design and they "own" it. Except the GTX 7800/RSX on the PS3 which again, it's an NVIDIA off the shelf chip which would require a license from NVIDIA in order for the PS3 to be emulated by Sony on PS4 or whatever console.

XBO can do OGXB no problem because the GPU can emulate Pixel/Vertex Shaders 1.1, but they need a license for the custom NVIDIA extensions that the games use.

See here: http://www.gamesindustry.biz/articles/nvidia-tech-licensed-for-xbox-360-backward-compatibility
 
Wrong. It is a matter of patents. The original Xbox had chips designed by their respective companies, so they keep the design under copyright law.

This is what is going to fuck up Backwards compatibility of the switch later on.

NVIDIA has to give a license to MS in order to emulate the NV2A inside the OGXB on the XBO. They gave a license for the 360.

This is why pretty much with the 360/PS3/Wii, companies order the design and they "own" it. Except the GTX 7800/RSX on the PS3 which again, it's an NVIDIA off the shelf chip which would require a license from NVIDIA in order for the PS3 to be emulated by Sony on PS4 or whatever console.

XBO can do OGXB no problem because the GPU can emulate Pixel/Vertex Shaders 1.1, but they need a license for the custom NVIDIA extensions that the games use.


So, thanks NVIDIA for fucking up Backwards compatibility. We will shit on you 20 years from now when we can't get Switch BC thanks to your shit.
Fair enough. Licensing is one thing to consider, but technical difficulties are another obstacle to overcome...

Keep in mind that OG XBOX was originally intended to be a DirectX Box (aka high-level API), which was stupid for a console environment with limited resources. Eventually MS caved in and they gave low-level GPU access with an appropriate API.

Even if they get a license from Nvidia, it'll be really hard to map NV2A low-level calls to the equivalent Radeon ones.

XBOX 360 had a Radeon GPU with unified (not discrete) shader pipelines and this time around MS didn't allow programmers to have that low level access to the "metal". Considering the fact that XB1 also has a Radeon GPU with unified shaders (even though the uArch is different), a fully virtualized OS structure and of course no legal obstacles, this makes BC infinitely easier. The CPU is not really emulated, they just recompile PPC binaries to the x86-64 equivalent, which seems to work fine for the most part.

Sony thanks to their Nvidia/RSX partnership are royally screwed legal and tech-wise, even if Cell is "emulated" somehow.
 
Bububu Nintendo won't be making hardware in two decades!

Bingo, LOL. I also remember back when many people (including 3rd party publishers) thought that the PSP would be the "Game Boy Killer." Look what happened:

  • Little to no advertising
  • Most of it's software sales suffering
  • Most of it's software consisting of being remakes along with Dreamcast, PS1 & PS2 ports that almost nobody wanted to buy

The main reason PSP sold as well as it did was because of people were buying it for piracy & for Monster Hunter. The only reason MH couldn't be on the original DS was because that it wasn't powerful enough to run the games on it properly. Now that Nintendo has brought out the 3DS that's more powerful than PSP is, Capcom no longer needed either PSP or PS Vita for MH anymore.

If not for that & piracy, PSP would've been completely dead like PS Vita is now.

People who underestimate Nintendo so much have always been proven to be wrong.
 
Bingo, LOL. I also remember back when many people (including 3rd party publishers) thought that the PSP would be the "Game Boy Killer." Look at what happened:

  • Little to no advertising
  • Most of it's software sales suffering
  • Most of it's software consisting of being Dreamcast, PS1 & PS2 ports that almost nobody wanted to buy

The main reason PSP sold very well was because of people buying it for piracy & Monster Hunter. The only reason MH couldn't be on the original DS was because that it wasn't powerful enough to run the games on. Now that Nintendo has brought out the 3DS that's more powerful than PSP is, Capcom no longer needed either PSP or PS Vita for MH anymore.

If not for that & piracy, PSP would've been completely dead like PS Vita is now.

People who underestimate Nintendo so much has always been proven to be wrong.
There were no smartphones/tablets when DS/PSP came out. 3DS will never reach DS sales.
 

Durante

Member
NVN API is probably a proprietary, closed, Nvidia-licensed API. Kinda like PhysX or GameWorks. Doesn't sound good to me, but I hope I'm wrong...
Generally, all console APIs are proprietary and closed. Far more so than PhysX and large parts of GameWorks actually.

I'm also not a fan of this state of affairs, but no one seemed to mind before.
 
Generally, all console APIs are proprietary and closed. Far more so than PhysX and large parts of GameWorks actually.

I'm also not a fan of this state of affairs, but no one seemed to mind before.
I don't agree. AMD provides documentation and source code for everything. They even made Mantle available to other vendors and this spurred DX12/Vulkan development.

GNM is similar to Mantle, so how is that closed and proprietary? I think you're confusing it with the PS4 SDK in general, which is not available to everyone.

There's a difference between Nvidia making a proprietary API for Nintendo and Nintendo getting documentation/source code from Nvidia to develop their own API.

And yeah, people mind it because of the NV2A/RSX fiasco. Nvidia is not exactly the most "open" company (G-Sync vs FreeSync, PhysX vs TressFX, GameWorks vs GPUOpen etc etc.) It's a perfectly valid concern IMHO.

What do you think will happen if AMD makes a competitive mobile SoC in the future and Nintendo wants to use that for Switch 3? Will it be compatible with the NVN API to offer Switch BC?
 

Durante

Member
GNM is similar to Mantle, so how is that closed and proprietary?
First of all, Mantle is proprietary, despite all the noise about being open. It wasn't even accessible to anyone but a select few partners for the years where it was a thing. I know, I tried to get the spec.

Secondly, can I go somewhere and download GNM, or at least a specification for GNM? How do I get certified as a GNM implementer? How do I extend it? I can't, it's closed and proprietary.

The only actually open graphics APIs in any relevant sense are OpenGL and Vulkan.
 
First of all, Mantle is proprietary, despite all the noise about being open. It wasn't even accessible to anyone but a select few partners for the years where it was a thing. I know, I tried to get the spec.

Secondly, can I go somewhere and download GNM, or at least a specification for GNM? How do I get certified as a GNM implementer? How do I extend it? I can't, it's closed and proprietary.

The only actually open graphics APIs in any relevant sense are OpenGL and Vulkan.
And why would want to use GNM for PC development? Vulkan is an excellent GNM substitute.

GNM is not a closed API developed by AMD and given to Sony. It's not the same thing as the NVN API. I want Nintendo to have full access to documentation/source code and develop their own API, even if it's closed. It makes a huge difference.

For me, it's a matter of corporate policy. Nvidia is not as open or flexible as AMD. It's like comparing Windows vs Linux.
 

Durante

Member
For me, it's a matter of corporate policy. Nvidia is not as open or flexible as AMD. It's like comparing Windows vs Linux.
I think you are attributing human qualities to corporations in a way that doesn't make much sense. They both operate by contracts and try both try to act in a way that they believe maximizes their profits, based on their market positions. That's it.

It's not at all like Windows and Linux. Comparing something like DirectX, GNM, Mantle, Glide or NVN to OpenGL or Vulkan is like Windows and Linux.
(Well, in a way, GNM, Mantle and NVN are more closed than Windows or DirectX. Even publishing the specification and documentation is prohibited)
 
I think you are attributing human qualities to corporations in a way that doesn't make much sense. They both operate by contracts and try both try to act in a way that they believe maximizes their profits, based on their market positions. That's it.

It's not at all like Windows and Linux. Comparing something like DirectX, GNM, Mantle, Glide or NVN to OpenGL or Vulkan is like Windows and Linux.
(Well, in a way, GNM, Mantle and NVN are more closed than Windows or DirectX. Even publishing the specification and documentation is prohibited)
Even if AMD is more open/flexible because of Nvidia's dominant position on the market, this still doesn't change the facts.

So, why is an Nvidia-developed API a good thing? Why is getting full access to documentaton & source code a bad thing? Do you know why open-source Radeon Linux drivers exist in the first place? Why is there no Nvidia equivalent?
 

Durante

Member
So, why is an Nvidia-developed API a good thing?
Compared to Nintendo handling it on their own? Because Nvidia is really good at providing state-of-the-art development tools. E.g. Nsight is absolutely amazing.

Why is getting full access to documentaton & source code a bad thing?
Where did I say that? That's silly.

Do you know why open-source Radeon Linux drivers exist in the first place? Why is there no Nvidia equivalent?
Because AMD is not willing to commit the internal resources to provide full production-quality and up-to-date Linux OpenGL drivers on their own. The Nvidia equivalent is a closed driver that actually works, is feature-complete, and fast. That's why people who need GPU graphics performance (or compute for that matter) on Linux overwhelmingly buy Nvidia.
 
Compared to Nintendo handling it on their own? Because Nvidia is really good at providing state-of-the-art development tools. E.g. Nsight is absolutely amazing.
Microsoft dev tools are also amazing, but they promote vendor lock-in...

The fact that we still don't have a vendor-agnostic physics API is Nvidia's fault ever since they bought Ageia. I don't like embrace and extend strategies. That's why I'm very wary of the NVN API.

At least DirectX is GPU vendor-agnostic, even though it's limited to Windows platforms...

Because AMD is not willing to commit the internal resources to provide full production-quality and up-to-date Linux OpenGL drivers on their own. The Nvidia equivalent is a closed driver that actually works, is feature-complete, and fast. That's why people who need GPU graphics performance (or compute for that matter) on Linux overwhelmingly buy Nvidia.
Fair enough. Open-source reduces company costs, but on the other hand it gives power to the community.

What will happen if Nvidia stops releasing closed-source drivers?
 

Durante

Member
What will happen if Nvidia stops releasing closed-source drivers?
People will have to buy different hardware. But their software (which is OpenGL) should work on everything.

I'm all for supporting true open standards like Vulkan or OpenGL. Seriously, I'm 110% behind you on that.
What I dislike is portraying one type of closed proprietary product as more "ethical" than another.
 

Thraktor

Member
I still strongly suspect that either NVN is Vulkan-derived (and not too far from the Vulkan spec), or they're also offering Vulkan as a non-afterthought alternative to NVN (unlike OpenGL on their previous consoles, which was apparently very poorly supported).

We know that Nintendo has been on Khronos's Vulkan Working Group since last year, and there's no reason for them to be there if they're not concerned about interoperability between software running on their platform and others.
 
Sony thanks to their Nvidia/RSX partnership are royally screwed legal and tech-wise, even if Cell is "emulated" somehow.

Agreed.

Unless Sony ports their other PS3 titles over to PS4 like they did with Beyond: Two Souls, Heavy Rain, TLOU, GoW 3 & Uncharted 1-3, they're sunk in future BC with PS3 games.
 

Buggy Loop

Member
I'm not the biggest Nvidia fan for the GPU PC market. (Mostly how they hide some gimped features to their customers, they are still very good hardware)

But this is probably the best move Nintendo could do in its current state. With Nvidia's help with tools, API & hardware development, they might finally have a competent product technologically. I hope Nvidia does some PR hype to showcase their chip around january.

Nvidia has been on a roll lately with Tegra chips in fact, they scored Nintendo, Tesla's autopilot which will be included in all tesla models from now on, they also scored some cancer diagnostic AI contract with the National Cancer Institute & Department of energy.
 

dr_rus

Member
- The fact that Nvidia are including Nintendo's business in their Gaming division, rather than their IP licensing division, would indicate that Nvidia are handling manufacturing and selling the final chips to Nintendo, rather than Nintendo licensing the design and handling manufacturing themselves (which was their arrangement with AMD for Wii U's GPU).
Doesn't mean anything, they are completely in their rights to include it anywhere they want. I would be _very_ surprised if NV handled anything more than the chip development for Nintendo here, it's very likely that the SoC itself will be Nintendo branded and produced at TSMC or Samsung by Nintendo. Zero reasons why NV should be involved here.

- Working with Nintendo for "almost two years" means design work on Switch's SoC likely started at the end of 2014/start of 2015. That would seem like a pretty typical timescale for a custom chip like this.
Switch SoC is Parker based and it's unlikely that they've been working on it _with_ Nintendo two years ago. What some people seem to be missing though is that NV is providing almost the whole software stack for Switch, with both OS and main API being developed by NV in a large part at least. This is the most possible area where their collaboration may have started almost two years ago.

This last bit is speculation on my part, but when he says "over the next" and then cuts off to talk about Nintendo using the same architecture for a long time, I get the impression that Nvidia already has contracts for future Nintendo hardware that he's trying not to talk about. Claiming that the relationship will last "two decades" is also pretty confident even for Huang, and I have a feeling that this is further evidence that Switch is the start of the "family of systems" which Iwata talked about, all revolving around ARM/Nvidia SoCs and a common software platform.

Unless Nintendo is already developing the console after Switch - which is unlikely in my opinion - there is no solid contracts to talk about but it's pretty apparent that NV will be the first option for Nintendo for any future h/w since maintaining backwards compatibility is important for them.
 
Status
Not open for further replies.
Top Bottom