This has been proven to be factually inaccurate.
3DVagabond, you asked what had been debunked, this claim right here.
Gameworks on the other hand directly impacts the performance of AMD cards, and since it is closed source, AMD cannot optimize their cards for the DirectX games that use it. The devs are also forbidden from helping AMD to optimize their cards for that game.
I think what he meant is there is nothing locking it to GCN. AMD has stated that in order for it to work, a card has to support a list of features. For AMD, GCN was the first card to support all those features. For nVidia, it is Fermi as I recall. Not sure what Intel GPU's would be supported. Most likely HD4000 and above.
I have never heard Fermi based cards will be capable of running Mantle. I find it hard to believe given the closer to metal nature of the api. My understanding of AMDs position on this is the API will be available to anybody provided they are using GCN. Which means it is is as open as PhysXs requirement the GPU be designed around CUDA for GPU-Physics.
I have never heard Fermi based cards will be capable of running Mantle. I find it hard to believe given the closer to metal nature of the api. My understanding of AMDs position on this is the API will be available to anybody provided they are using GCN. Which means it is is as open as PhysXs requirement the GPU be designed around CUDA for GPU-Physics.
This has been proven to be factually inaccurate.
3DVagabond, you asked what had been debunked, this claim right here.
Mantle is open
Mantle is not open. AMD has sole discretion about what is included and how. They have complete control. That's the direct opposite of open.
Yep sadly got to agree; until sdk kit is released which will be later this year as its still in beta.......its closed as of right now; but won't be later on.....I think that is the difference......
Can AMD develop a card/driver that uses/interprets Cuda code? That's the difference between open and closed. CUDA equals for Nvidia hardware only, Mantle means everyone can use it even if AMD is the only one that does.It doesn't matter at all when the SDK is available it will always be a closed API just like CUDA. Having a SDK doesn't make it a open source period, by that definition CUDA is open as well.Mantle is optimized for GCN architecture and it will be a stupid endeavor for NV to optimize for it when the DX12 is just around the corner.Also I don't believe you can just optimize it for other architectures as you wish, we need to see the eula first. In the end I like both the vendors are pushing for higher fidelity in PC gaming, it is not just the struggle for existence but for supremacy as well.
DX is an open standard, OpenGL is an open standard, DDR is an open standard, Mantle will be an open standard, and so on.
You think open standard means that anyone and their brother can throw in their opinions and have them be taken seriously during development.
Can AMD develop a card/driver that uses/interprets Cuda code? That's the difference between open and closed. CUDA equals for Nvidia hardware only, Mantle means everyone can use it even if AMD is the only one that does.
....
Yes, that's exactly what I think, because that's exactly what open means. Anything other than that is not open.
The opposite of open is when one entity gets to dictate 100% of the development. That's what is happening with Mantle.
Open just means free in this respect. DDR3 is an Open Standard of Jedec where RDRIMM was an closed standard for Rambus (they got royalties for RDRIMM manufacturing and for a little while DDR manufacturing). That means anyone can manufacturer memory to the DDR3 specs and call it DDR3 as long as it met all the requirements. But that doesn't mean you tell Jedec what to include. Still a big boy club that decided the specs.....
Yes, that's exactly what I think, because that's exactly what open means. Anything other than that is not open.
The opposite of open is when one entity gets to dictate 100% of the development. That's what is happening with Mantle.
Well, why not just say that it's open for anyone to implement if they so desire. Good or bad, that's pretty much the short of it. That said, NV doesn't have ANY efforts even closely related to that.
That's my point even if a driver wrapped could be made for Cuda to work with GCN, it wouldn't matter. It's a proprietary API (closed)that Nvidia refuses to license out to other hardware manufacturers which means AMD can't, until the day where Nvidia at least opens up for licensing, develop Cuda capable hardware. This is the same for Physx. The only thing stopping Nvidia from making Mantle is 1.) it isn't an Open standard yet (hasn't been "released" to the public) and 2.) their own opinions on how to handle competition at this time.CUDA only works for NV and even if it didn't it would be foolish for AMD to optimize for it.
That's my point even if a driver wrapped could be made for Cuda to work with GCN, it wouldn't matter. It's a proprietary API (closed)that Nvidia refuses to license out to other hardware manufacturers which means AMD can't, until the day where Nvidia at least opens up for licensing, develop Cuda capable hardware. This is the same for Physx. The only thing stopping Nvidia from making Mantle is 1.) it isn't an Open standard yet (hasn't been "released" to the public) and 2.) their own opinions on how to handle competition at this time.
And why should they?
Why should they spend development time on something that AMD could change arbitrarily and unilaterally?
AMD controls Mantle. Mantle is not open. It doesn't matter if you don't have to pay for the SDK.
I dunno... because it's in the best interest of gamers (i.e. us, the customer)? I say why not give it a shot. Hell, for all we know, NV's implementation could be even better than AMD's.