Originally Posted by Serios
You are the one that doesn't understand. Even if AMD would have published Mantle's source code right then when they announced it Nvidia would still not support it. That is what they said.
It was never about "we would support it but we can't" it was always "we won't support it".
Things are simple.
What are you talking about?? Async compute is not AMD's tech, they don't own it, they simply implemented in their GPUs when Nvidia didn't. How can AMD give Nvidia something it doesn't own?
Also nvidia's software implementation is limited, AMD has full hardware support, they don't need a software implementation.
AMD still supports Mantle for those developers that want to implement it, they just won't release general drivers and improvements any more.
There are two things: DX12 and Vulkan are identical to Mantle so it doesn't make sens for AMD to invest in a 3rd API and Nvidia was asked about Mantle support before DX12 and Vulkan were on the table as options and they have denied support right from the very start.
“Khronos Group has chosen the best and brightest parts of Mantle to serve as the foundation for Vulkan.”
Which was acknowledged by Khronos when they thanked AMD.
OpenCL and CUDA are not identical. For example OpenCL has better support for heterogeneous computing and can work on both CPUs and GPUs while CUDA is more concentrated around the GPU and it's main advantages it that it gets updated faster.
The problem is not Nvidia killing CUDA but what will happen as things go foreword.
AMD said "it will be open source" but they also said they will control it. Would you agree to your competitors controlling your drivers and code, before that code was even published?
it is just a spin to make them look good. AMD would never have published it. And Nvidia would not accept AMD to controlling an "open source" API, which was not even published and was favouring AMD's hardware.
If you had listened to nvidia talk in the podcasts when mantle was announced, you would have understood instead of blabbering. Nvidia said that as long as mantle is controlled by AMD and it is not published, and hardware specific to their hardware, they will never support it.
If AMD first published it as open, gave full control to someone else (like khronos which is also supported by nvidia as well as AMD), nvidia would possible support it. But not closed "open source".
DX12 and vulkan are not identical to mantle. Especially since mantle is very hardware dependent on AMD internal API (read about the BF4 nvidia fiasco to optimise the game and why it took so long).
They are taking ideas
from mantle, not the API or the hardware dependencies in order to allow several manufacturers. This is not the same. They also took DX12 ideas into, nvapi, openGL, they are working not with just AMD but also Nvidia.
The same as Mantle took ideas from nvapi (which was way before mantle was even considered) or from OpenGL.
This is the reason why nvidia will support it. Because it is not controlled by AMD.
Regarding CUDA, it is not only concentrated about the GPU. And it will not die until someone else comes along. Which will also bring the end of OpenGL as it is now.