AMD and NVIDIA only want to use low-level APIs, why?

As we say, it is not a single process or single decision that has made this decision, far from it, but sales are marked by one thing: performance / value ratio.

And it was here that the two of them played a few years ago cake to distribute between the two, because the excuse was to gradually force all users to accept Windows 10 due to Microsoft's refusal to send DX12 to Windows 7 or less.

Low quality APIs, a holy grail for efficiency and efficiency

It's all about low-level APIs, work that AMD started with Mantle and that advanced APIs at that time like OpenGL and Direct3D took months to see. But why did both companies jump into the arms of Microsoft and DX12 when Windows 10 came out?

We all know the story behind Mantle and DX. The official implementation of the standard API meant that the first one to use it, interact with it, and support it to its drivers would get performance from the same machines.

High-quality API has a very clear advantage over a single low level: it supports a large number of GPUs without the effort of a game plan or driver to make it all work together.

The problem is that the advanced API also has a very important problem and since then both NVIDIA and AMD felt they had to run away.

The DX9, DX10, and DX11 provide efficient fines

Even though we dealt with this at the time, it is important to understand it so that you understand why everything is currently about customization and optimization. We all know that the API most commonly used is Microsoft & # 39; s DirectX, and as a result, the company offered some benefit in its own unique versions.

This has prevented AMD or NVIDIA from wasting R&D resources out of the staff to better their drivers with game engines, and at the same time made it easier for developers to put the management part in Microsoft's hands. engine and GPU services.

DirectX was therefore up to version 11 as a resource manager between Hardware and software that prevented both from seeing each other more than necessary and therefore did not interfere with direct resource management and management. However administrators become so sophisticated and sophisticated that APIs begin to implement what is called "more" in a more continuous way.

API calls to Hardware and software have doubled the complexity of both, and it was first recognized that improvements between developers, APIs and Hardware were a disaster in many cases. For this reason, AMD has been working for a long time in the development of Mantle and Microsoft is doing the same with the DX12.

DX12, an API that just finished putting everything in its path

Popular mobile phones are no longer a problem with DX12, NVIDIA and AMD already have a legitimate background vendor that will work with their APIs and developers, but now what has changed about DX11 and its predecessors? So much for managing resources and accessibility alike. DX12 is a low-level API, similar to Mantle, but much improved (Mantle was "donated" to the Khronos team to create the Vulkan).

What distinguishes DX11 from DX12 is the so-called abstraction, that is, with DX12 the game engine engineer or driver still has full access to their code synchronization rather than hardware. This has a definite benefit, and there's nothing more than doing well for software developers in relation to AMD and NVIDIA Hardware, it's achieved great performance thanks to the fact that drivers are often "very clear" when it comes to operating and allocating resources. on hardware.

As a whole, this has a very clear counterpart: it requires more resources from engineers and program managers, that is, it requires more R&D, more staff and more work to change the code in each design. Although these GPU structures are almost identical as a rule change to another and are therefore compatible in terms of driving conditions, technologies such as Ray Tracing or DLSS traverse the invention and require new, low and complete APIs.