Microsoft has published a new DirectX 12 Agility development kit, to include new features that in the hands of programmers should be key to improve gaming performance.
You already know that DX 12 is the latest version of Microsoft’s Application Programming Interface (API). They are libraries used for the creation of multimedia and video applications, especially video games, and although there are other development standards (such as OpenGL and its promising version Vulkan), DX is used massively by the industry and is one of the keys for Windows to maintain its huge market share in computer desktops. Hence, each novelty that it provides is of general interest.
DirectX 12 Agility 1.710.0
The new version of this development kit for libraries launches two functions called GPU Upload Heaps and Non-normalized Sampling. They are intended for programmers, but they should improve game performance at the end-client level by making better use of PC hardware resources.
Especially important is the first function so that the CPU and GPU access VRAM (graphics memory) simultaneously. It must be explained that historically a processor did not have access to the graphics memory of a dedicated one, which forced programs to have to copy large amounts of data to the GPU through the PCI bus.
Most modern GPUs have introduced VRAM Variable Size Base Address Register (BAR) which allows Windows to manage GPU VRAM with WDDM 2.0 or later drivers. The latest version of the DirectX 12 Agility SDK will allow the CPU to access graphics memory at the same time as the graphics card itself. eliminating the need to copy data from the CPU to the GPU.
Microsoft says that this optimization “could offer many performance benefits in computer games”in a context in which video games are huge memory gobblers as their sophistication and visual complexity increase, and VRAM is key to their operation.
A shared memory pool between the CPU and GPU will eliminate the need to duplicate data and will reduce the flow of data that must be transferred under the PCI-Express interface and it should also free up the RAM used by the CPU in this type of task. Memory speed shouldn’t be an issue considering that the dedicated memory that comes with a modern graphics card is a very fast GDDR6. In fact, there will probably be an improvement in latency and access times.
Microsoft has already delivered the new version of the SDK to developers and it should be a matter of time before games take advantage of these types of optimizations. The big three graphics card vendors will also support it.. NVIDIA has already included it in its Game Ready and Studio drivers (version 531.41 or newer), as has Intel for its A and Xe series (with driver 31.0.101.4255 or newer). AMD Radeons will support it shortly and the end customer will only have to update the drivers.
It is difficult to quantify the performance improvement that can be achieved with this function and it will depend a lot on each game and the ability of the developers to take advantage of it. In any case, promising on paper.
