Microsoft wants to significantly accelerate ray tracing in games with DXR 1.2

Microsoft is launching a new version of "DirectX Raytracing". The company announced this at the Game Developer Conference 2025 (GDC) in San Francisco. This DXR will remain part of DirectX 12 and will be available as a "preview" in the development kits from April. The company has not yet announced when it will be released as a Windows update, as was the case with DXR 1.0 in fall 2018 and DXR 1.1 two years later.

Since ray tracing in Windows games can still significantly reduce frame rates compared to pure rasterization processes, Microsoft has implemented two functions to alleviate this. Firstly, the "Shader Execution Reordering" (SER) introduced by Nvidia with the RTX 4000 generation becomes part of DXR. Developers can leave the order in which shader threads are transferred to the GPU to the driver. This is intended to resolve dependencies and make execution more efficient.

OMM should solve the transparency brake

In addition, a second function invented by Nvidia will become part of DXR, namely "Opacity Micro-Maps" (OMM). This is additional data that is provided for complex objects such as vegetation after they have been checked for transparency in the render pipeline (alpha testing). When the ray tracing then encounters these objects, processing is faster. Even textures with transparencies such as grids should hardly slow down ray tracing thanks to OMM.

With this step alone, Microsoft expects a performance increase by a factor of 2.3 in the case of path tracing, as can be activated in Cyberpunk 2077, for example. Microsoft quotes this figure in a blog post on DXR 1.2. This is likely to be a best-case scenario, as such optimizations are heavily dependent on the scene in question. In addition to SER and OMM, there are, as Microsoft explains in the same place, some updates to tools such as the debugger and profiler "PIX" for DirectX.

Four GPU manufacturers support DXR 1.2

What is remarkable about the new DXR 1.2 is that both SER and OMM are supported by all four well-known providers of ray tracing GPUs for games: AMD, Intel, Nvidia and Qualcomm. The functions are therefore no longer Nvidia-proprietary, but can be addressed via DirectX using a standardized interface. Hardware support is probably not yet available everywhere. In a presentation slide published by Microsoft, Intel states that SER will be supported soon, but OMM will only be supported with future GPUs. Qualcomm refers to upcoming GPUs for both SER and OMM; AMD is not quoted there.

(nie)