Microsoft Shader Model 6.10 Delivers LinAlg Matrix APIs And Enhanced GPU Raytracing Performance

Microsoft Shader Model 6.10 Delivers LinAlg Matrix APIs And Enhanced GPU Raytracing Performance

Microsoft Shader Model 6.10 Release Advances Graphics Engineering Through LinAlg Matrix APIs Flexible Memory Management And Enhanced Raytracing Tools

Microsoft has officially distributed the Agility SDK version 1.720 preview along with the DirectX Shader Compiler version 1.10.2605.2. Shader Model 6.10 provides developers with essential tools for linear algebra capabilities together with enhanced memory management functions through its newly launched version. The Microsoft DirectX developer blog states that this update optimizes GPU performance for executing complex mathematical computations and raytracing tasks.

The main feature of Shader Model 6.10 introduces the LinAlg Matrix APIs to users. The feature combines two separate projects known as Cooperative Vectors and WaveMMA into one single functional system. Microsoft describes LinAlg as a mid level API that enables hardware drivers to sustain high context levels during their matrix processing tasks. High bandwidth matrix multiplication serves as a crucial requirement for both neural rendering and machine learning systems.

The new system operates across three distinct scopes. Thread scope matrices serve as the standard matrix format which traditional rendering pipelines use to perform lighting inference tests. Wave scope matrices provide access to dedicated multiplication hardware for image processing. The driver now uses ThreadGroup scope to tile management across extensive networks which include large language models as their primary application. The system enables users to maintain performance consistency through one code implementation across different hardware systems.

Shader Model 6.10 introduces new memory usage capabilities which enable GPUs to access previously restricted memory resources. The traditional 32 KB cap on groupshared memory has been replaced with a variable system. The developers can use MaxGroupSharedMemoryPerGroup query to retrieve the actual maximum memory capacity of a specific GPU. The authors of shaders can use GroupSharedLimit attribute to specify their shaders maximum memory usage. The software specification changes enable developers to create advanced algorithms for tile culling and software rasterization which were previously limited by the actual hardware capabilities of their systems.

The update adds a new feature called Group Wave Index. The GetGroupWaveIndex and GetGroupWaveCount intrinsics enable shaders to detect the wave structure which exists within a thread group. The group index division by lane count becomes unnecessary through this method. The change allows wave level cooperation to function across all hardware wave sizes without breaking any code logic.

The special features of TriangleObjectPositions and ClusterID intrinsics provide users with upgraded raytracing capabilities. The first enables any hit or closest hit shaders to retrieve vertex positions which correspond to a particular triangle. ClusterID currently exists but its complete functionality will launch in 2026 after DXR clustered geometry reaches production status. The tools enable raytracing systems to control their geometry interactions at a detailed level.

Microsoft has developed Batched Asynchronous Command List APIs for D3D12. Legacy commands require strict sequential execution which results in GPU delays between each command. The new batching feature allows the driver to overlap independent work within a single call. Modern clear and fill commands provide support for block compressed formats through ClearTextureSubresources which operates directly without needing descriptor heaps. The ergonomic enhancements work to maximize hardware throughput by eliminating all implicit serialization requirements.

Major graphics vendors provide different levels of support for these features. AMD provides support through the Agility SDK Developer Preview Edition driver for Radeon RX 7000 and 9000 series products. Intel has confirmed support for the Arc B Series Graphics with plans for broader implementation later this year. Developers who work with NVIDIA should reach out to their representatives in order to obtain access to developing drivers. The Agility SDK preview and the latest compiler binaries are available for download through the official Microsoft DirectX devblogs and the DirectXShaderCompiler repository on GitHub.

About the author

mgtid
Owner of Technetbook | 10+ Years of Expertise in Technology | Seasoned Writer, Designer, and Programmer | Specialist in In-Depth Tech Reviews and Industry Insights | Passionate about Driving Innovation and Educating the Tech Community Technetbook

Join the conversation

Newsletter Subscription