Microsoft has rolled out DirectX 12 Agility SDK 1.620 Retail and 1.722 Preview, bringing major upgrades for ray tracing, GPU workload efficiency, shader delivery, and developer tools. The latest preview is especially important for game developers because Microsoft describes its new DirectX Raytracing improvements as the biggest DXR update since the feature first arrived in 2018.
The main focus of DirectX 12 Agility SDK 1.722 Preview is making ray tracing faster and more practical in highly detailed, dynamic game worlds. To achieve this, Microsoft is introducing Clustered Geometry, Partitioned Top-Level Acceleration Structures, and GPU-driven acceleration structure builds.
Ray tracing depends on acceleration structures, often called bounding volume hierarchies, to help the GPU quickly determine how rays interact with objects in a scene. These structures are essential for realistic reflections, shadows, and lighting, but they can be expensive to build and update, especially in modern games with huge environments, streaming assets, animated characters, and constantly changing geometry.
Clustered Geometry is designed to reduce that cost. Instead of treating geometry as one massive collection of triangles, Microsoft’s approach divides it into smaller reusable clusters. Each cluster can contain up to 256 triangles and vertices. Developers can then assemble these clusters into larger ray tracing acceleration structures without rebuilding the same geometry repeatedly from scratch.
This could be a major step forward for PC game graphics. In many games, developers have to balance visual quality and performance by using simplified geometry for ray tracing. With Clustered Geometry, it may become easier to use the same high-detail assets for both traditional rasterization and ray tracing. That could result in more accurate reflections, better shadows, and more consistent lighting without dramatically increasing GPU overhead.
Another important feature is Partitioned Top-Level Acceleration Structures, or PTLAS. This allows developers to update only the parts of a scene that have changed instead of rebuilding the entire top-level structure. In large open-world games or scenes with many moving objects, this could save a significant amount of processing time.
Microsoft is also adding GPU-driven acceleration structure builds. By allowing the GPU to handle more of the work directly, developers can potentially reduce CPU bottlenecks and improve performance in complex scenes.
To show what these technologies can do, Microsoft demonstrated a development sample comparing traditional Bottom-Level Acceleration Structure rebuilding with the new Clustered Geometry approach. In one test running on an AMD Radeon RX 9070, performance increased from 1.2 FPS using traditional triangle-based geometry to 79 FPS with Clustered Geometry. That is roughly a 66x improvement in that specific scenario.
The same test showed acceleration structure build time dropping from 788 ms to just 3.11 ms. GPU ray tracing time also improved from 72.6 ms to 9.27 ms. Total acceleration structure memory use fell from about 7.3 GiB to 774 MiB, a reduction of nearly 90%.
Here is a simplified breakdown of the sample results:
Traditional BLAS performance: 1.2 FPS
Clustered Geometry performance: 79 FPS
GPU acceleration structure build time: reduced from 788 ms to 3.11 ms
GPU ray tracing time: reduced from 72.6 ms to 9.27 ms
Total acceleration structure memory: reduced from around 7.3 GiB to 774 MiB
Microsoft cautions that these numbers come from a specific development sample and should not be treated as a guaranteed performance boost for every game. Real-world improvements will depend on the game engine, scene complexity, implementation quality, driver support, and GPU architecture.
Even so, the results point to a promising future for ray-traced games. As developers push toward richer environments with more detailed assets, more animation, and more dynamic lighting, reducing the cost of acceleration structure updates becomes increasingly important.
Microsoft also notes that similar ideas have already appeared in high-end graphics technology demonstrations and upcoming games. By bringing a standardized version of these capabilities into DirectX, the company may make advanced ray tracing techniques more accessible across different GPU brands.
Current preview support includes all NVIDIA GeForce RTX GPUs, AMD Radeon RX 9000 Series graphics cards, Intel Arc B-Series GPUs, and supported Intel Core Ultra integrated graphics. Microsoft says the features do not require entirely new hardware beyond existing ray tracing support, but compatible drivers are still required.
Alongside the ray tracing improvements in Agility SDK 1.722 Preview, Microsoft has also released DirectX 12 Agility SDK 1.620 Retail with updates to Advanced Shader Delivery.
Advanced Shader Delivery is designed to improve how shaders are distributed, compiled, and reused. Shader compilation stutter has been a common issue in PC gaming, and better shader management can help developers reduce loading times, compilation overhead, and runtime performance issues.
One of the key additions is Partial Graphics Programs. This feature allows developers to reuse precompiled shader components across multiple pipeline configurations. In practice, that could reduce shader package sizes and make shader handling more efficient.
Advanced Shader Delivery now has broad support across the Windows 11 ecosystem, including compatible GPUs from AMD, Intel, NVIDIA, and Qualcomm. However, the new Partial Graphics Programs optimization is currently limited to AMD Radeon RX 7000 and RX 9000 graphics cards. Microsoft says support from other GPU vendors is expected in the future.
The retail SDK also includes several developer-focused improvements. DirectX Dump Files will help diagnose GPU crashes more effectively, making it easier for developers to identify and fix graphics-related problems. Fence Barriers Tier 1 and Batched Asynchronous Command List APIs are also included, along with new DirectX Compute Graph Compiler interfaces aimed at machine-learning workloads.
Shader Model 6.10 remains in preview for now. Microsoft plans to release it in retail form with DirectX 12 Agility SDK 1.621, expected in early December.
For players, these changes will not automatically make existing PC games run faster. Developers need to adopt the new tools and build support into their engines and games. However, for future titles, especially those targeting advanced ray tracing, large environments, and complex rendering pipelines, these DirectX 12 updates could be highly important.
With Clustered Geometry, PTLAS, improved shader delivery, and better diagnostic tools, Microsoft is giving game developers more ways to build visually ambitious PC games while managing performance more efficiently. If widely adopted, these technologies could help shape the next generation of real-time ray tracing on Windows PCs.
