Microsoft Agility SDK for DirectX Introduces Shader Execution Reordering, Opacity Micromaps Support, and Major Ray Tracing Enhancements for NVIDIA Hardware

Microsoft has unveiled significant enhancements to its Agility SDK with the release of DirectX, introducing remarkable advancements in ray tracing capabilities through Shader Execution Reordering (SER) and Opacity Micromaps (OMM).

Transformative Ray Tracing Features Integrated into DirectX Agility SDK

The recent updates in Microsoft’s Agility SDK DirectX signify a leap forward in graphical performance, particularly in ray tracing. With new features like SER and OMM, developers can exploit these improvements to achieve remarkable visual fidelity and efficiency in their applications. Below, we detail the key features across both the preview and retail versions of the Agility SDK:

Key Features of Agility SDK 1.717 Preview

Cooperative Vectors: This feature introduces robust hardware acceleration for vector and matrix computations. It simplifies the integration of neural rendering techniques into real-time graphics pipelines, enhancing performance and efficiency.

Shader Execution Reordering (SER): This innovative element in DirectX ray tracing allows applications to optimize GPU thread execution. By reducing thread divergence and enhancing parallel processing, hardware supporting SER can achieve up to twice the performance in path tracing games. For an in-depth exploration of SER, see our comprehensive blog post or view the GDC DirectX State of the Union Recording.

Direct3D Video Encoding Enhancements: This update includes several new features:

HEVC Reference List extension for complex long-term reference scenarios.

Two-pass encoding, incorporating a lower resolution for the initial pass.

Inclusion of Peak Signal to Noise Ratio (PSNR) metrics within encoded frame output statistics.

Highlights of Agility SDK 1.616 Retail

Opacity Micromaps (OMMs): A powerful addition to DirectX Raytracing, OMMs leverage hardware-accelerated alpha testing to streamline ray tracing tasks. By significantly minimizing or even eliminating AnyHit shader calls, OMMs enhance efficiency without compromising visual quality. In path traced games, this could translate to performance improvements of up to 2.3 times, allowing hardware to better manage complex transparency. For more on OMMs, refer to our detailed blog post or check out the GDC DirectX State of the Union Recording.

D3D12 Tiled Resource Tier 4: This update introduces support for tiled texture arrays with a complete mip chain, overcoming previous constraints on packed mip resources. This enhancement allows for more efficient texture streaming and greater layout flexibility. Detailed specifications can be found at Tiled Resource Tier 4. Supported drivers include:

AMD: Expected support for Tiled Resource Tier 4 is set for early June 2025.

Intel: Intel has made its support for Tiled Resource Tier 4 available through the latest driver.

NVIDIA: Full support for this SDK release is confirmed, contact your developer relations representative for more details.

For further insights, please visit Microsoft Dev Blogs.

DirectX State of the Union: Raytracing and PIX Workflows (Presented by Microsoft) (https://www.youtube.com/watch?v=CR-5FhfF5kQ)

Focusing on D3D12 Opacity Micromaps (OMMs), Microsoft emphasizes that this feature allows hardware to efficiently process alpha-tested geometry, significantly reducing the reliance on costly AnyHit shader invocations.

Microsoft has previously highlighted performance gains of up to 2.3x in path-traced titles utilizing OMM. A striking demonstration by NVIDIA showcased over 60% performance improvement, with a reference scene achieving 55 FPS contrasted against 90 FPS when OMM was enabled. Currently, NVIDIA is the sole provider of driver support for OMM on RTX GPUs, with other manufacturers expected to follow suit in the near future.

Notably, the game “Alan Wake”by Remedy exemplifies the advantages of alpha-tested geometry, featuring scenes with upwards of 9.3 million triangles, 5.2 million skinned vertices, and 2.2 thousand skinned instances. This title is recognized for its visual intensity, employing path tracing with high settings that require generating 10 rays per pixel—totaling approximately 36.9 million rays per frame. Under these demanding conditions, an RTX 4090 without optimizations renders a scene in approximately 16.8 milliseconds, but with the enhancements from SER and OMM, this can drop to approximately 10.2 milliseconds.

The advancements presented through these next-generation technologies are promising. As developers embrace and implement these innovations into upcoming titles, gamers can expect to experience the cutting-edge enhancements facilitated by modern GPU hardware.

Source & Images