Microsoft has released DirectX Agility SDK version 1.619, bringing Shader Execution Reordering (SER) and Opacity Micromaps (OMMs) out of preview status and into a stable, production-ready release. The update, which also formally launches Shader Model 6.9, represents a significant step in standardising ray tracing efficiency tools across the PC gaming industry — technology that was previously confined to vendor-specific implementations.
The headline figure grabbing attention is a 90% framerate uplift recorded on Intel's Arc B-Series (Battlemage) GPUs during Microsoft's own internal demo testing. Tom's Hardware reports that an NVIDIA GeForce RTX 4090 showed a roughly 40% performance boost under the same demo conditions. These are striking numbers, though they require careful interpretation before anyone starts upgrading their hardware on the strength of them.
What SER Actually Does
Ray tracing is computationally brutal because light rays bounce off surfaces unpredictably, creating what engineers call divergent shader execution. When different rays require different processing paths, GPU threads end up waiting on each other, burning clock cycles on idle time rather than actual rendering. SER addresses this directly: when different shaders need to be processed, rather than forcing every thread in a group to wait for the others, the technology stores ray hit results and reorders them based on spatial location and shader similarity, allowing execution to proceed in a far more coherent manner.
Opacity Micromaps, the companion feature also moving out of preview, enable hardware to handle alpha-tested geometry far more efficiently than relying on costly AnyHit shader invocations. In practical terms, the GPU gains precise information about which parts of a scene need to be calculated as opaque and which can be skipped entirely. The two technologies are designed to work together: SER groups similar workloads, while OMMs remove unnecessary ones. A demonstration using the game Alan Wake 2, shown at GDC 2025, revealed that combining OMMs and SER reduced ray tracing costs by roughly one third.
Who Benefits, and Who Doesn't
Nvidia introduced hardware-level SER with its RTX 40-series GPUs back in 2022, giving that architecture a head start on the efficiency gains the technology provides. With the feature now standardised inside DXR 1.2 as part of Agility SDK 1.619, the playing field broadens — at least on paper. OMMs are supported only by RTX 40-series GPUs and above, while SER with active reordering is confirmed on NVIDIA RTX 40-series and newer chips and on Intel Arc B-Series GPUs, but AMD's RX 9000 series offers API compatibility without performing actual reordering in hardware.
Microsoft's own figures show that Intel Arc B-Series configurations each produced a 90% higher framerate versus running without SER, though these figures are specific to Microsoft's sample application and its configuration, not a shipped game benchmark. That distinction matters enormously. The performance improvements were recorded on a demo only, and actual gains in real games could be lower; further, existing ray-traced games cannot simply add SER, as a game's ray-traced rendering pipeline must be rebuilt or modified to take advantage of it.
What the Standardisation Means for Developers
The Agility SDK is Microsoft's mechanism for shipping newer DirectX 12 features without requiring a full Windows operating system update; games can bundle the newer runtime, and players need only a compatible GPU driver and hardware. That approach lowers the barrier for developers who want to adopt SER without waiting for broad OS adoption. SER is now a required feature of Shader Model 6.9, providing a standardised way for GPUs to sort disorganised ray-tracing workloads into coherent, parallel threads dynamically.
To use SER, developers must target Shader Model 6.9 and use Agility SDK 1.619 or newer; while driver support is mandatory under SM 6.9, actual hardware-level thread reordering depends on device implementation, and developers can query whether a GPU actively performs SER sorting or treats it as a no-op. That last point is significant: the standardisation guarantees that the API exists across hardware, but it cannot guarantee that every GPU will deliver hardware-accelerated gains.
Implementation will likely require months of work before SER is incorporated into an actual shipping game, meaning consumers should temper any expectation of immediate performance dividends. The technology is real, the efficiency gains in controlled conditions are genuine, and the direction of travel is clear. But the gap between a promising tech demo and a measurable improvement in a commercially released title is one the industry has crossed before at its own pace. How quickly game studios integrate SER into production pipelines will ultimately determine whether these benchmark numbers translate into something players actually notice.