Microsoft: DirectX SER sample shows +40% on RTX 4090, +90% on Intel Arc B-Series
This update follows the Agility SDK 1.619 release that brought Shader Model 6.9 to the release runtime, which we covered separately.
© INTEL Microsoft has published a DirectX Developer Blog post confirming the official release of Shader Execution Reordering (SER) for DirectX Raytracing. Microsoft describes SER as an addition to DXR that lets shader code signal how hardware can find coherence across rays, so the GPU can sort work for better parallel execution. Microsoft also says SER support is a required feature in Shader Model 6.9 , so drivers must accept shader code that uses SER. Device-side acceleration still depends on hardware support.
Microsoft describes SER as a response to disparities in ray tracing workloads. According to the company, DXR applications often see divergent shader execution and data access because ray work is stochastic. SER adds HLSL primitives that enable application-controlled reordering for improved execution and data coherence.
© NVIDIA Microsoft highlights a minimal SER demo called D3D12RaytracingHelloShaderExecutionReordering in the DirectX-Graphics-Samples repository. The sample draws a full-screen quad, then uses artificial “light” and “heavy” shading work, with heavy work appearing in vertical bands, to create divergence and show how sorting can help. Microsoft says the shader exposes compile-time configuration options so developers can compare SER modes against a no-SER path.
Using SER with the settings below running on an NVIDIA RTX 4090 showed a 40% framerate increase versus not using SER, and a couple of configurations of Intel Arc B-Series GPUs each showed a 90% framerate increase.
Keep in mind, this is just a demo, not a real game. However, in their testing, Microsoft engineers found that using SER with the listed settings on an NVIDIA GeForce RTX 4090 produced a 40% higher frame rate versus not using SER. Microsoft also reports that “a couple of configurations” of Intel Arc B-Series GPUs each produced a 90% higher frame rate versus not using SER. These figures are specific to the Microsoft sample and its configuration, not a shipped game benchmark. Also, they do not explain if that means “Battlemage” or Panther Lake B-series iGPUs, so I suppose both?
Source: Microsoft
