TECH NEWS – Intensive rendering workloads now perform better on Intel Arc B graphics chips.
Microsoft‘s latest experiment to optimize rendering workloads has delivered impressive performance gains, with Intel Battlemage GPUs showing a significant uplift. As rendering pipelines grow more complex, traditional approaches can create bottlenecks under these workloads, and Microsoft‘s SER solution targets a key limitation. SER stands for Shader Execution Reordering and is a Shader Model 6.9 feature, meaning SER will become standardized with the next driver code release.
Microsoft tested SER in a sample and observed a 90% frame rate increase with Intel B-series GPUs, and a 40% increase with an Nvidia GeForce RTX 4090. Interestingly, SER is not entirely new, as Nvidia has already integrated it with path tracing optimizations, producing significant gains in rendering workloads. However, Microsoft‘s tests do not fully reflect real-world gaming workloads, so the difference may be smaller.
SER improves ray tracing efficiency, especially when a ray hits multiple objects that each require different shaders. Typically, when different shaders must be processed, every thread in a warp has to wait for the others, resulting in substantial idle time. With SER, objects hit by the ray are stored and then reordered based on spatial locality and shader similarity, enabling much more coherent execution. According to Microsoft, combining HitObject and SER is particularly effective, allowing execution and data coherence to be rearranged using HitObject information and additional user-provided hints, further improving hit/miss processing coherence.
SER is integrated with Microsoft Shader Model 6.9, which requires AgilitySDK 1.619. While there are currently no known hardware limitations, based on Microsoft‘s presentation it is likely that Ada Lovelace and newer Nvidia graphics cards (GeForce RTX 4000+) and Intel Battlemage GPUs will support SER.
