Microsoft has found a new way to squeeze more performance out of today’s increasingly demanding rendering pipelines, and early results look especially promising for Intel’s latest Battlemage “B-series” GPUs.
The technique is called Shader Execution Reordering, or SER, and it’s designed to tackle a growing bottleneck in modern graphics workloads—particularly ray tracing, where the GPU can end up wasting time as different threads wait on each other. SER is part of Shader Model 6.9, which means it’s positioned to become a more standard, widely supported feature as new driver code and tools roll out.
In Microsoft’s own sample test, SER delivered a massive jump in performance. Intel’s Battlemage GPUs reportedly saw frame rates rise by as much as 90%, while NVIDIA’s GeForce RTX 4090 posted gains around 40%. Those are attention-grabbing numbers, but it’s worth keeping expectations realistic: Microsoft’s sample test isn’t the same as a full real-world gaming workload, so performance improvements in actual games may end up smaller depending on the title and how it uses ray tracing.
So what exactly is SER doing? Ray tracing workloads often require a single ray to interact with multiple objects, and those objects may each need different shaders. In the traditional approach, when threads inside a warp diverge—because they’re running different shaders or taking different paths—some threads end up idle while others complete their work. That idle time can add up quickly, reducing overall GPU efficiency.
Shader Execution Reordering aims to reduce that inefficiency. Instead of processing divergent shader work immediately in a scattered way, SER groups and reorders work so that similar tasks run together more coherently. In plain terms: it tries to line up ray tracing work in a smarter order, based on factors like shader similarity and spatial locality, so the GPU spends less time waiting and more time rendering.
Microsoft also highlights how SER becomes even more effective when combined with HitObject, enabling improved scheduling and better data coherence for hit and miss processing. The idea is that better information about ray-object interactions helps the system reorganize work more effectively, leading to smoother, faster rendering.
For developers and users wondering how this arrives on PCs, SER is integrated into Shader Model 6.9 and requires AgilitySDK 1.619. Microsoft hasn’t detailed strict hardware limits yet, but based on the demo results, support is expected on newer high-end architectures—likely including NVIDIA Ada Lovelace and newer, along with Intel Battlemage GPUs.
The big takeaway is simple: as ray tracing and complex rendering effects continue to push GPUs harder, efficiency improvements like SER could play a major role in delivering higher frame rates without brute-force hardware scaling. If SER adoption expands across engines and games, it may become one of the more meaningful optimizations for next-generation real-time rendering.
