Microsoft has rolled out its biggest DirectX Raytracing update since the technology launched back in 2018, adding two features called clustered geometry and partitioned top-level acceleration structure, or PTLAS. Together they promise smoother ray-traced frame rates and VRAM savings of up to 90% in supported scenes, a change that could matter most in sprawling open-world games where memory and compute overhead have long been a bottleneck.
Key takeaways
- ✓Microsoft's DirectX Raytracing update is its biggest change to DXR since the feature launched in 2018.
- ✓Clustered geometry lets ray tracing reuse small triangle clusters instead of rebuilding large acceleration structures.
- ✓Partitioned TLAS (PTLAS) updates only changed regions of a scene, easing VRAM and compute load.
- ✓Microsoft and NVIDIA report ray tracing VRAM savings of up to 90% in supported scenes.
- ✓The two features can be used together or independently, with open-world games the biggest likely beneficiary.
Ray tracing has always carried a heavy cost because of how scenes are built behind the scenes. Every object that moves or deforms needs its own acceleration structure, a kind of index that tells the GPU where geometry sits in 3D space so it can trace light rays accurately. Rebuilding these structures constantly, especially in busy or animated scenes, eats into both VRAM and processing time. This DirectX Raytracing update tackles that problem directly, rather than relying on incremental tweaks to existing methods.
Table of Contents
What clustered geometry changes in DXR
Clustered geometry breaks animated and procedurally generated objects into small, reusable groups of triangles instead of treating each object as one large, constantly rebuilt structure. Microsoft’s approach means developers no longer need to fall back on simplified proxy models just to keep ray tracing affordable. Instead, the same detailed geometry used for standard rasterised rendering can now be fed straight into the ray tracing pipeline.
That matters for visual accuracy as much as performance. When ray-traced shadows and reflections are built from a simplified stand-in rather than the real, currently visible model, mismatches can creep in, such as a reflection that looks subtly different from the object casting it. By reusing the full-detail geometry, Microsoft says shadows and reflections should line up with exactly what is on screen, without developers needing to maintain two separate versions of every asset.
Why rebuilding acceleration structures was so costly
Large bottom-level acceleration structures are expensive to regenerate every frame, particularly in scenes full of moving characters, destructible objects, or dense foliage. Clustered geometry sidesteps much of that cost because small triangle clusters can be reused and recombined rather than torn down and rebuilt from scratch each time something on screen changes.
PTLAS and the open-world VRAM problem
The second piece of this update, PTLAS, addresses a different pain point. A standard top-level acceleration structure (TLAS) tracks every object’s position across an entire scene, which becomes punishing in large open-world titles where the playable area can stretch for kilometres. Keeping all of that data resident and current demands significant VRAM and GPU cycles, even for areas the player cannot currently see.
Partitioned TLAS lets the GPU update only the regions of a map that have actually changed, while leaving acceleration-structure data for the rest of the world untouched. In practice, that means a game can refresh the ray tracing data for a small neighbourhood the player just entered without re-processing an entire city or continent. Microsoft notes that clustered geometry and PTLAS are not dependent on each other. Developers can adopt either feature on its own or combine both for maximum benefit, depending on what a given game needs most.
| Feature | Problem it solves | Main benefit |
|---|---|---|
| Clustered geometry | Costly rebuilding of acceleration structures for animated or generated objects | Full-detail geometry in ray tracing, better shadow and reflection accuracy |
| Partitioned TLAS (PTLAS) | Heavy VRAM and compute demand from whole-scene acceleration structures | Updates only changed regions, cutting VRAM use by up to 90% |
What this means for gamers and developers
For players, the payoff should show up as higher and more consistent frame rates in ray-traced games, particularly on GPUs with limited VRAM that currently struggle once ray tracing is switched on in demanding titles. Lower memory pressure also means ray tracing could become viable on a wider range of hardware, rather than being reserved for high-end graphics cards.
Developers get the more interesting trade-off. Clustered geometry and PTLAS remove some of the long-standing excuses for skipping ray tracing in large, dynamic worlds, since the heaviest technical penalties are now considerably smaller. That said, adoption still depends on game engines and individual studios building support for these DirectX Raytracing features into their pipelines, so the benefits will roll out gradually rather than overnight. Microsoft’s push to squeeze more performance from ray tracing hardware follows NVIDIA’s own neural rendering work with DLSS 5, suggesting both companies are converging on smarter, less brute-force approaches to real-time lighting.
Expect the first practical benchmarks to come from engine vendors and early adopter studios rather than finished games, since this kind of low-level API update typically needs months of integration work before it shows up in a shipping title. Players chasing every extra frame in the meantime may still find value in automated tools designed to claw back extra frames, or general tips for reducing lag and improving frame rates while waiting for ray tracing-heavy titles to catch up with Microsoft’s latest DXR capabilities.
Frequently asked questions
What is clustered geometry in DirectX Raytracing?
Clustered geometry breaks a scene’s animated or generated objects into small, reusable groups of triangles, allowing ray tracing to use the same detailed models as standard rendering instead of simplified proxy versions.
What does PTLAS stand for and what does it do?
PTLAS stands for partitioned top-level acceleration structure. It lets a GPU update only the parts of a scene that have changed, rather than rebuilding acceleration data for an entire open-world map every frame.
How much VRAM can this DirectX Raytracing update save?
Microsoft and NVIDIA report VRAM savings of up to 90% in scenes that take advantage of both clustered geometry and PTLAS, though actual results will vary by game and hardware.
When will games use these new DXR features?
There is no fixed timeline. Studios need to integrate clustered geometry and PTLAS into their engines and rendering pipelines, so visible benefits in shipping games are likely to appear gradually over coming months.
