What we know so far: Many ray tracing games still reflect light off cheaper, low-detail copies of their scenery rather than the models players actually see on screen. This is mainly because it simply takes too long to recover the GPU’s ray tracing data for every swaying tree and moving character in every frame. A new DirectX feature currently in preview aims to address this issue. It could also free up a large portion of video memory currently consumed by ray tracing.
Microsoft detailed Clustered Geometry and a companion feature called Partitioned TLAS on its DirectX developer blog this week, describing the pair as the biggest change to DirectX ray tracing since the API’s introduction in 2018.
Right now, every time a mesh deforms or flows in with a new level of detail, the GPU has to rebuild that object’s acceleration structure from scratch. Clustered Geometry instead splits models into small pieces of up to 256 triangles, which are created once and reused. So when an object changes, developers can reassemble existing clusters or recreate just a handful of them.
Partitioned TLAS applies similar thinking to the entire scene. It divides the world into regions so that in each frame only the parts where something has actually moved are updated.
Microsoft’s own test pattern makes the profits appear enormous. On a Radeon RX 9070, switching from traditional rebuilds to cluster geometry increased the frame rate from 1.2 fps to 79 fps. The creation time per frame dropped from 788 ms to just over 3 ms, and the memory for these structures shrank from about 7.5 GB to 774 MB, about 90% less. However, this is only ray tracing data, not the entire VRAM footprint of a game.
Still, Microsoft issued a headline warning that wins near this magnitude were not expected in games. Cluster builds also require far more temporary memory, so developers need to take this into account.
Nvidia has been delivering the same idea as RTX Mega Geometry since last year. In Alan Wake 2, both frame rates and image quality have been improved. Earlier this month, Gears of War: E-Day used it to ray trace Unreal Engine 5’s detailed nanite geometry, instead of the static low-poly proxies the engine normally relies on.
The difference now is that it’s part of DirectX, giving developers a path that works across vendors. No new hardware is required either: Nvidia supports it on every RTX card, and Intel covers its Arc B-series GPUs and current Core Ultra chips. On the AMD side, however, only the Radeon RX 9000 series has so far made it onto the list.
The waiting time will be a little longer for players. Everything is still in preview and needs to be enabled to turn on Windows Developer Mode, so games based on it are still a ways off. And according to Gears, even Nvidia’s version requires a GeForce card with at least 12GB of VRAM before it can be activated.
