RTX 5080 Jumps From 1.4 to 310 FPS in New DirectX Ray Tracing Test
Microsoft’s new DirectX Raytracing technology just made an RTX 5080 look 221 times faster. In a demanding synthetic test, the card jumped from 1.4 FPS to 310 FPS once Clustered Geometry was switched on, while its acceleration structure memory use fell by 94.3 percent. The result comes from independent testing reported by Wccftech, and it also shows the RTX 5080 running about 3.4 times faster than AMD’s RX 9070 XT in the same sample. Before you start dreaming about 300 FPS path tracing in your favorite game, though, there is a catch. This was a synthetic sample, not a game, and the tester himself warns against reading too much into the headline number.
The test results in full
Graphics enthusiast Osvaldo Pinali Doederlein ran Microsoft’s D3D12RaytracingClusteredGeometry sample on an RTX 5080, comparing the traditional bottom-level acceleration structure (BLAS) rebuild against the new Clustered Geometry path. He paused the animation at the exact same frame for both runs, so the comparison is apples to apples. Here are the numbers Wccftech reported:
| Metric | Traditional BLAS rebuild | Clustered Geometry |
|---|---|---|
| Frame rate | 1.4 FPS | 310 FPS |
| Acceleration structure rebuild | 720 ms | 1.0 ms |
| Ray dispatch time | 1.8 ms | 2.0 ms |
| Acceleration structure memory | 4,716 MB | 267 MB |
That works out to roughly 221 times higher frame rates, a 720 times faster acceleration structure rebuild, and 94.3 percent lower memory consumption. The one number that barely moved is the ray dispatch time, which actually rose slightly from 1.8 ms to 2.0 ms. That detail matters, because it shows the huge gain came from cutting rebuild overhead, not from making the rays themselves travel faster.
RTX 5080 vs RX 9070 XT
Osvaldo ran the same sample on AMD’s Radeon RX 9070 XT earlier, which makes the head to head comparison unusually clean. With Clustered Geometry enabled, the RX 9070 XT reached 91 FPS, putting the RTX 5080 about 3.4 times ahead in this particular test.
- With Clustered Geometry on: RTX 5080 at 310 FPS vs RX 9070 XT at 91 FPS.
- With Clustered Geometry off: the RX 9070 XT actually led, at 1.7 FPS against the RTX 5080’s 1.4 FPS.
- Memory use: the RTX 5080 also consumed substantially less acceleration structure memory than the RX 9070 XT, both with the feature on and off.
- Why NVIDIA leads here: Osvaldo credits part of the advantage to NVIDIA’s more mature bounding volume hierarchy (BVH) compression technology, which dates back to its Turing architecture.
The flip between the two runs is the real story. When the feature is off, both cards choke on the same rebuild cost and the 9070 XT edges ahead. When it is on, NVIDIA’s compression advantage shows up and the gap explodes. The bottleneck was never raw ray tracing throughput. It was the cost of rebuilding acceleration structures, over and over, every frame.
What Clustered Geometry actually changes
Ray tracing works by building acceleration structures that tell the GPU where every piece of geometry sits in a scene. When detailed meshes move, deform, or change, those structures have to be rebuilt, and in a complex scene that rebuild can cost more than the ray tracing itself. Clustered Geometry attacks that exact problem. It groups triangles into reusable clusters, so the GPU can reuse most of the structure and only rebuild the parts that changed.
A companion feature, Partitioned Top-Level Acceleration Structures (PTLAS), does the same thing at the scene level. It lets developers update individual sections of a large environment instead of reconstructing the entire top-level structure every time something moves.
NVIDIA has published a technical whitepaper explaining how these DirectX features essentially standardize key parts of its existing RTX Mega Geometry technology. The tech is already out in the wild in at least one game: Remedy’s Control Resonant already uses Clustered Geometry and RTX Mega Geometry. NVIDIA has also demonstrated its RTX Mega Geometry foliage technology for the upcoming The Witcher 4, which should give future open world games a clear path to far denser ray traced vegetation.
A 60 million plant forest shows where this is heading
To show what the technology makes possible, NVIDIA built a DirectX 12 demonstration featuring a 25 square kilometer forest with 60 million individually animatable plants, rendered with real-time path tracing and dynamic vegetation. The demo uses 636 reusable tessellation patterns to manage terrain geometry, while partitions shift dynamically with the camera position and conditions such as wind.
The point of the demo is not the plant count. It is that a scene of that density would be unthinkable with traditional acceleration structure rebuilds, because every swaying plant would force the GPU to rebuild huge chunks of geometry data. With clustered, reusable structures, the cost stays manageable. That is the kind of headroom that could let future open world games pack far more detailed ray traced environments without a proportional performance collapse.
Keep the 221x number in perspective
Osvaldo himself cautioned against extrapolating these numbers to actual games, and that warning deserves to be front and center. The Microsoft sample has no level of detail (LOD) system, so unnecessarily complex geometry stays visible even in distant areas where a real game would swap in simpler models. Real world path traced games use far more sophisticated optimizations, which makes such extreme performance gains highly unlikely in practice.
It also helps to remember what the sample is designed to do. It is a stress test for one specific bottleneck: acceleration structure rebuilds in a scene packed with animated geometry. It proves how much overhead smarter geometry management can remove. It does not prove your games will run 221 times faster.
On the practical side, support is already here. According to Wccftech, GeForce Game Ready Driver 621.01 already supports Clustered Geometry and PTLAS across every GeForce RTX generation, with no dedicated preview driver needed. The DirectX 12 features themselves remain part of Microsoft’s Agility SDK 1.722 Preview, and AMD currently provides a separate developer preview driver for compatible Radeon GPUs. In other words, the hardware and driver side is ready. The wait now is on game developers adopting the new DirectX features.
What this means for PC gamers
- Denser ray traced worlds are coming. The first games to adopt Clustered Geometry should be able to pack far more detailed geometry into ray traced scenes without the same performance penalty. Open world games stand to gain the most.
- No new hardware is required. If you own any RTX graphics card, the current Game Ready driver already supports the feature. The bottleneck is game support, not your GPU.
- Watch the RTX 5080 line closely. The card keeps making headlines, and board partners are still waiting on design kits for the rumored 24GB model. Read our coverage of the RTX 5080 24GB release delay for the full picture.
- DLSS 5 is the other half of the story. Geometry management cuts the cost of building ray traced scenes, while upscaling cuts the cost of rendering them. See how the DLSS 5 October update stirred debate over image quality to understand both sides of that trade off.
The 221x headline is a synthetic number from a synthetic test, and it will not show up in your games. What will show up is quieter but more important: ray traced worlds that can finally afford the geometric detail developers have always wanted to build. Clustered Geometry removes one of the biggest hidden taxes on ray tracing, and the RTX 5080 test shows just how large that tax was.
Source: Wccftech
