DirectX Raytracing Update Hits 66x FPS Gain in Microsoft Test

Hakkı ALKAN

October 30, 2026

Microsoft has released a preview of the biggest DirectX Raytracing update since the technology debuted in 2018. Announced on October 28, 2026, the new Clustered Geometry and Partitioned TLAS features cut how much work a GPU spends rebuilding ray tracing data every frame, and in a Microsoft test scene on an AMD Radeon RX 9070 the frame rate jumped from 1.2 FPS to 79 FPS, roughly 66 times higher.

Microsoft is clear that this number is not a gaming benchmark. Its own developer blog warns in capital letters that games should not expect wins of this size and calls the scene “just a test playground.”

What the DirectX Raytracing Update Changes

Ray-traced games keep a structure that describes where every object in the scene sits, so the GPU can quickly work out what each ray hits. When characters animate, foliage moves or new areas stream in, large parts of that structure have to be rebuilt, which is expensive.

The preview, part of Agility SDK 1.722 preview, attacks that cost in two ways:

• Clustered Geometry splits models into small reusable clusters of up to 256 triangles and 256 vertices. Games can reassemble existing clusters for animation, streaming and level of detail instead of rebuilding everything from raw triangles.
• Partitioned TLAS (top-level acceleration structure) limits updates to the regions of a large world that actually changed, leaving the rest untouched.

According to Microsoft’s DirectX blog, cheaper builds can make it practical to ray trace the same full-detail geometry a game already renders, rather than simplified stand-in models, so ray-traced shadows and reflections match what players see on screen.

Where Does the 66x Figure Come From?

Microsoft compared a traditional rebuild with Clustered Geometry in the same sample scene on an RX 9070:

• Frame rate: 1.2 FPS vs. 79 FPS (about 66x higher)
• GPU acceleration structure build time per frame: 788 ms vs. 3.11 ms (about 254x less)
• Ray tracing time per frame: 72.6 ms vs. 9.27 ms (about 7.8x less)
• Total acceleration structure memory: 7,519 MiB vs. 774 MiB (89.7% less)

The trade-off is a much larger temporary scratch buffer during builds. Microsoft adds that real games may see far smaller gains, but that build time savings can still outweigh slightly slower ray tracing in many cases.

Which GPUs Support It?

No new hardware is required beyond existing ray tracing support, but each vendor needs to ship drivers. Preview drivers are available now for:

• NVIDIA: all GeForce RTX GPUs
• AMD: Radeon RX 9000 series
• Intel: Arc B-Series graphics and Core Ultra 200V and Series 3 processors
• Qualcomm: Snapdragon X2 support targeted for the first half of 2027

Support for some older ray tracing GPUs is still undecided, Microsoft says, because driver work may reveal performance problems.

What It Means for Gamers

The idea is not entirely new. NVIDIA already offers it on RTX cards under the name RTX Mega Geometry, and Gears of War: E-Day is an early adopter. Bringing it into DirectX gives developers one standard path that works across NVIDIA, AMD and Intel hardware, which should make it far more common in future PC games. NVIDIA has been pushing ray and path tracing on several fronts lately, from RTX Remix 1.6 with Minecraft path tracing to the RTX Spark platform with Microsoft, while AMD keeps widening upscaling support with FSR 4.1.

For now, this is a developer preview that requires Windows Developer Mode, Shader Model 6.10 and preview drivers. Players will only benefit once games adopt it, and Microsoft says developer feedback will decide how quickly it moves to a final release.