Tom Clabault

I'd like to share @DaqiLin and I co-authored a chapter in the latest GPU Zen 4 book. We cover the engineering details of implementing ReSTIR GI & PT in UE5's NvRTX branch, adopted in NVIDIA's Zorah project showcased at CES and GDC last year. The book is now available on Amazon.

Thanks for a great 2025! Here's a quick look back through some of my favorite renders I made this year

We just dropped a full 11 part course on Neural Shading 👀 If you’re looking to integrate the latest in real-time graphics tech into your projects, this is for you. Watch here 👉 nvda.ws/4p8ZyhI

Did a quick and dirty implementation of a spatial hash structure to speedup RTAO, ray results are stored in cells indexed by pos/normal/cell size and after storing a few rays occlusion can be queried from the cell instead of raytracing it. 3x faster RTAO with no denoising.

I had previously used triangle meshes to render giant planetary atmospheres but after many unsuccessful attempts to fix all the artifacts, especially when intersecting atmosphere geometry, I decided to add support for spheres to my renderer. It works much better

New personal project! Jungle environment and highly detailled breakdown - 3DS Max/Houdini/Vray/Nuke A 40 page breakdown awaits you : artstation.com/artwork/K3rNbB I had a ton of fun making this, exploring so many workflows and sharing so many tips! #ArtStationHQ

Curious to know how many cycles each CPU instruction costs? The instruction matrix from Google's highway (a cross-platform SIMD library) has interesting details. github.com/google/highway…

If early samples are possibly more noisy than later samples, do we need some more clever way of averaging rendered frames than just an average? Something that weighs all frames based on variance and takes those weights in consideration during averaging?

(At the end of section 3.2.2) The effect of the online learned visibility estimate system is that renders with complex occlusion initially converge slowly, but as the learning system’s quality improves during the renderer’s initial iterations, the convergence rate improves [...].

So I was reading the "Cache Points for Production-Scale Occlusion-Aware Many-Lights Sampling and Volumetric Scattering" paper from @yiningkarlli et al. and came up with a question about visibility learning (quote from the paper):

@Auzaiffe @VrKomarov @Auzaiffe Hey there! Any way to convert the VDB to the grid format taken by the grid to LEB3D converter? I downloaded the wdas cloud but it's VDB format 🤔

@Auzaiffe Impressive results! @VrKomarov You may be interested : )

The source code for the Longest edge bisection, tetrahedral based, volume representation for faster path tracing project is online available on GitHub and is under MIT license. Feel free to grab it and play with it! github.com/AnisB/leb3D

Worth watching presentation, also provides some background info on how ReSTIR came to be and where it is heading: "ReSTIR: Traveling the Path of Reuse" youtube.com/watch?v=MRg91Z…

Implemented BSDF evaluation and PDF query for microfacet-based normal mapping to support more light transport algorithms. For dielectrics, I need to make some adjustments to the equations in the original paper. Now I need to find a beautiful scene and test it.😉

@yiningkarlli That's cool stuff! Is your renderer available to the public online?

Installed macOS Tahoe in a VM and screwed around with getting my work-in-progress next-gen hobby renderer building with *･ﾟ✧ Liquid Glass ✧･ﾟ*. Most of my UI is dear imgui, but the window chrome and stuff is native AppKit. Wow Liquid Glass windows are really rounded off.

Neural Visibility Cache for Real-Time Light Sampling boksajak.github.io/files/NVC_pape…

Tried DLSS RR in my path tracer and it's pure magic. I think it's quite expensive for games (~7ms for this image on laptop 3080), but for previz in path tracing it's really good, very resposive and high quality

BVH Quality: Beyond SBVH – Jacco’s Blog jacco.ompf2.com/2025/05/20/bvh…

__builtin_amdgcn_image_bvh8_intersect_ray in gfx12-insts github.com/llvm/llvm-proj…