Do Raytracing cores speed up volume rendering significantly compared to fragment shaders?

I’m currently doing volume rendering using fragment shaders (similar to what is shown in Volume Rendering with WebGL).

What speedup could I expect when using Raytracing cores via Optix?

Hi @abiramlon,

OptiX and RTX hardware is built to accelerate surface rendering, e.g., meshes, curves, & other surface-based geometry, as well as accelerate common use-cases with surface geometry such as instancing and motion blur.

As volume rendering typically does none of that, there isn’t currently a strong reason to try to use OptiX or RTX hardware for volume rendering, and one shouldn’t expect any speedup. That said, one reason to use OptiX for volume rendering is when integrating volume rendering and surface rendering together. In that case you may wish to integrate volume rendering into an OptiX surface renderer for ease of development or to avoid memory transfers between two different kinds of render kernels. There is an OptiX SDK sample that demonstrates this, called optixVolumeViewer, which uses nanovdb for traversing the volume data.