NVIDIA Micro-Mesh

NVIDIA Micro-Mesh is a graphics primitive for extreme geometry and real-time ray tracing. Accelerated by NVIDIA® GeForce RTX® 40 Series GPUs, this technology efficiently stores opacity and displacement and allows assets to be used in their full fidelity, directly rasterized or ray traced. Micro-Mesh increases real-time performance and memory compression for complex geometry such as scanned artifacts, creatures, rocks, and trees.

Learn More

NVIDIA Micromesh can store opacity or displacement for complex geometry such as fossils, creatures and nature.

Sample micro-mesh composed of 16K individual micro-meshes (left half), expanding to 2 million micro-triangles (right half), consuming ~1 byte per micro-triangle; from: threedscans.com

Explore the Benefits of NVIDIA Micro-Mesh

More Efficient Pipeline for Complex Geometry

Micro-meshes more efficiently encode the opacity and displacement of complex materials, allowing you to fill a scene full of photorealistic geometry with increased frame rate and smaller memory footprint.

Superior Ray-Tracing Performance

Built to ray trace from the ground up and handle objects with extremely high geometric detail, micro-meshes provide headroom for more photorealistic rendering.

Open Source and Cross Platform

Micro-meshes are available to all developers, may be used across platforms, API’s and independent hardware vendors (IHVs), and are hardware accelerated by GeForce RTX 40 Series GPUs.

Take a Deeper Dive into Micro-Mesh Technologies

Get started with two SDKs below: Displaced Micro-Mesh provides superior ray tracing performance and memory compression while Opacity Micro-Map provides increased frame rates in ray traced scenes.

You can use them separately or together.

Displaced Micro-Mesh SDK

With Displaced Micro-Mesh, you can build highly detailed and complex geometries that are very compact and efficient to render. Built on a structured graphics primitive using micro-triangles, assets can be used in their full fidelity, designed to be directly rasterized or ray traced in real time without conversion or expansion. Currently, standard graphics primitives are not designed for highly detailed organic surfaces, characters, or objects. This is the only technology built from the ground up for real-time ray tracing with up to a 50X increase in geometry.

Access GitHub

Build detailed and complex geometry quickly with Displaced Micro-mesh SDK

Sample micro-mesh composed of (first image) 175K individual micro-meshes, expanding to (second image) 57 million micro-triangles, consuming ~4 bits per micro-triangle; from Smithsonian 3D

Map detailed textures to complex objects with Opacity Micro-Map SDK

Ada uses Opacity Micro-Maps to increase performance by avoiding any hit shader invocation (running them only in yellow areas)

Opacity Micro-Map SDK

With Opacity Micro-Map, you can more efficiently map intricate geometries onto triangles and micro-meshes. This technology encodes the opacity of micro-triangles and makes it possible to ray trace, at high performance, extremely detailed scene elements such as vegetation and foliage. A traditional alpha texture requires a more costly any-hit shader to produce the same result. Opacity Micro-Map was designed to support real-time ray tracing of extremely detailed elements, like ferns or door screens, with little cost, creating headroom for dramatically increased richness and realism.

Get Started

"We are very excited to partner with NVIDIA to enable game creators with the Simplygon SDK, to compress super-detailed objects by an order of magnitude unmatched by other solutions. With NVIDIA Displaced Micro-Mesh technology, developers can pursue crafting environments at unprecedented levels of fidelity, density, and variety – immersing players in new stunning and beautiful worlds in games."

— Magnus Isaksson, Studio Head for Simplygon at Microsoft

"We at Adobe are excited about NVIDIA’s Displaced Micro-Mesh technology with native ray tracing support, which has the potential to unlock ultra-detailed real-time ray traced scenes with minimal memory cost.”