As someone who spent a lot of their childhood talking to trees, I’m of the opinion that videogame forestry doesn’t need to look any better than Alan Wake 2’s Pacific Northwest woodlands. That hasn’t stopped AMD from innovating though; by leveraging ‘tetrahedral cages’ to help render a vast scene of greenery, researchers at the company say they’ve managed to cut the potential ray tracing memory usage by 98%.
Typically, a scene of around 25,000 independently animated plants made of hundreds of thousands of triangles could hoover up 80 GB of GPU memory. Not only that, but it would take a Radeon RX 9070 XT GPU 300 ms per frame to check which surfaces within a scene are intersecting with a ray (the data structure that organises this information is called a Bounding Volume Hierarchy). In short, using traditional tech to render such a densely populated scene would be slow and choppy.
Holger Gruen, a Fellow of AMD’s Advanced Graphics Program, recently shared on GPUOpen that the team’s tetrahedral cage demos only use “1.7 GB of BVH memory and approximately 3.3 ms for all BVH updates per frame” in comparison. Touch grass? Not when we’ve got trees like these.
Latest Videos FromPC Gamer
Anyway, let me break down how this tech works: a tetrahedral cage essentially wraps around the dense mesh of all that plant matter. Rather than many polygonal leaves bobbing in the simulated breeze, only the simplified cage deforms instead. So, rather than looking straight at the many wibbly-wobbly vertices of a virtual tree, rays are instead traced along the simplified surface structure of the tetrahedral cage around it.
At the risk of really oversimplifying, I think of this tech a bit like those 3D models or laser-engravings you sometimes see inside cubes of crystal—but maybe that’s just me.
“The dense geometry of an animated object and its BVH structures remain static and reusable. During runtime, only the cage deforms. Rays are transformed back into a static reference frame before intersecting them with the dense set of triangles,” Gruen explains. “This means animation cost only scales with cage complexity, not with triangle count. This comes at the expense of losing animation control over every vertex, but this is fine for many animations.”
In other words, for a scene where the focus is less on the luscious fronds of a single plant and more on a field of foliage, devs likely won’t miss the ability to finetune individual animations. But for 3D trees with personality, tetrahedral cages won’t be as useful.
It’s also important to note that the tech is still a research demonstration, rather than a fully fledged feature coming to a game near you soon. That said, the team is already “hard at work” on making “DXR samples and a header-only C++ library that allows building high-quality tetrahedral cages for skinned objects, keyframe-animated objects, and static objects.” You can also read more about the team’s research in the open access paper ‘Ray Tracing Massive Amounts of Animated Geometry’.
Best gaming rigs 2026
All our favorite gear