GPU-driven rendering
A rendering architecture where the GPU culls the scene and builds its own draw lists with compute passes and indirect calls, sparing the processor from touching one object at a time.
In the classic model the CPU walks the scene, culls against the frustum, sorts, and submits one draw call per object. At hundreds of thousands of objects the call count itself becomes the bottleneck. GPU-driven rendering moves the whole loop onto the GPU: a compute pass reads all the geometry, culls it against the frustum and Hi-Z, and writes a compacted list that feeds straight into an indirect draw.
Ubisoft (Assassin's Creed Unity) and Sebastian Aaltonen were early movers; the talk "GPU-Driven Rendering Pipelines" at SIGGRAPH 2015 became the reference. The approach assumes bindless resources and instancing, and is now standard in large open-world engines. The price is complexity — debugging on the GPU is anything but pleasant.