NVIDIA has released a demo of Turing mesh shader technology
being used to create smart culling systems to manage the level of detail (LOD) and improve rendering efficiency. Turing brings a new compute programmable model to the graphics pipeline that is based on task and mesh shaders. This is accomplished in the asteroids demo by the GPU hardware checking the asteroids for visibility to determine the LOD(s) to use. The new graphics pipeline uses cooperative thread groups to create compact meshes (meshlets) on the chip using application-defined rules. These meshlets are tested by the mesh shader. The remaining triangles are culled by the GPU hardware. Older GPU technologies would have to cull each triangle individually which created huge workloads for the GPU and CPU. In the video it shows 3.5 trillion triangles being culled to only 50 million triangles actually drawn on screen. The Asteroids demo
can be found here.