GPU instancing
Drawing many copies of the same mesh in a single call by handing the GPU a small table of per-instance data instead of asking for each copy separately.
The bottleneck in real-time graphics is rarely the GPU's raw horsepower but the number of draw calls — each one costs CPU time for state changes and chatter with the driver. Instancing sidesteps this by putting whatever differs between copies, typically a transform and a colour, into a separate instance buffer. The GPU walks that buffer and re-draws the mesh over and over, with no CPU involvement between iterations.
APIs grew direct support with OpenGL's glDrawArraysInstanced and Direct3D's DrawInstanced, and it is the standard answer to grass, foliage, crowds and particles — anything that is one shape repeated a thousandfold. Where LOD simplifies each object, instancing attacks the cost per object.