On some devices the instances cannot fit into the vertex buffer for a draw call. I should be able to split up the instancedMeshes based on the maxvertexbuffer for device.