Writeup doesn't make sense. The problem is supposedly a fundamental limit in Direct3D 11.2, which would be unable to handle the large nunber of draw calls. Yet the Nvidia 970 and 980 are claimed to offer great ('ideal') performance. Using D3D 11.2. Wut?
Maybe AMD should spend more time optimizing their D3D path rather than spending their limited development resources on the Mantle detour that benefits noone but their own hardware.