And how exactly the material quoted means they will apply an slider to decrease unit number?
Comparing unit quantity/draw call amount in a RTS to tesselation factor is just BS, tess is just eye candy which can be easily controlled with an slider, unit quantity is a core element in making a RTS and having less draw calls without compromising unit quantity just so Maxwell doesnt look so bad is just asinine.
Do people really think for a second before smashing the keyboard like this?
Units are made of triangles. Triangles form polygons. NVIDIA GPUs have a superior GTris/s rate. Many units favors NVIDIA.
Each units has its own light source, light sources require shaders, shaders are compute intensive. Compute intensive favors AMD.
The ground is made up of tessellated, geometry bound, triangle meshes. NVIDIA have superior tessellation.
Smoke effects are compute bound. They require a shader.
AotS uses NVIDIA and AMD optimized shaders.
The advantage AMD have is not due to any scrupulous bias on the part of Oxide. AMD focused their architecture on compute parallelism. When many compute effects permeate a scene, this results in a compute bound scenario but this in no way indicates a bias towards AMD. Look at the DX11 performance for AotS, NVIDIA dominate, because DX11 can't feed AMD GCNs CUs.
Move over to DX12 and we've got a different situation. AMD GCN can be properly fed. NVIDIA are still executing the same shaders optimized for their architecture.
Add Async Compute + Graphics and AMD GCN can execute compute and graphics work in parallel. This isn't a bias, it's simply part of the DX12 multi-engine specifications.
NVIDIA lacking this capability is sad but that's NVIDIA lacking a truly beneficial aspect of DX12.