Pretty sure the GPU doesn't understand DirectX, which is used for Windows games, so it must be run through a translation layer.

Wrapping graphics APIs adds effectively zero overhead if the featureset of the hardware and drivers are a close enough match.

Sure, you are translating DirectX to Vulkan and that work is done on the CPU. So it may need emulation. But the Vulkan instructions passed to the GPU are effectively native. The amount of work the GPU has to do to execute against those calls is the same regardless of CPU architecture.