The cost of translating input calls from Linux to Windows APIs through Wine/Proton is so much less than the cost of rendering a single frame that it's effectively 0. The only real overhead Wine/Proton might be adding is in the rendering pipeline.

As Steve explained in the video, the Steam Deck's input-to-photon latency is lower than the Aya Neo's simply because it has a much more powerful GPU.

The only real way to test if there's any extra input-to-photon latency would be to compare the exact same game on a Steam Deck running SteamOS to a Steam Deck running Windows.