On Windows, Thunderbolt power management is handled by the ICM. Because we are asserting "force power" and disabling the ICM, we lose any hope of responsible energy usage. According to LWN, the controller uses 2W on idle. Macs implement the power management for the Ridge through ACPI. Why not through the OSX driver like most other PCI devices? The answer is in the strange architecture of the Ridge Thunderbolt controllers. Essentially, two host controllers are exposed to the operating system: the Thunderbolt NHI controller and the USB XHCI controller. Both are controlled by their respective drivers and uses PCI power management (D-states) and can support PCI active link power management. Additionally, USB 2.0 legacy devices can be supported through a "companion" controller (usually the PCH XHCI controller) instead of directly by the Ridge. All this means that for effective power management beyond that offered through the PCI specifications (which leaves us with the 2W under idle), the three controllers must communicate with each other. However, the code can get rather messy so Apple opted to do the dirty work in ACPI instead.