No. Linux drivers are supposed to implement runtime_pm callbacks. These are called based on the usage counter of those devices. ACPI is an Intel/microsoft standard for power state definition with spotty compliance. How and when you put those devices into whatever power state, ACPI or not is dictated by those callbacks. Bugs seem to most often occur when you turn off something when it is still being used cuz someone failed to claim it or released it too soon. Also asic bugs in their definitions of their power states.