What's the overhead of having the Intel(R) VTune(TM) Amplifier XE for Linux* modules loaded but un-used and no collection is happening?
If there is no collection in progress, there is no execution time overhead of having the driver loaded. There would also be very little overhead wrt memory.
You can let the module be loaded at boot time (e.g., via the install-boot-script script, which is the default) and not worry about it. Unless data is being collected by VTune Amplifier XE, there will be no latency impact on system performance.