Whether it's average or max here depends on what fraction of time you're taking into account and how often are you sampling and of course how precise your measurement equipment is in this matter.
If your equipment is slower than the time between change of max power spikes and low power cycles than it'll average the measured values.
To get accurate measurements on power draw on 50Hz power line you'd need 100Hz sampling rate. Meanwhile most of the resources you listed show graphs with range of 1 minute. Of course their equipment may be precise enough but we don't know that.
Not sure if you noticed this graph at Tom's:
and that their "Graphics Card Overall" maximum result below is 437W while the average is 186W, and this graph:
where their "Graphics Card Overall" maximum result below is 332W while the average is 182W.
If your equipment is slower than the time between change of max power spikes and low power cycles than it'll average the measured values.
To get accurate measurements on power draw on 50Hz power line you'd need 100Hz sampling rate. Meanwhile most of the resources you listed show graphs with range of 1 minute. Of course their equipment may be precise enough but we don't know that.
Not sure if you noticed this graph at Tom's:
and that their "Graphics Card Overall" maximum result below is 437W while the average is 186W, and this graph:
where their "Graphics Card Overall" maximum result below is 332W while the average is 182W.