The numbers I'm seeing in the tomshardware article doesn't really support that
Could be that when you are testing full system's load at idle, most of the power is actually used on secondary components and CPU, not the GPU. Would make sense as most of the today's graphics cards have power saving features which kick in when you idle, thus consuming very little power compared to the rest of the system. That'd explain the small gap even if you swap cards.
We are venturing pretty far in the land of offtopic, though, aren't we