It's a bit strange to test a gaming PC with 8hours/day, 365days a year.
I think no person is capable of doing such a thing.
The situation is different with desktops that are used in work environments.
It is more feasible to test a gaming pc/graphics card @ 2-4 hours a day, 300 days a year. (I mean, you gotta take a break sometime, no?)
With this test one could say his card costs him 150$ per year over another card, while in real life it costs him eg: only $40 per year more because he only plays games 2 hours a day, 250 days a year.
Then there's less the urge to buy a newer graphic card because at the end of the year he'll lose this amount anyways in electric bills.
If one is really going to game 356x8 hours/year one better buys a more expensive card, since one is playing a lot.
to some, power requirement will determine which card to take.
I mean, if could choose between 2 AMD and 2x Nvidia cards,that would be powerful enough to run my games in full detail, and where cheap enough to buy, probably the power requirement would be the determining factor after technology advancement (eg: DX9 / 10 / 10.1 card),and extra's like HDMI/DVI/...
Ps: some numbers seem a bit off on the power chart of the AMD cards, where a card used more in 2D than 3D (4550,and 3650 seem a bit off here).