There's been various driver improvements since release for both cards, rendering most benchmarks wrong anyway, so evidence is pretty meaningless, when for example, the latest AMD drivers claim to offer 20% extra performance in certain games than the drivers at release did.
Also, I believe passmark probably looks more at 3D modelling and stuff rather than actual gaming, where nvidia may benefit from cuda cores. In gaming, results are different.
Also, your test puts a 670 above a 690? And a 480 above a 590, so I'm guessing all tests were done immediately after release where drivers were pretty terrible. things will have changed a lot since then.