The figures basically show peck peformance as
600x16=9600 mpixels/s
430x24 = 10320 mpixels/s
Without any optimising, although the above isnt strictly true, the 7800 only has 16ROPS so it can only output a maximum of 16pixels per clock cycle, this means its the same as the faster clocked R520 so in terms of plain pixels the R520 will be quicker.
Although went it comes to advanced shaders etc you dont kickout 16pixels a clock anyway so the Nvidia with the longer pipeline will have an advantage.
Also theres rumours that Nvidia achieve 30% more performance(in some applications) by decreasing the AA sample rate and altering the the sample grid. Although that was only tested my modifing drivers, which may just mean that some of Nvidias "speical" 7800 optimising was ignored.