With all the 2048 shaders enabled the comparison is not fair.
If they want an apples-to-apples comparison they should test the same amount of everything, eg number of shaders, frequency, memory bandwidth, etc...
Seriously, this will be impossible to do anyway... or this will mean 1 AMD shader ( ALU ) is the same of an Nvidia shader ( Cuda Alu ), and both architecture are exactly the same ... At this point we can OC the 680 for match the Tflops of the 7970 too . ( lol for DP, the 680 will need to do 5ghz without turbo . )
Both cards are oc easy, and you can find retail cards using similar clock on each brand ... so it is interessant to see if you buy y Gigabyte 7970 who run at 1050mhz in a shop what it is worse against the 680 who have a turbo clock speed gling from 1058mhz to 1110mhz ( in the best case, as some do 1058, 1084 or 1110 in retail ) ... It is too interessant to see if both architecture are clocked at the exact same clock speed, what happend . or if a 7970 retail OC from a brand x or y is worse against a 680 from a brand x or y . They are so close ... so it is even more important... nobody say you need to clock the 7970 at 1300mhz for reach the performance of a 680 at stock speed .. we speak about card you can find in shop .. or if you have a reference 7970, just put 75mhz more and be happy.
What is really strange is, we dont see so much review comparing 680 retail OC vs 7970 retail OC or 680 stock review who have add 7970 retail cards OC ... Price wise this could be a good comparaison too.
For the good or the bad, i remember thoses last 2 years this was more the case.. specially if price was close . Now i see 670SC vs 670 stock ( interessant ofc ), 680 OC vs 680 stock vs 7970 stock ( who is in the chart ), but 7970Msi Lighthning vs 680 Gigabyte OC vs stock ... not so much ... I have not even seen a Asus DCII Top 7970, vs Asus DCII 680 vs Asus stock 680 and stock 7970 .. If you want to do the comparaison, you will need read 3 review minimum on the same site, and like thoses last years, read one review is allready not enough for draw any conclusion, you can multiply this by 2 -3 sites (as no one have the same results and you need do an average for draw a conclusion ... ( erf my headache start again )