A gtx670 overclocked to 1250 (the average any competent OC'er can hit) vs an overclocked 7950 at 1175 (which is on the high-end here) both cards are about equal in gaming benchmarks, with a clear advantage given to the 670 in a lot of the newer games with heavy tessellation and physics.
As i said in that thread, this comparison is skewed in favor of the 7950. They're comparing a non-reference 7950 that has improved cooling to a reference 670. That alone accounts for the small difference as non-reference models are usually about 4-5% better than their reference counterparts at overclocking. But then you also need to take note that the 7950 is at very good OC for the card compared to the 670 which is poorly overclocked to only 1200mhz which is about 50hmz less than average and 150mhz less than the high-end. They also didn't overclock the 670's memory very well, only to 1555mhz when the 670 can easily hit 1800+mhz on the memory too. Then you also need to notice they switched motherboards between reviews which isn't a huge deal, but it could account for as much as 1-3% difference. So considering all that, and the fact the 670 in that review beat the 7950 in 4 out of the 7 games and only lost by an average of 10% in the other 3 (and 2 of those 3 were crysis and crysis 2, so really it's only 2 additional games), and that the 670 in that review is underclocked by about 10% below average, it's safe to say that a FULLY overclocked 670 will beat a FULLY overclocked 7950 by a fairly large margin. Then you also have all of the other benefits nvidia currently has over AMD (for example, the 670 uses 52 less watts when overclocked, it's quieter, it has better driver/software support, less micro-stuttering, and better SLI/CF implementation), it's pretty obvious to me which is the winner here.Edited by SeanPoe - 5/13/12 at 4:09pm