erm nVidia was only able to maintain the performance by scaling the die size to make up for the lack of die shrink, AMD took a more creative route and optimized, sacrificing silicone count to maintain original planned die size yet still came close to comparable nVidia products performance wise
this would explain the fluffed naming convention, had the 32nm happened, then chances are the 6xxxx would have comparable performance to their respective 5xxxx counter parts
personally i think this could well mean AMD has significantly more performance headroom when the die shrink does happen, there is only so much silicone real estate nVidia can add before it becomes unpractical, simply put, if AMD became desperate for the performance crown they can easily just fill out the difference in silicone real estate and probably trump nVidia, maybe AMD is leaving the ball in nVidia's court to see what their next move is