At the risk of getting a lot of flames: I've been in this category a few times, but never consistently over many months or years.
I.e. getting a challenge at work: "Mobil Oil left the meeting when we said you guys could develop this (safety) system in 3 months, with just one month to the first deployable version. They had calculated that it would take at least a calender year independent of the number of developers!"
My coworker and I hid away in a meeting room for three days, at which point we had written the entire first version, including a separate machine with a full sw simulation of all the missing hw parts, with programmable (Monte Carlo) error rates for all components and tracking of any resulting errors in the user output.
If I could do this day in and day out I would deserve that "rock star" title, but I know very well that I cannot.
Most of the time I'm quite happy working out interesting algorithms, shooting the breeze over at comp.arch or just spending my time figuring out why a given application/system doesn't work (or perform too badly).
I'm actually getting paid for that last part, so that is good.
Besides, I also want time for my wife & kids, my hobbies (orienteering, xc skiing/snowboarding, windsurfing/kiting, rock climbing etc), so I limit my work hours to the regulation 40h/week.
OTOH I have known/met a few real "rock stars", John Carmack is way up on that list and so is Anders Heijlsberg (who I first met way back when here in Scandinavia when he was a young punk who had just sold Turbo Pascal to Borland). Mike Abrash isn't quite as bright as Carmack, but he is incredibly persistent as well as consistently good.
All three of these come across as really nice guys.
Terje