And that's why I said it may disappoint. TDP doesn't decrease, and neither does power consumption.
What TDP measures is how much heat the chip puts out on average, or how much heat Intel "thinks" it puts out on average. In Penryn's case, even though power consumption dropped, heat was still about relatively the same, but in exchange, IPC improved, and also frequency scaling improved.
Like you said, Anand's article makes it seem like power consumption isn't significantly different compared to Sandy.
Where is less power drain?
Regarding the top 2, do take note that they are hexa-core (6-core) processors as opposed to quad. That's why the difference is so large compared to the 2600K. Compare the highlighted part to the 2600K and you'll see the quad vs quad result. In which case, it's not much difference.