The last top500 list was published during ISC 2006 held this year in Dresden, Germany.
I was sad not to find in it a great system I was lucky to install.
With "only" 1.916 TFlops, it was not one of the 500 fastest system in the world 
The 500th system had a measured performance of 2.026 TFlops ... I was kind of shocked to see that such a system ( a bunch of nodes connected via a gigE network ) had an efficiency of 35.19 %, too bad for me to call this HPC !
"my" system, with a peak performance of 2.288 TFlops has an efficicency of 83.74%. If you consider only this number, it would be
ranked at position 41 ! And if you compare only dual core opteron systems, it is the first in its category 
Of course, linpack is not the ultimate goal in HPC, that being said, I do not believe bunch of nodes should be called HPC neither.
