They have 5 different systems. Most of them (maybe all, I do not remember) are running on 20 gigabit infiniband networks.
At least two of the systems have made it into the top 50-75 fastest supercomputer list.
The newest one is a "HP Linux cluster with 1,083 HP ProLiant BL280c G6 blade servers, each with two-socket, quad-core Intel Nehalem processors, with a QDR InfiniBand (IB) interconnect."
Here is a link that gives more info (though not much) about the hardware.
The director of admin was very informative and helpful. He was more then happy to answer questions about costs (since it is all tax money and thus public record somewhere). Most programs are written in C and Fortran. We talked about administrating the system, since I was looking at trying to do an internship there. (Just to say I worked on a supercomputer... that is just awesome.)
We walked through the production floor, and he showed off the cooling systems, and talked about the history.
If everyone is interested in a tour, we need to figure out tentative dates, and I will contact the guy and try and set it up.
Ultimately, the experience has driven me to start pulling my spare computers together, and stand up my own rudimentary cluster.