What impressed you about the supercomputers of 5 years ago that no longer impresses you about today's supercomputers? The Cray approach (http://www.nccs.gov/wp-content/uploads/2011/03/UserMeeting20...) and the IBM approach (http://en.wikipedia.org/wiki/Blue_Gene#Blue_Gene.2FQ) both seem more interesting than just connecting together as many CPUs/RAM as you can afford.

There are interesting OS challenges for the compute nodes, interesting hardware challenges on interconnect, interesting filesystem challenges, interesting programming challenges to manage the parallelism, competing architectural approaches between CPUs and GPU-accelerated clusters ... it's a pretty interesting space!