A GPU-powered computer cluster created by students at NCSA.

Students Build Leading ‘Green’ Supercomputer


A GPU-powered computer cluster created by students at NCSA, which placed third in the Green 500 competition for energy-efficient supercomputing.

A team of students from the National Center for Supercomputing Applications (NCSA) have earned third place in the Green 500 list for energy efficient supercomputing with a 33 Teraflop system that was competition’s greenest self-built cluster. The NCSA entry finished behind a prototype system from IBM and Japan’s top supercomputer.

With guidance from Computer Engineering Professor Wen-mei Hwu, Computer Science professor Bill Gropp, and cluster experts at NCSA, students got hands-on experience building their own supercomputer with NVIDIA C2050 graphics-processing units. NVIDIA donated 128 C2050 units to the Illinois CUDA Center of Excellence, led by Hwu, and NVIDIA research scientist Sean Treichler spent time on campus helping to plan and build the cluster, which landed spot 403 on the Top500 list and then took third place when the Green500 was posted

“This project was very unique in that it gave students access to something few people ever have the opportunity to do,”said NCSA’s Mike Showerman. “We are already running some physics codes on the system with good performance results.”

This video shows the NCSA students assembling the system. It runs just under 2 minutes.

For more coverage of information about supercomputing, check out our High Performance Computing Channel. For additional video, check out our DCK video archive and the Data Center Videos channel on YouTube.

About the Author

Rich Miller is the founder and editor at large of Data Center Knowledge, and has been reporting on the data center sector since 2000. He has tracked the growing impact of high-density computing on the power and cooling of data centers, and the resulting push for improved energy efficiency in these facilities.

Add Your Comments

  • (will not be published)


  1. Back calculating from numbers in their press release, this supercomputer consumes about 36 KW, with equipment mounted on bread racks. Note the double rows of perforated floor tiles in front, and no visible efforts to separate hot air from cold air. Data center designs like this often have PUEs over 2, approaching 3, so this supercomputer might need another 36-72 KW for cooling. Maybe the Green 500 should multiply IT load by PUE before calculating the best performance per watt.

  2. Knightlink

    From my understanding of green computing is to use less energy but produce the same results, from what I can see here is they are basically the same build as a rack mount but without a housing. Still have the 128 power supplies creating excess heat that is not needed in this build. Looks interesting...

  3. Rocky, thanks for the comment! As for the PUE, the current projection for the facility is 1.1 or 1.2 -- you can read more on our website: http://www.ncsa.illinois.edu/AboutUs/Facilities/npcf.html