Today's administrators face difficult decisions when they allocate limited university funds. Areas of basic research compete for scarce dollars, and the cost of a typical high-performance computing (HPC) cluster rings up at around a million (US) dollars. A new NSF-funded study may be able to tip the balance of the argument.
The study from the Clemson School of Computing, US, shows researchers with local access to supercomputers are more productive than those without. Even though remote access is available to US scientists through XSEDE or INCITE portals, the authors found that closer proximity enhances technical efficiency and thus research output.
The authors excluded grid computing from the study, and defined HPC as any "specialized computer systems designed to solve challenging computational problems that cannot be solved using desktop or similar, commonly available machines." They considered the typical HPC to consist of hundreds or thousands of central processing units.
To arrive at their conclusion, researchers divided a US National Research Council sample of 212 universities into two camps - those with and those without HPC. They weighed broad input variables of faculty numbers and graduate student Graduate Record Examination scores against output measurements of PhDs awarded and publications per year.
Of the eight disciplines considered, chemistry, civil engineering, physics and evolutionary biology evinced the greatest benefit.
"It's unequivocal that HPC provides an advantage in doing research in several fields," says Amy Apon, lead author of the study. "It's not uniform across all fields. But for fields where it matters, it matters a lot."
Read more about how supercomputers give universities a competitive edge here.