« Back

Engineering Student Assesses Supercomputing Performance

An Ohio State University engineering student has established side-by-side performance comparisons for three different supercomputing technologies, a project that could help researchers determine the best architecture for their specific parallel computing needs. The specific architectures evaluated included a 128-processor Cray massively-multithreaded computer at the US Pacific Northwest National Laboratory; a 16-processor IBM shared-memory machine at the Ohio Supercomputer Center; and a system consisting of 240 Nvidia GPUs. Ohio State graduate student Saniyah Bokhari tested the systems using a subset-sum problem. The results showed that the Nvidia system works best for small problem sets; the IBM computer is good for medium-sized sets; and the Cray is ideal for large problems. Bokhari published her complete results in Concurrency and Computation: Practice and Experience. (PhysOrg.com)(Ohio Supercomputer Center)

Comments
Trackback URL: