Always a supporter of research and the education market, Acer is proud to jointly announce with Taiwan's National Cheng Kung University (NCKU) a revolutionary system design for high performance computing. Making user of Acer's Altos R380 F2 with two Intel Xeon Phi cards per node, this new architecture uses a daisy-chain design and intelligent networking to do away with the need for expensive switching technology. In so doing, this architecture eliminates the bottlenecks which are often associated with fat tree, switch-based topologies, and can help some of the costs associated with parallel computing.
A leader for supercomputing in Taiwan, NCKU collaborated with an international team of researchers both in Asia and the US to design, implement , and benchmark the performance of the system - reaching results comparable to conventional HPC architectures that employ a single switch or series of switches. NCKU used eight Altos R380 F2 servers with Intel Xeon E5-2630 v2 processors and Intel Xeon Phi 7120P coprocessors to achieve 15 TFLOPS, equal to the performance expected with more conventional architecture. Moreover, the design used adds flexibility to users who may expand their systems overtime without having to worry about port limitations in a switch or network of switches.
Evis Lin, AVP of commercial products at Acer, commented, "The results of NCKU represent the ability of Acer to work with and enabling our cutting-edge customers with hardware that can keep-up with their most aggressive demands. Acer is committed to the education sector no matter if it is providing students with a notebook computer, or a research institute, like NCKU, looking for the latest in supercomputing technology."
Director Huang of NCKU also stated, "We were impressed with not only the hardware of the Acer systems, but also the company's ability to work with us to help achieve these results. Acer brings to bear a growing amount of technical computing experience to Taiwan and East Asia, and has been a great partner for us."
Users or researchers interested in the results of NCKU may visit the NCKU website at http://supercomputer.ncku.edu.tw/bin/home.php