VISUALIZATION

By Uwe Harms, Munich -- With a faster clocked vector processor NEC improves the performance the NEC SX-6 from 8 to now 9 GigaFlop/s. An other feature of the additionally optimized machine is the doubled size of the main memory. The performance of biggest system with 1024 processors (128 nodes) grows to 9.2 TeraFlop/s. In the mid of its life, NEC has risen the clock of the vector supercomputer NEC SX-6. The peak vector performance grew to now 9 GigaFlop/s. The main memory is now 128 GigaByte for a node with 8 processors. The peak vector performance of the biggest system sums up to 9.2 TeraFlop/s with a main memory of up to 16 TeraBytes. Now NEC gives a definition of the peak performance of a system. They add the performance of the add and multiply vector processing pipes plus the performance of the divide unit plus the performance of the scalar processor. Now a node, 8 processors, has a peak system performance of 93,3 GigaFlop/s. The theoretical peak performance of the biggest system now sums up to 11.9 TeraFlop/s. One of the specific feature of a vector computer is the high data transfer rate between main memory and processor. It grew from 256 GByte/s to 288 GByte/s in the node - 8 processors - of the new machine. Thus the peak transfer rate of the biggest system is 38,8 TeraByte/s. The new NEC SX-6 models are available in October 2003. www.hpce.nec.com Uwe Harms