The Mole-8.5 machine uses Nvidia Tesla C2050 graphics processors and they have a theoretical peak performance of 1Pflops in double precision.
The supercomputer is made up of 372 computing and data processing nodes based on the FT72-B7015 board, made by Tyan and IPE. Each platform incorporates a 4U two way barebones server, coupled with two Xeon 5520 CPUs and six Tesla C2050 per each node. That gives a theoretical peak performance of around three Tflops in double precision.
The Linpack figure is 207.3 Teraflops and the leader of the Mole system at the Institute of Process Engineering at the Chinese Academy of Sciences is very excited about it.
Professor Wei Fe said the combo was “fantastic”. He said that he believed the combo is the right choice for many applications.