The Department of Energy’s Oak Ridge National Laboratory is expected to take delivery of a new IBM system, Summit, in early 2018. Summit will be capable of 200 peak petaflops, making it twice as fast as Sunway TaihuLight, China’s new system announced this past Monday. As of now, TaihuLight is listed in the TOP 500 list of fastest supercomputers in world. TaihuLight packs a Linpack benchmark score of 93 petaflops and a claimed peak of 124.5 petaflops. Linpack benchmark has become the perferred yardstick for measuring performance of supercomputers, as it doesn’t record overall performance but rather performance of the system when solving dense systems of linear equations. This gives the best approximation of real-world performance. The Summit however, will emply IBM power 9 and Nvidia Volta GPUs. Using about 3,400 nodes, Summit will deliver five times the computational performance of Titan’s 18,688 nodes. Over half a terabyte of coherent memory in each node, as well as 800GB of non-volatile RAM, serving as a burst buffer or extended memory. Cray also announced this week that it’s XC systems are now available with the latest Intel Xeon Phi (Knights Landing) processors. The XC systems should deliver a 100% performance boost over prior generations, with an adaptive design that supports multiple processor and storage technologies in the same architecture. Cray also revealed the Sonexion 3000 Lustre storage system, which will deliver speeds of almost 100 GB/sec in a single rack.
If you would like to educate yourself in more detail about the information presented in this blog post please visit: IBM to deliver 200-petaflop supercomputer by early 2018; Cray moves to Intel Xeon Phi