TOKYO-BASED Semiconductor Energy Laboratory will use a hybrid high-performance supercomputer to research new semiconductor technologies.
The large-scale computing system provided by SGI will facilitate the research and development of field-effect transistors that use oxide semiconductors, as well as a number of other products related to active matrix displays, non-volatile memory and lithium-ion batteries.
As SEL is focused on exemplary research and development in the semiconductor field, it has worked to ensure that the computational resource driving its research has improved in parallel with its own offerings. This marks the second time that SEL has implemented a large-scale overhaul of its systems.
The computer combines SGI's large-scale shared-memory UV 2000 server and ICE X blade-type cluster server into a unified, large-scale system.
Application performance has increased as much as 2.6-fold compared to SEL's previous system, with a theoretical overall peak performance of 180 teraflops.
A large-scale shared-memory server allows a single system to make use of extensive memory space, while distributed cluster servers allow high-speed processing through massively parallel computations. By combining these two server types into a single, unified system, and by appropriately designating computation processes according to their content and characteristics, SEL has been able to create a computational environment of unprecedented efficiency, facilitating cutting-edge research and development into ground-breaking semiconductor technologies.
The SGI UV 2000 large-scale shared-memory server has 768 cores from the Intel Xeon processor E5-4600 v2 product family with 16TB of memory. The SGI ICE X large-scale blade-type cluster server combines 7,200 cores from the Intel Xeon processor E5-2600 v2 product family with 46TB of memory. The systems' nodes are connected by high-speed InfiniBand 4x FDR.