In the month of November 2014, in New Orleans (USA), Intel introduced a number of technologies designed to strengthen its leading position in the corporate segment of HPC (High Performance Computing, HPC). In particular, the conference Supercomputing Conference (SC14) announced a new generation of processors Intel Xeon Phi (codenamed Intel Knights Hill) and the new architecture Intel Omni-Path for high-speed interconnects.
Intel presented the Intel Xeon Phi 3rd generation (codenamed Intel Knights Hill), which will be created using the 10-nanometer manufacturing process and to support Intel Omni-Path Fabric. Intel Knights Hill will be presented followed by preparing for release of Intel Knights Landing, the first system on the basis of which will be released next year. Industry investments in processors Intel Xeon Phi continue to increase. It is expected that more than 50 companies offer systems based on the new version of Intel Knights Landing. Also, in many systems will be used as a module in the expansion-card interface PCIe. At the moment, the performance of systems based on the Intel Knights Landing more than 100 petaflops.
In particular, Intel Knights Landing is used in supercomputer Trinity, which is a joint project of the Los Alamos and Sandia National Laboratories, and supercomputer Cori National Science Computing Centre for Energy Studies US Department of Energy. In addition, the company DownUnder GeoSolutions recently announced large commercial projects on the basis of the current generation of co-processors Intel Xeon Phi, and the National Supercomputing Center IT4Innovations recently announced the creation of Europe’s largest computing cluster based on Intel Xeon Phi.
Intel introduced the information that Intel Omni-Path Architecture will provide a data rate of 100 Gb / s and up to 56% lower switching delay in clusters of medium and large size compared to alternative solutions based on InfiniBand1. Intel Omni-Path Architecture will use the 48-port switching chip to provide higher port density and a higher level of scalability in comparison with developments based on InfiniBand to 36 ports. It is expected that the provision of up to 33% more nodes to 1 chip commutation will reduce the required number of switches to simplify the design of systems and reduce infrastructure costs.
Note that according to the 44th edition of the list TOP500, which was presented at a fraction of systems based on Intel products accounted for 86% of all supercomputers included in the list, and 97% of new systems that have been added to the list. Since the release of the first generation of Intel Xeon Phi was 2 years ago, the share of multi-core systems accounted for 17% of the performance of all computers TOP500.
This was all about the latest advancement and technology in the era of processors world. And yes the new generation Intel Xeon Phi 3rd generation (code named Intel Knights Hill) will define the new standards and limits of the processing world, stay tuned for more information.