RSC Tornado SUSU is one of the first RSC's liquid-cooled projects. It was deployed by RSC's specialists in Super Computer Center of the South Urals State University in Chelyabinsk in 2010. In 2013 it underwent a planned upgrade and rose to 127th position in Top500 list (November 2013).
This supercomputer is still in use. Its peak performance is 473,6 teraflops with Linpack performance of 288,2 teraflops.
Project timeline
October 2013 - January 2014
Location
Chelyabinsk, Russia
Features
- Achieving higher performance. To achieve maximum performance, supercomputer was supplied based on Intel® Xeon® 5600 processor platform. The existing infrastructure was later upgraded to RSC Tornado SUSU supercomputer with Intel® Xeon Phi™ coprocessors.
- Compact size and ultra-high computing density. RSC Tornado SUSU supercomputer system was developed by RSC Group on the basis of innovative RSC Tornado architecture providing highest computing density in the industry with standard server boards and x86 processors. With direct liquid cooling on server board level we managed to deploy the entire solution (computing part, power and cooling systems, communication equipment, storage system) just in 50 sq. m. room (3 times less as compared to air-cooled cluster with similar performance).
- High energy efficiency and record PUE level. Combination of Intel® Xeon® processors и Intel® Xeon Phi™ coprocessors with direct liquid cooling allowed to reduce annual power costs by 60% on the average. Supercomputer was built on standard x86 architecture (without GPU acceleration) and was one of the most energy-efficient supercomputers in Russia and CIS according to Green500 global rating.
- Improved reliability. Liquid cooling provides high-precision heat dissipation increasing the lifetime of electronic components and improving fault resistance of the entire solution. The design has no moving elements such as fans and regular hard disks (Intel® SSD solid state drives are used instead). This almost completely removes the noise, prevents vibration and significantly improves system reliability.
