Tianhe-1A Supercomputer (天河-1A超級計算机)

by Prof. Xing Zuocheng, National University of Defense Technology, China

 :  20 Oct 2011 (Thu)
 :  4:00 - 5:10pm
Venue  :  Citigroup Lecture Theater (LT-A), HKUST
(Language : Mandarin)


Supercomputing has drawn significant attention from both research communities and industries around the world. As the pinnacle of computer engineering, supercomputers empower large-scale applications for scientific and technological advances. In this talk, we will introduce Tianhe-1A supercomputer (TH-1A, 天河-1A) developed by National University of Defense Technology (国防科学技术大学). Independently developed, TH-1A features high performance, high energy efficiency, high security, and high usability. For the first time in the world, TH-1A explored and implemented a heterogeneous high-performance parallel computing architecture of both GPUs and CPUs. Its peak speed is 4,700 TFLOPS, and the LINPACK benchmark result is 2,566 TFLOPS. In November 2010, TH-1A claimed the top ranking on the TOP500 List of the world’s top supercomputers, and it is currently ranked No. 2 worldwide.

TH-1A is a historic breakthrough of China’s supercomputing technology and highlights China’s recent advances in science and technology. It is tasked with making significant contributions to the national scientific advancement, the national strategic technical innovation of new industries, the development of national economy, and the training of innovative hi-tech workforce. TH-1A consists of a hardware system and a software system. The hardware system mainly consists of five subsystems: the computing system, the interconnect communication system, the I/O system, the monitoring and diagnostic system, and the infrastructure system. TH-1A team developed two networks, three ultra-large-scale integrated circuits, four types of nodes, and 15 different boards. The operating system, the compilation system, the parallel programming environment, and the scientific virtualization system were developed for the software system. TH-1A system has been installed at the National Supercomputing Center in Tianjin. It has served and is serving many applications, such as weather forecasting, oil and gas exploration, space exploration, bio-medical research, animation design, exploitation of new energy sources, material research, remote sensing data processing, and financial risk analysis.


Prof. Xing Zuocheng received his B.S. from Gullin University of Electronic and Technology, and M.S. and Ph.D. from National University of Defense Technology (国防科学技术木学). He currently is a full professor and PhD advisor in National University of Defense Technology. He is the Director of Parallel and Distributed Processing Architecture in the State Key Laboratory of Parallel and Distributed Processing, and the academic leader of High-Performance Computer Architecture, Microelectronics, and Solid-State Electronics.

Prof. Xing has led and engaged in over 20 China’s key research projects in computing systems, including Tianhe (天河) series supercomputers, Galaxy (银河) series supercomputers, and FT (飞騰) series high-performance general-purpose CPU. He headed the hardware system designs of Tianhe and Galaxy series supercomputers, and successfully developed a family of northbridge and system interconnect chips. Prof. Xing and his team developed FT-1000/1000A/1500 series multi-core processors, the world's first general-purpose 64-bit stream processor for scientific computing, and the world's first CPU compatible with Intel Itanium processor.

Prof. Xing is the project leader and deputy chief designer in many projects funded by the National High-Tech R&D Program of China (863計划). He is currently co-organizing the CPU developments under National S&T Major Project (国家科技重大專項) program.  Prof. Xing is awarded the First Prize of National Science and Technology Advancement Award (国家科技进步一等奬) twice.  He is on the editorial board of Journal of Computer Science & Technology and International Journal of Advance Research in Computer Science.  Prof. Xing published more than 80 papers in international and domestic journals and conferences including ISCA. He holds 16 patents on CPU architectures.