China’s Tianhe-2 still the fastest supercomputer in the world, but the US is catching up

China’s Tianhe-2 still the fastest supercomputer in the world, but the US is catching up

The results are in, and they’re not a huge surprise: For the fifth consecutive time, China’s Tianhe-2 remains the fastest supercomputer in the world, with a Linpack benchmark performance of 33.86 petaflops, or quadrillions of floating-point calculations per second. That’s the word from the 45th edition of the twice-annual TOP500list of the world’s most powerful supercomputers.

Despite the expected result, some interesting bits of info come out of this latest roundup. The United States still has the most systems in the list of any country, with 233 (up from 231 six months ago and down from 265 in late 2013); the second and third-place systems are both from the United States as well. Meanwhile, Europe has 141 machines on the list. Significantly, three new systems belong to China’s Lenovo, although China itself is represented less this time around, with 37 supercomputers as opposed to 61 last year.

The average performance of the TOP500 has gone up considerably in the past six months. The total combined performance of all 500 supercomputers has reached 363 petaflop/s, compared with 309 last November and 274 one year ago. 98% of the systems use processors with six cores or more, while 88.2% have at least eight cores per CPU. Eighty-eight of the 500 systems use accelerator/coprocessors, including Nvidia (52), ATI Radeon (4), and Intel Xeon Phi (33); four of the systems use a combination of Xeon and Nvidia processors.

The top 10 in general consist of systems installed in either 2011 or 2012, with the exception of a new Saudi Arabia entry at #7 on the list. Here’s a rundown of the top 10:

1. Tianhe-2: TH-IVB-FEP Cluster; National Super Computer Center in Guangzhou, China; 3.12 million cores (33.86 petaflop/s).

2. Titan: A Cray XK7 system at the Department of Energy’s Oak Ridge National Laboratory (17.59 petaflop/s).

3. Sequoia: An IBM BlueGene/Q system located at the Department of Energy’s Lawrence Livermore National Lab in California, with 1.57 million cores.

4. K Computer: A SPARC64 system with 705k cores at RIKEN Advanced Institute for Computational Science in Japan.

5. Mira: IBM BlueGene/Q; DOE/SC/Argonne National Laboratory, US; 786k custom IBM cores.

6. Piz Daint: Cray XC30 with 116k Xeon and Nvidia cores; located at the Swiss National Computing Centre in Switzerland.

7. Shaheen II: A Cray XC40 at King Abdullah’s University of Science and Technology in Saudi Arabia, marking the first appearance of a Middle East supercomputer in the top 10 (5.536 petaflop/s).

8. Stampede: A Dell PowerEdge C8220 system with 462k Xeon Phi cores at the Texas Advanced Computing Center/University of Texas in the US.

9. JUQUEEN: BlueGene/Q, 459k custom IBM cores, located at Forschungszentrum Juelich, Germany.

10. Vulcan: BlueGene/Q, 393k custom IBM cores, DOE/NNSA/LLNL in the US.

All of this could change if IBM (or someone else) can figure out how to build a real quantum computer. IBM has gone on record as saying that if one could be built with just 50 quantum bits (qubits), instead of the current maximum of four, and that can detect both types of quantum errors and be scalable to larger systems, “no combination of today’s TOP500 supercomputers could successfully outperform it.”

In the meantime, the DOE has commissioned two IBM-and-Nvidia systems in a $425 million deal that’s scheduled to bear fruit in 2017 or 2018, with a peak performance target of 150 petaflops. For the moment, though, China’s lead seems secure.