With the help of supercomputers, scientists can perform simulations in various areas, among other things. Whether it’s space research, the discovery of oil and gas reserves, weather forecasts or nuclear weapon research, the supercomputers are involved and help researchers.
The performance of supercomputers is measured in Floating Point Operations Per Second, FLOPS for short. It specifies how many floating-point operations (additions or multiplications) can be performed per second. The systems consist of many individual computation units, the so-called nodes, which are interconnected and thus become a massive supercomputer. In these, nodes are often traditional Intel chips, such as the Xeon processor to find. Depending on the application, graphics units from Nvidia can also be found in the systems.
On this article, we made the list of the most powerful supercomputers in operation worldwide. Unsurprisingly, govt agencies are the operators of the expensive facilities.
10. Cori (USA)
Cori, named after the American biochemist Gerty Cori, she was the first woman with a Nobel Prize in Medicine. The system is in the National Energy Research Scientific Computing Center, and it can reach 14.01 petaflops (Rmax).
This supercomputer splits into two systems. While in the first part is of Intel’s Haswell-Xeons on the 2,388 nodes are used (32 cores per node), and the second part “Knights Landing” Xeon 7250 used, each with 1.4 GHz. Here are another 9,688 nodes with 68 cores each.
9. Trinity (USA)
Just ahead of Cori is also operated by the US Department of Energy Trinity supercomputer, which is located in the Los Alamos National Laboratory. Again, the Node uses Intel’s Xeon 7250 with 68 cores. With a total of 979,968 cores, the system can reach up to 14.14 petaflops.
8. Sequoia (USA)
Sequoia is another supercomputer from the US in the top 10 of the current supercomputers. The Lawrence Livermore National Laboratory’s (LLNL) system, which runs at 17.17 petaflops, relies on IBM’s Power BQC, each with 16 cores. Overall, the supercomputer is composed of 1,572,864 cores. The system has been running in this configuration since mid-2012.
7. Titan (USA)
Titan manages 7th place at 17.59 petaflops at the Oak Ridge National Laboratory also from the US. As a CPU manufacturer, it has not been decided for Intel or IBM, but since the end of 2012 on AMD’s Opteron CPU back. Specifically, according to the operator, these are 16-core processors; in total it should be 299,008 cores. The GPUs are 18,888 Nvidia Kepler K20X used.
6. Piz Daint (Switzerland)
The Piz Daint baptized system is found at the Swiss Federal Institute of Technology in Zurich and achieves sixth place with 19.59 petaflops.
The Swiss National Supercomputing Center uses Intel’s Xeon processors in its supercomputer and consists of two primary node types. For example, 1,813 Cray XC40 nodes are used, each equipped with two Xeon E5-2695 v4 (2.1 GHz) and having 2 × 18 cores. There are also a further 5,320 XC50 nodes, each running a Xeon E5-2690 v3 (2.6 GHz, 12 cores).
5. AI Bridging Cloud Infrastructure (Japan)
The new system operated by the Japanese National Institute of Advanced Industrial Science and Technology (AIST) is named ” AI Bridging Cloud Infrastructure ” (AIBC) and it can reach 19.88 petaflops.
The system, supplied by Fujitsu, consists of Primergy CX2550 M4 nodes, which use Intel’s Xeon Gold 6148 with 20 cores and 2.4 GHz. In total, 391,680 cores are available to the supercomputer. The GPU is Nvidia’s Tesla V100.
4. Tianhe-2 (China)
The first Chinese participant in the Top 10 is Tianhe-2 (also known as Milky Way-2), who stands at the National Supercomputer Center in Guangzhou, China, this supercomputer can reach 61.44 petaflops.
The system has been recently equipped with new processors, but it still does not make it to the podium prior to the upgrade it could only 33.9 petaflops. Inside, the Nodes are now Intel’s Xeon E5-2692 v2, each with 12 cores, which are clocked at 2.2 GHz. Overall, the supercomputer has 4,981,760 cores.
3. Sierra (USA)
In third place is again a system from the USA called Sierra, it is in the Lawrence Livermore National Laboratory, which is operated by the US Department of Energy.
Each node, for a total of 4,320, is based on IBM’s 3.1GHz Power9 chip. GPU is Nvidia Volta GV100 is used. Overall, the 1,572,480 cores reaching up to 71.6 petaflops.
2. Sunway TaihuLight (China)
The second place in the world of supercomputers goes to Sunway TaihuLight in China. This system is located at the National Supercomputing Center in Wuxi has 10,649,600 cores reaching up to 93 petaflops. Sunway TaihuLight was in first place since June 2016 but now had to give up to a new leader.
1. Summit (USA)
Summit, stands in the 1st place in the world of supercomputers found in the Oak Ridge National Laboratory. The Department of Energy of the United States has control of this system.
This supercomputer has 4,600 individual nodes, each of these systems has two IBM Power9 processors with 22 cores and six GPUs in the form of Nvidia’s Volta V100. Also, there are 512 GB of DDR4 memory and 96 GB of “High Bandwidth Memory” (HBM2) for the GPUs. Each CPU core supports up to four hardware threads and it can reach up to 122.3 petaflops.