Most of us consider a PC fast enough if it can play 8K footage or the latest Far Cry without stuttering at 60 frames per second. On the other hand, many complex activities need billions of computations per second, which a desktop with an i9 CPU cannot do.
That is where supercomputers can help. What exactly are they?
They perform well, allowing governments and businesses to take on tasks that traditional computers couldn’t manage. Today’s supercomputers are pushing science to new heights that human intelligence would be unable to achieve on its own.
The power of these computers has increased substantially over the last few decades. In 1985, the world’s most powerful supercomputer, Cray-2, could handle 1.9 billion floating-point operations per second (FLOPS), or 1.9 gigaflops. The parameter is used to assess the power of these machines.
The performance of a supercomputer is expressed in floating-point operations per second (FLOPS). FLOPS is a more precise figure in scientific computing than measuring instructions per second.
A modern PlayStation 4 gaming console, for comparison, achieves 1.84 teraflops, over a thousand times greater. There are at least 500 supercomputers today that can do more than one petaflop (one billion flops).
Let’s have a look at the top supercomputers.
1. Fugaku (Japan – 442,010.0 teraFLOPS)
In June 2020, the Japanese supercomputer Fugaku sped past all opponents to take the top place in the world’s most powerful computing machines’ twice-annual list.
Fujitsu’s Fugaku supercomputer, constructed in collaboration with the federal Riken research facility, was able to do nearly three times as many operations per second as the previous leader of the list, the Summit supercomputer from the United States.
According to the Riken lab, no other computer has ever accomplished this feat. In addition, Fugaku possessed the most cores of any computer in the competition, the best theoretical peak performance for calculations, and the biggest power capacity. Meanwhile, Sunway TaihuLight, China’s fourth-ranked supercomputer, has more cores than Fugaku, with 10.6 million vs. 7.6 million.
The machine’s power is also utilized to investigate the coronavirus in the instance of Fugaku, especially to do simulations of how respiratory droplets flow through the air in various situations such as workplaces and train carriages.
The Japanese government has accomplished a great deal, but developing such a formidable system has not been cheap. The project’s R&D, purchases, and application development have cost the government over $1 billion since 2014.
2. Summit (USA – 148,600.0 teraFLOPS)
It’s also the third most energy-efficient supercomputer, with 14.66 gigaFLOPS per watt power efficiency.
IBM established the Oak Ridge National Laboratory in Tennessee for the US Department of Energy. Its 2.41 million cores give it a performance of 148.6 petaflops, equal to two basketball courts.
The network comprises 185 miles of fiber optic cable and consumes enough energy to power 8,100 residences.
Summit broke the exascale threshold for the first time in 2018. As a result, it reached a maximum throughput of approximately 2 billion computations per second when processing genetic data.
Summit is a step toward the United States’ goal of creating a fully functional exascale computing ecosystem for scientific purposes.
3. Sierra (USA – 94,640.0 teraFLOPS)
Sierra, the second most powerful supercomputer on the list, is also manufactured by IBM. Lawrence Livermore National Laboratory in California houses it. The hardware is similar to Summits.
In addition, it combines IBM’s Power 9 CPUs with NVIDIA’s Volta GPUs on a single chip.
Sierra was created to evaluate the performance of nuclear weapon systems in mind. It’s utilized in stockpile management, the US dependability testing program, and nuclear weapon maintenance without atomic testing.
4. Sunway TaihuLight (China – 93,014.6 teraFLOPS)
With TaihuLight, a machine constructed by the National Centre for Engineering Research and Parallel Computing Technology and deployed at the National Supercomputing Centre in Wuxi, China, stood at the forefront of global supercomputing until Summit and Sierra came into operation in 2018.
TaihuLight’s computational power comes from a multi-core SW26010 CPU developed in-house, containing computing and management processing features.
Because of its 260 processing components, a single SW26010 may achieve peak performance of almost three teraFLOPS (integrated into one CPU). In addition, in most applications, a scratchpad memory that functions as a user-controlled cache is built into each computer processing device, decreasing the memory bottleneck.
TaihuLight has been utilized to model the cosmos using 10 trillion digital particles and life sciences and pharmaceutical research. On the other hand, China is attempting far more. The government has proclaimed its intention to be the world leader in artificial intelligence by 2030.
In some ways, TaihuLight results from the trade war between China and the United States. However, it was built entirely without American technology due to US limitations.
5. Perlmutter (USA – 70,870.0 teraFLOPS)
In June 2021, Perlmutter joined the TOP10 list of the TOP500. It came in 5th place in the most recent edition. It is built on the HPE Cray EX235N “Shasta” platform and runs on AMD EPYC 7763 CPU.
It has accelerated nodes from NVIDIA A100 SXM4. The Slingshot-10 connection links all of the cores together. In the Linpack performance benchmark, Perlmutter scored 70,870 TFlop/s.
HPE built it, housed in the National Energy Research Scientific Computing Center in the United States. Nuclear fusion and cosmology researchers have exploited Perlmutter’s work.
6. Selene (USA – 63,460.0 teraFLOPS)
It has a 555,520-core AMD EPYC 7742 processor. It is powered by an NVIDIA DGX A100 GPU, with Mellanox HDR Infiniband connecting the cores.
In the Linpack performance benchmark, it earned a score of 63,469 TFlop/s. NVIDIA created Selene at Argonne National Laboratory in the United States.
Selene has been employed in quantum chemistry studies and is now being used to learn more about coronaviruses.
7.Tianhe-2A (China – 61,444.5 teraFLOPS)
Tianhe-2A is the world’s most extensive installation of Intel Ivy Bridge and Xeon Phi processors, with over 16,000 computer nodes. Milky Way-2A is another name for it.
Even though each node has 88 gigabytes of memory, the overall memory (CPU+coprocessor) is 1,375 terabytes.
The National University of Defense Technology produced Tianhe-2A, housed in China’s National Supercomputer Center. It is primarily utilized to ensure the security of the government.
This supercomputer cost China 2.4 billion yuan (about $390 million). It’s now primarily utilized for simulations, analysis, and government security.
8. Juwels Booster Module (Germany – 44,120.0 teraFLOPS)
JEWELS isn’t a store-bought computer. However, being one of Germany’s top research institutions. It runs on a 449,280-core AMD EPYC 7402 CPU.
Mellanox HDR Infiniband/ParTec ParaStation ClusterSuite is used to connect the cores. In the Linpack performance benchmark, it earned a score of 44,120 TFlop/s.
It is manufactured by Atos and installed in the Juelich Forschungszentrum in Germany. It’s Europe’s most powerful supercomputer for particle and nuclear physics research.
The JEWELS cluster is in great demand even though it is yet incomplete. According to JSC, eighty-seven projects have been assigned computing time on the machine, which is scheduled for the next three months.
Targeted applications cover a wide range of scientific and engineering disciplines, emphasizing brain research. For example, JUWELs is one of the primary computational resources for the EU’s Human Brain Project in the coming years, thanks to JSC’s expertise in this field.
Complex simulations with a vast number of variables are done on supercomputers. Economic and climate modeling, neurological research, and nuclear science are all examples of typical applications.
Although it appears that the United States, Japan, and China have a grip on supercomputing technology, this is not always the case. Germany, the United Kingdom, and Saudi Arabia are also rivals with amazing breakthroughs in computer engineering. Only time will tell if these technological behemoths can maintain their grip on the top 500 rankings.
Read More Ten Scientists Who Changed The World