In the 56th edition of the TOP500, Japan, USA, Germany, Italy, and France have been the most dominant of in the 100 top of supercomputers with the most power. Japan’s Fugaku supercomputer remained as the top computer system which was closely followed by IBM’s Summit and Sierra in the U.S.
The Top500 project is a database of those supercomputers with the highest performance around the world. It is released every two years. The non-profit organization published its November 2020 rankings last week. In this year’s report, two new systems made it into the top 10 list, the entire list had the lowest number of entries that were new since the beginning of the project at the beginning of 1993.
This is why Top500 representatives said that the current list shows a flattening growth curve.
What are the results?
The world’s supercomputers have become not only more efficient and more efficient, but they are also smarter and are able to handle a greater range of tasks.
About 70 percent (140) of devices that are listed on the TOP500 list including 8 of the Top 10 have been powered by NVIDIA’s technology as well as “increasingly making use of AI to assist researchers in making discoveries more quickly.”
NVIDIA’s complete HGX AI supercomputing platform helps speed up scientific computing as well as data analytics and AI jobs.
“The move to incorporate AI to HPC (High-Performance Computing), and a platform that goes beyond the traditional supercomputing centers is a major shift in this field, which was, ever since Seymour Cray’s CDC 6600 was launched in 1964, has been focused on harnessing ever-larger and more powerful computers for computationally intensive simulation and modeling,” NVIDIA said in the blog post.
In October of this year in October of this year, the Italian interuniversity group CINECA announced plans to create the world’s most powerful AI supercomputer. The brand new “Leonardo” machine is predicted to offer the equivalent of 10 Exaflops of FP16 speed in AI tasks. It will have approximately 14,000 Nvidia A100 GPUs as well as 3500 Intel Sapphire Rapids CPUs.
When it comes to supercomputers, the capabilities in performance are outlined by a standard rate known as FLOPS, or floating-point operation per second. This is the rate that indicates the number of floating-point arithmetic calculations that a system is able to perform in a second. In contrast, one petaFLOPS (PFLOPS) computer system is able to do the equivalent of one-quadrillion (10^15) floating-point calculations per second. a 1 exaFLOPS (EFLOPS) computer system is able to perform 1 quintillion (10^18) floating point operations in a second.
SIX WAYS THE WORLD’S FASTEST SUPERCOMPUTERS HAVE CHANGED OUR LIFESTYLE
However, Intel has remained the most popular processor manufacturer with a majority of systems being running its Xeon as well as Xeon Phi chips.
With 221 systems included that are listed, China has exerted unequivocal dominance over the list. It is being followed by U.S. with 113 systems as well as Japan having 34. In regards to overall performance it is it is the U.S. continues to lead with 668.7 petaflops, as compared the Chinese’s 564.0 petaflops. However, Japan has an aggregate performance of 593.7 petaflops.
Supercomputers, with the unfortunate consequence of using huge quantities in energy. They are being ranked on the Green500 according to energy efficiency. In the Green500 list, the brand new NVIDIA DGX SuperPOD located in the U.S. was identified as the most efficient in energy use. It was closely followed by MN-3 in Japan as well as the Atos-built JUWELS Booster Module that was built in Germany.
Top 10 supercomputers
Some notable developments within the top 10 are an addition to the top 10 of brand new systems, as well as a new benchmark created in Fugaku. Fugaku supercomputer.
In the wake of the introduction of new hardware that boosted this system Fugaku has increased the High-Performance LINPACK (HPL) benchmark up to 442 petaflops. This was in comparison with the previous 416 petaflops that the system had when it was launched in June 2020 . That’s an increase of 6.4 percent speed increase.
It is the HPL benchmarkthat measures the efficiency of a specific system in solving a complex set of equations, is used to identify the Top500.
1. Fugaku, Japan
Created in collaboration with Fujitsu, Fugaku is installed at the RIKEN Center for Computational Science (R-CCS) located in Kobe, Japan. With the addition of hardware, the system has set an all-time record in the world of 442 petaflops on HPL which is three times faster than the second most powerful system on the list of.
Director of RIKEN, Satoshi Matsuoka, said that the improvements was due to them “finally having the ability to use the whole machine instead of just a small portion or a portion.”
Since June’s competition the group has managed tweak the code to ensure the best performance. “I do not think we are able to make any progress,” Matsuoka said.
2. Summit, U.S.
The facility is located in situated at Oak Ridge National Laboratory (ORNL) located in Tennessee, Summit was built by IBM and is the most efficient technology within the US. It was released in the year 2018 It boasts a speed of 148.8 petaflops and includes 4356 nodes, each has 2 22-core Power9 CPUs as well as 6 NVIDIA Tesla V100 GPUs.
Recently two teams that worked on Summit were awarded the famous Gordon Bell Prize for exceptional achievement in high-performance computing. This is often known as the Nobel Prize’ in supercomputing.’
3. Sierra, U.S.
A system located at that is located in the Lawrence Livermore National Laboratory (LLNL) located in California, Sierra has an HPL number that is 94.6 petaflops. Each one of the 4,320 Nodes armed with 2 Power9 CPUs , and 4 NVIDIA Tesla V100 GPUs it features similar architecture to that of Summit. Summit.
Sierra also reached the 15th spot in the Green500 List of the world’s energy efficient supercomputers.
4. Sunway TaihuLight, China
It is located by the China’s National Supercomputing Center located in Wuxi, Sunway TaihuLight previously was ranked as the top computer in Wuxi for two seasons ( 2016-2017). However, its position has since dropped. It was in third place last year, it is now fourth.
It was developed for China’s National Research Center of Parallel Computer Engineering & Technology (NRCPC) which achieved the record of 93 petaflops in it’s HPL benchmark. It is exclusively powered with Sunway SW26010 processing units.
5. Selene, U.S.
In-house installed by NVIDIA Corp, Selene jumped to fifth place from the seventh spot in the June ranking. Following a recent upgrade, Selene achieved 63.4 petaflops on HPL which is nearly double the previous score that was 27.6 petaflops.
NVIDIA revealed Selene Selene, their AI supercomputer in June of this year, after having built and operating Selene within less than one month. Its primary uses include the development of systems and tests and in-house AI applications, as well as designing chips.
6. Tianhe-2A, China
It has 61.4 petaflops Tianhe-2A (aka MilkyWay-2A) has climbed one place to sixth place. It was developed through the Chinese National University of Defense Technology (NUDT) It is housed in Guangzhou’s National Supercomputer Center located in Guangzhou.
Tianhe-2A runs on Intel Xeon CPUs and NUDT’s Matrix-2000 DSP accelerators. It is designed to support analysis, simulation, as well as security applications for the government sector. It was the number one position from June 2013 through November 2015.
7. JUWELS Booster Module, Germany
The Atos-designed JUWELS Booster module is the latest addition to the list of. Its BullSequana machine was recently put in place on the Forschungszentrum Julich (FZJ) in Germany and is the most powerful machine in Europe with a capacity of 44.1 HPL petaflops.
Based on the modular systems architecture JUWELS runs on AMD processors as well as NVIDIA GPUs, which are similar to Selene. Selene system.
8. HPC5, Italy
An Dell PowerEdge computer system called HPC5 has been installed by Italian firm Eni S.p.A and is located within the Eni’s Green Data Center in Italy. HPC5 is among the most robust and environmentally green computing systems available HPC5 can be used to discover the development of new sources of energy..
With a capacity that is 35.5 petaflops, it’s the most powerful machine on the list of systems used for commercial use on a customer’s site. It is powered by Intel Xeon Gold CPUs and NVIDIA Tesla V100 GPUs.
9. Frontera, US
Installation at the Advanced Computing Center at the University of Texas at Austin in September, Frontera is a Dell C6420 system that’s powered with Intel. Utilizing 448,448 Intel Platinum Xeon cores, Frontera can achieve 23.5 petaflops.
Frontera helps research across all fields of science, from to quantum mechanicsand drug development and eradicating emerging viruses and the physics behind black holes.
10. Dammam-7, Saudi Arabia
Dammam-7 is the next release on the list. It was installed by Saudi Aramco in Saudi Arabia The HPE Cray CS-Storm systems utilize Intel Gold Xeon CPUs and NVIDIA Tesla V100 GPUs. It has a performance of 22.4 petaflops and is the second supercomputer commercially among the top 10.
