During the past month of June Top 500 update introduced (It is updated in June and November of each year) and tracks the evolution of the performance and architecture of the most powerful supercomputers in the world, this being a report in which we can know the speeds at which they are based on the High Performance Linpack benchmark (HPL), which measures the ability of systems to solve a dense system of linear equations.
And in this new edition that was presented, in the top ten in the last ranking, five are in the United States, three in China, one in Germany and one in Italy.
In this new list of 2021 the Fugaku continues in the first position since last year and now it scored 442 petaflops in the HPL benchmark, with 7.630.848 cores. This result is identical to that obtained in November 2020, but higher than the 415,53 petaflops obtained last year when it topped the list.
The supercomputer It is based on Fujitsu's custom Arm A64FX processor and does not use a GPU, It is installed in the RIKEN Center for Computational Sciences (R-CCS) in Kobe, Japan, this machine has been fundamental in the fight against covid-19.
In the second place we can find with 148.8 petaflop to the Summit at Oak Ridge National Laboratory, USA, which is now almost three years old and uses 27.000 Nvidia Tesla V100 GPUs paired with 9.000 IBM Power9 CPUs.
In third place is the supercomputer «Sierra», which is similar in design to the Summit and is operated by the Lawrence Livermore National Laboratory (United States).
In fourth place with 94,6 petaflop, there is the Sunway TaihuLight (China).
While the fifth place is occupied by Perlmutter, installed at the Berkeley Lab's National Energy Research Scientific Computing Center. The HPE Cray EX system is powered by AMD Milan processors and 100GB Nvidia A40 GPUs, linked by HPE's Slingshot technology. According to the HPL benchmark, the system delivered 65,69 Linpack rMax petaflops over a potential peak (rPeak) of 93,75 petaflops, which corresponds to a Linpack efficiency of 70%. Perlmutter is the largest Nvidia A100 GPU system and the largest HPE Cray EX (public) system deployed to date, based on the architecture formerly known as "Shasta".
In sixth place is «Selene» Nvidia's system with 63,46 petaflops and in which Nvidia's SuperPod A100 DGX modular architecture is implemented with AMD Eypc Rome processors and Nvidia's 100GB A80 GPUs.
In seventh place is Tianhe-2A, built in China, which lost one spot in this ranking and delivered 61,4 Linpack petaflops. It is equipped with Intel Xeon chips and custom Matrox-2000 accelerators. It is installed in the National Supercomputing Center in Guangzhou.
In eighth position is the «JUWELS Module», Europe's most powerful supercomputer, an Atos BullSequana supercomputer equipped with Epyc CPUs and A100 GPUs. Located in Germany, it is capable of 44,1 petaflops.
Inside the main details of this top 10, we can find that Dell remains the number one academic provider and the number one commercial supplier with HPC-5, used by Italian fossil fuel giant Eni and Frontera from the Texas Advanced Computing Center at the University of Texas in ninth and tenth positions, respectively.
Besides that China still leads in terms of number of systems, despite a significant drop to 186 systems, from 214 machines six months ago and 226 machines a year ago.
The Top 500 noted that there hasn't been much definitive evidence why China's poor performance is happening, although this may be related to mounting sanctions against Chinese supercomputing companies and geopolitical tensions between the United States and China. These factors make it difficult for companies and research labs to acquire supercomputing equipment and are less likely to engage in a voluntary rating system that could place them in the target line of business for regulators.
AMD performed well, since with this new edition of the Top 500 we can notice a "marked increase" in its use. This is in line with AMD's overall progress in the server and supercomputing space, which is taking Intel shares.
Finally if you are interested in knowing more about it, you can check the details In the following link.