We often see powerful computers in science fiction that are able to process a lot of data very quickly. They make the modern personal computer look like a toy in comparison. However, what many people don’t know is that today’s supercomputers are already extremely powerful. They are blurring the line between science fiction and reality almost yearly. In this list, we will take a look at the world’s eight most powerful supercomputers.
How Supercomputers Are Ranked?
Before we can start to assess and rank supercomputers, we need to see how their performance is measured. Older computers were once rated on a modest scale called IPS (instructions per second). But nowadays, supercomputers are evaluated by their FLOPS (floating point operations per second).
1. Frontier (United States)
Frontier was the world’s first exascale supercomputer, built in 2022 by American multinational information technology company Hewlett Packard Enterprise in collaboration with its subsidiary Cray. It had a total of 8,730,112 cores and scored 1.1 EFLOPS (exaflops) on Linpack benchmark tests.
Frontier was based on the latest HPE Cray EX235a architecture and used a combination of AMD’s 3rd generation 64-core 2GHz 7A53s CPUs and MI250X GPUs.
Frontier is a supercomputer that is used for efficiency. It has a power efficiency rating of 52.23 gigaflops/watt. Each of its 74 computing cabinets weigh about 8,000 pounds (~3.63 tons), and the entire system costs a whopping $600 million in total.
2. Fugaku (Japan)
In 2020, the Japanese IT company Fujitsu created Fugaku as a successor to their older K computer. They designed it specifically to solve some of the world’s most pressing issues, with a focus on climate change. Fugaku is the fastest computer in the world and is helping researchers find new ways to combat global warming.
Fugaku was once the world’s most powerful supercomputer. It was beaten by Frontier in May 2022. It has 7,630,848 cores and is very fast, scoring 442 PFLOPS (petaflops), meaning it can do 442 quadrillion calculations per second.
The Fujitsu A64FX 48-core 2.2GHz processor is at the heart of the system, and it is very inefficient by today’s standards. The system cost more than $1 billion, and each of its cabinets weighs 1.6 tons, totaling ~700 tons.
3. LUMI (Finland)
LUMI (Large Unified Modern Infrastructure) was built by HPE in 2022 and is located in Finland. LUMI has a total of 1,110,144 cores and clocks a speed of 151.9 PFLOPS.
LUMI is a powerful and efficient supercomputer. It runs on the same processor as Frontier, and has a power efficiency rating of 51.63 gigaflops/watt. This makes it the second most efficient supercomputer in the world.
4. Summit (United States)
Summit was designed with scientific research in mind and built by IBM in 2018. It’s located at Oak Ridge National Laboratory in the same spot as Frontier. In total, Summit has 2,414,592 cores which allows it to hit 148.6 PFLOPS.
Summit runs on IBM’s POWER9 22-core 3.07GHz CPUs and Nvidia Tesla V100 GPUs and has a power efficiency rating that is on par with Fugaku.
5. Sierra (United States)
Sierra was built in 2018 and is very similar to Summit in that both supercomputers use the same IBM POWER9 22-core CPU architecture and Nvidia Tesla V100 GPUs. However, Summit was made for scientific research while Sierra was made for running simulations to test and maintain the United States’ nuclear weapons.
By running nuclear weapon simulations, the National Nuclear Security Administration (NNSA) of the US can check the effectiveness of its nuclear weapons without physical testing, which helps save costs. According to the Top500 list for November 2020, Sierra has a total of 1,572,480 cores and scores 94.64 PFLOPS. In terms of power efficiency, it has a rating of 12.72 gigaflops/watt.
6. Sunway TaihuLight (China)
Ranked at sixth on the list, we have the Sunway TaihuLight supercomputer. It was created in 2016 and is used for things such as weather forecasting, pharmaceutical research, and life sciences research, among other purposes.
The Sunway TaihuLight is a bit of an oddity on this list. It’s a 10,649,600 CPU core behemoth that runs on the Sunway SW26010 260-core 1.45GHz processor. With a power efficiency rating of 6.05 gigaflops/watt, it’s the most inefficient supercomputer currently in operation.
7. Perlmutter (United States)
In 2021, HPE built Perlmutter, the world’s seventh most powerful supercomputer. It was named after Nobel Prize winner Saul Perlmutter. Just like Sunway TaihuLight, Perlmutter can be used for nuclear fusion simulations, climate projections, material and biological research, and computational cosmology.
Perlmutter is a high-performance computer that is used by the United States Department of Energy. It is located at the National Energy Research Scientific Computing Center (NERSC). Perlmutter uses AMD 7763 64-core 2.45GHz CPUs and Nvidia A100 GPUs. It has a total of 761,856 cores and clocks at a speed of 70.87 PFLOPS. Its power efficiency rating is 27.37 gigaflops/watt.
8. Selene (United States)
Selene was built by Nvidia in 2020. It is named after the Greek goddess of the moon. Selene’s most prominent purpose was realized in 2020 when Argonne National Laboratory used it to study the coronavirus and a potential cure. Selene is also used to train Google’s natural language processing model BERT.
Selene is a strong computer with 555,520 cores. It has a good power efficiency rating of 24 gigaflops/watt and scores 63.46 PFLOPS on benchmarks. Selene is built on the NVIDIA DGX SuperPOD reference architecture and runs on AMD 7742 64-core 2.25GHz processors.