(Pocket-lint) - Supercomputers, the behemoths of the tech world and inventions by man often to put to specific use to solve incredible problems mere mortals couldn't fathom alone.
From studying the decay of nuclear materials to predicting the path of our planet due to global warming and everything in between, these machines do the processing and crunch the numbers. Calculating in moments what it would take mere mortals decades or more to decipher.
- Star Trek in real life: Best Starfleet gadgets and toys you can buy
- Iconic gadgets of the 90s: Amazing gadgets and gizmos from yesteryear
Earth Simulator was the world's fastest supercomputer between 2002 and 2004. It was created in Japan, as part of the country's "Earth Simulator Project" which was intended to model the effects of global warming on our planet.
The original Earth Simulator supercomputer cost the government 60 billion yen but was a seriously impressive piece of technology for the time, with 5120 processors and 10 terabytes of memory.
It was later replaced by Earth Simulator 2 in 2009 and Earth Simulator 3 in 2015.
Blue Gene / P From Argonne National Laboratory
IBM Blue Gene
The original Earth Simulator supercomputer was surpassed in performance by IBM's Blue Gene/L prototype in 2004.
Blue Gene was designed to reach petaFLOP operating speeds while maintaining low power consumption. As a result, the various Blue Gene systems have been ranked as some of the most powerful and most power-efficient supercomputers in the world.
The Blue Gene supercomputers were so named because they were designed to help analyse and understand protein folding and gene development. They were most well-known for power and performance though, reaching 596 TFLOPS peak performance. They were then outclassed by IBM's Cell-based Roadrunner system in 2008.
U.S. Army Photo
ENIAC (Electronic Numerical Integrator And Computer)
ENIAC was one of the very first supercomputers. It was originally designed by the US Army to calculate artillery firing tables and even to study the possibility of thermonuclear weapons. It was said to be able to calculate in just 30 seconds what it would take a person 20 hours to do.
This supercomputer cost around $500,000 to build (over $6 million in today's money).
Notably, the Electronic Numerical Integrator and Computer was later used to compute 2,037 digits of Pi and it was the first computer to do so. Even that computation took 70 hours to complete.
SUNWAY TAIHULIGHT SYSTEM REPORT
In 2018, the Chinese supercomputer known as Sunway TaihuLight was listed as the third-fastest supercomputer in the world. This system sported nearly 41,000 processors, each of which had 256 processing cores, meaning a total of over 10 million cores.
This supercomputer was also known to be able to carry out an eye-watering 93 quadrillion calculations per second. IT was designed for all sorts of research from weather forecasting to industrial design, life sciences and everything in between.
The difference engine
The Difference Engine was crafted by Charles Babbage in 1822. This was essentially the first computer or at least one of them. It could be used to calculate mathematical functions but unfortunately cost an astronomical amount for the time.
This machine was impressive for what it could do but also for the machines it inspired in the years and decades that followed.
IBM's Roadrunner supercomputer was a $100 million system built at the Los Alamos National Laboratory in New Mexico, USA.
In 2008, it managed to become one of the fastest supercomputers on the planet, reaching a top performance of 1.456 petaFLOPS.
Despite taking up 296 server racks and covering 6,000 square feet, Roadrunner still managed to be the fourth-most energy-efficient supercomputer at the time.
The system was used in order to analyse the decay of US nuclear weapons and examine whether the nuclear materials would be safe in the following years.
Summit is one of the most recent and most powerful supercomputers built by man. Another incredible system built by IBM, this time used at Oak Ridge National Laboratory and sponsored by the U.S. Department of Energy.
Between 2018 and June 2020, Summit (also known as OLCF-4) achieved the record of being the fastest supercomputer in the world, reaching benchmark scores of 148.6 petaFLOPS. Summit was also the first supercomputer to hit exaflop (a quintillion operations per second) speeds.
Summit boasts 9,216 22-core CPUs and 27,648 Nvidia Tesla V100 GPUs which have been put to work in all manner of complex research from Earthquake Simulation to Extreme Weather simulation as well as predicting the lifetime of Neutrinos in physics.
U.S. Department of Energy
The Sierra is another supercomputer developed by IBM for the US Government. Like Summit, Sierra packs some serious power, with 1,572,480 processing cores and a peak performance of 125 petaFLOPS.
As with IBM Roadrunner, this supercomputer is used to manage the stockpile of US nuclear weapons to assure the safety of those weapons.
Tianhe-2 is another powerful supercomputer built by the Chinese. It's located at the National Supercomputer Center in Guangzhou, China and cost a staggering 2.4 billion Yuan (US$390 million) to build.
It took a team of 1,300 people to create and their hard work paid off when Tianhe-2 was recognised as the world's fastest supercomputer between 2013 and 2015.
The system sports nearly five million processor cores and 1,375 TiBs of memory, making it able to carry out over 33 quadrillion calculations per second.
The CDC 6600 was built in 1964 for $2,370,000. This machine is thought to be the worlds first supercomputer, managing three megaFLOPS, three times the speed of the previous record holder.
At the time, this system was so successful that it became a "must-have" for those carrying out high-end research and as a result over 100 of them were built.
The Cray-1 came almost a decade after the CDC 6600, but quickly became one of the most successful supercomputers of the time. This was thanks to its unique design that not only included an unusual shape but also the first implementation of a vector processor design.
This was a supercomputer system that sported 64-bit processor running at 80 MHz with 8 megabytes of RAM which make it capable of a peak performance of 250 megaflops. A significant move forward compared to the CDC 6600 which came a mere decade before.
The Frontera supercomputer is the fastest university supercomputer in the world. In 2019, it achieved 23.5 PetaFLOPS making it able to calculate in a mere second what it would take an average person a billion years to do manually.
The system was designed to help teams at the University of Texas to solve massively difficult problems including everything from molecular dynamics to climate simulations and cancer studies too.
Los Alamos National Laboratory
Trinity is yet another supercomputer designed to analyse the effectiveness of nuclear weapons.
With 979,072 processing cores and 20.2 petaFLOPS of performance power, it's able to simulate all manner of data to ensure the country's stockpile of weapons is safe.
In 2019, IBM built Pangea III, a system purported to be the world's most powerful commercial supercomputer. It was designed for Total, a global energy company with operations worldwide.
Pangea III was an AI-optimised supercomputer with a high-performance structure but one that was said to be significantly more power-efficient than previous models.
The system was designed to support seismic data acquisition by geoscientists to establish the location of oil and gas resources. Pangea III has a computing power of 25 petaflops (roughly the same as 130,000 laptops) and ranked 11th in the leaderboards of the top supercomputers at the time.
The Connection Machine 5
The Connection Machine 5 is interesting for a number of reasons, not simply because it's a marvellous looking supercomputer but also because it's likely the only system on our list to be featured in a Hollywood Blockbuster. That's right, this supercomputer appeared on the set of Jurassic Park, where it masqueraded as the Park's central control computer.
The Connection Machine 5 was announced in 1991 and later declared the fastest computer in the world in 1993. It ran 1024 cores with peak performance of 131.0 GFLOPS.
It's also said to have been used by the National Security Agency back in its early years.
HPC4 is a Spanish supercomputer that's particularly well-known for being energy efficient while still sporting some serious processing power that includes 253,600 processor cores and 304,320GB of memory.
In 2018, the updated HPC5 supercomputer was combined with HPC4 to result in 70 petaFlops of combined computational capacity. That means this system is capable of performing 70 million billion mathematical operations in a single second.
Selene is Nvidia's supercomputer built on the DGX SuperPOD architecture. This is an Nvidia-powered supercomputer sporting 2,240 NVIDIA A100 GPUs, 560 CPUs and an impressive record that includes being the second most power-efficient supercomputer around.
Selene is particularly impressive when you discover that it was built in just three weeks. We also like that it has its own robot attendant and is able to communicate with human operators via Slack.
Perlmutter was revealed in the middle of 2021. A powerful GPU-accelerated supercomputer built by HPE in partnership with Nvidia and AMD. This supercomputer uses the power of 6,159 Nvidia A100 GPUs and 1,500 AMD Milan CPUs to analyse the effect of dark energy on the universe’s expansion.
The supercomputer is essentially helping to map the visible universe and is able to process the data provided by Dark Energy Spectroscopic Instrument. This map will span 11 billion light-years and give an impressive view of our universe.
Writing by Adrian Willings. Originally published on 24 August 2020.