There is a lot of talks everywhere about the best personal computers, whether desktop or laptop and almost we can say that most of us have a clear idea about this topic and some of the names of the computers hesitated on our ears to become part of our ordinary vocabulary, but what for supercomputers those we can hear about Little, and what are the most powerful supercomputers in the world? Do we use the same modules that we normally use when computing PC specifications? In other words, are gigabytes and terabytes sufficient to describe such supercomputers?
As in the world of personal computers whose development has not slowed down over time and the race between producers has not ceased, so do supercomputers on a parallel line that we are witnessing their development and the race between producers at its peak.
First, let’s explain some of the concepts before going into comparison.
Supercomputers and a glimpse of their evolution
One supercomputer is thousands of processors connected to perform large-scale work that personal computers cannot do, such as climate research, weather forecasting, oil and gas exploration research, and other important and large-scale work that requires a huge amount of calculations to complete.
The first supercomputer was made in 1964 and called CDC 6600, my friend imagined that it was only a single processor performing 3 million calculations per second, thousands of times less than the processing capacity of an iPhone mobile.
But that did not last and began racing organizations and companies to innovation, production and improve the performance of the least problems, was one of the race criteria is the number of processing operations per second and called floating operations interval in the second (floating-point operations per second ( FLOPS, in addition to many other standards The evolution of the transistors, CPUs, and cooling industries, of course, will be accompanied by an increase in heat dissemination, a problem that needs to be solved.
This race track has seen the participation of companies of different nationalities during the different years since the 1960s, some of which have been delayed and some of them retreated. According to the Top500 website.
Because of the high cost of energy and time to operate such computers – your laptop with its modest capabilities, it may take several minutes to boot and imagine it for a supercomputer – companies have changed the industry and turned supercomputers into Nodes that take off in parallel and eventually work together as a single unit.
It is important to note that the theoretical peak performance of the supercomputer is theoretically the maximum performance that Rpeak stands for.
Top 10 supercomputers in the world
Summit – America
After China continued to hold the lead for several years, America came to this supercomputer to come back and top the list of the best and fastest computer in the world, where he achieved a record for the peak of theoretical performance Rpeak The number of processes processed per second FLOP up to 200 petaFlops-beta equivalent to 1000 terabytes – or the equivalent of 200,000 trillion – one trillion equivalent to 1000 billion – a different process per second, using processors with 2.4 million nuclei.
On the other hand, for mixed and interdependent operations, it can accomplish 3 billion – or 3.3 billion Exa Operation Per Second. Today of the largest research is carried out which needs a huge volume of operations.
The computer is powered by an IBM POWER9 22C 3.07GHz processor (3.07 GHz frequency, 22 kernels), and 6 Nvidia Tesla V100 graphics cards. His predecessor ranked TaihuLight where he was deploying power by 15 MW.
Sierra – America
Serra is second only to boosting America’s lead. Speak achieved more than 125 petaflops using 1.6 million kernels and 1.4 MB of memory with the IBM POWER9 22C 3.1GHz processor, deploying 7MW of power.
Sunway TaihuLight – China
China has been on the top of the list of supercomputers for several years thanks to this computer, which can provide performance at the top of the Rpeak up to 125 petaflops using 10.6 million kernels, and memory exceeds 1 Petabyte, processor Sunway SW26010 260C 1.45GHz, but it differs very much compared to previous computers To 15 MW.
Tianhe-2A – China
Also known as MILKYWAY-2 was the fastest in 2013, with a Rpeak performance of 33.86 Petaflops using about 5 million kernels in 16,000 node computers made from Intel Xeon E5-2692v2 12C 2.2GHz processors and memory up to 2 Petabytes, but it spreads heat exceeding 18 MW. Chinese supercomputers seem to be overheated.
Piz Daint – Switzerland
The Swiss computer in this competition on the scale of Rpeak up to 27 petaflops using 387 thousand kernels – this is very few for others – and memory exceeding 365 terabytes – terabytes equivalent to 1000 GB – and processor Xeon E5-2690v3 12C 2.6GHz, and deploy power exceeds 2 megawatts – Lowest power with the least potential yet.
Trinity – America
America is still in the top ten, and this supercomputer was designed for the Los Alamos National Lab for near-site nuclear tests, with Rpeak performance of up to 41 petaflops using 979,000 kernels and an Intel Xeon Phi 7250 processor 68C 1.4GHz, diffuses power by 7 MW.
ABCI – Japan
This computer appears to be Japan’s only contribution to the race, ranking seventh, with Rpeak performance exceeding 32 petaflops using 391 thousand cores with 417TB of memory and Xeon Gold 6148 20C 2.4GHz processor and spreading heat up to 1 MW.
SuperMUC-NG – Germany
Germany has entered the racetrack with this supercomputer, delivering Rpeak performance of up to 27 petaflops using 305,000 kernels, more than 718 terabytes of memory and a Xeon Platinum 8174 24C 3.1GHz processor, using the warm-water cooling system that the designer claims Lenovo will reduce electricity consumption by 45% compared to conventional air cooling systems.
Titan – America
It is the first supercomputer that combines AMD processors with NVIDIA Tesla processors to achieve its highest performance 27 petaflops using 560 thousand cores and memory size up to 710 terabytes, and processor type Opteron 6274 16C 2.2GHz, consuming the power of 8 MW.
Sequoia – America
Built to measure the dangers of nuclear war, 98,304 computer nodes with 1.6 million nuclei deliver Rpeak performance exceeding 20 petaflops, 1 bit more memory, and a Power BQC 16C 1.6GHz processor, consuming more than 7 MW of power.