With the launch of Fugaku, the next generation of super-powerful supercomputers is here – the ramifications would be extensive.
With the world of AI and data science becoming more and more computationally intensive, the need for greater computing power for carrying out high-level simulations is peaking. The new Japanese-engineered Fugaku supercomputer may just turn out to be the benchmark for ensuing solutions.
The First and Last of the Supers
Although the term ‘supercomputing’ was first introduced to refer to the massive custom-built tabulators designed for Columbia University by IBM in the 1920s, the first commercially successful transistorised computer (and inarguably the fastest at the time) was the CDC 1604, dating back to only the early 1960s. Yet, the world’s first recognised supercomputer was, in fact, an eventual iteration of its first model; the CDC 6600: working at a then-mind-boggling top speed of 40MHz, or 3 million FLoating-point Operations Per Second (MegaFLOPS).
Today, however, such speeds are considered painstakingly slow. For comparison’s sake: the base model of Apple’s newly-launched iPhone 12 works at about 2.99 GHz — about 75 times faster than the world’s first supercomputer. In fact, ever since the inception of the first supercomputer in 1964, the global race to strengthen computing power has become exceptionally tight. Interestingly, the record for the world’s fastest supercomputer was set only in June 2020 at an astonishing 1.4 ExaFLOPS — until, of course, the record was shattered by the very same makers very recently.
Super Needs, Super Power
Originally envisioned in 2014 as a project collaboration, Japanese computing firm Fujitsu and Research Institute RIKEN have now successfully combined to produce for commercial use the world’s fastest supercomputer — the Fugaku. The Fugaku has been developed as a core system for the innovative High-Performance Computing Infrastructure (HPCI) promoted by Japan’s Ministry of Education, Culture, Sports, Science and Technology (MEXT).Topping out at an outrageous 2 ExaFLOPS, it currently stands as the most powerful supercomputer in the world — around 1.42 times faster than the previous fastest.
According to TechRepublic: “Fugaku has controlled the top spot on the list of fastest supercomputers since mid-2020, and it continues to hold that spot in the most recent Top 500 list published in November (2020). With three times the computing power of runner-up IBM Summit, Fugaku is likely to remain the supercomputer to beat until the LUMI is completed in Finland.”
Boasting exceptional computing power, Fugaku allows the running of various kinds of ‘high-resolution’, ‘large-scale’, ‘multiple-case’ and ‘long-duration’ simulations and is expected to play a key role in “solving familiar social problems to helping improve understanding of fundamental science.” In fact, given the magnitude of the machine, RIKEN President Hiroshi Matsumoto has even opined: “Fugaku is a key national technology, and we will manage it responsibly with the goal to achieve research results that will help build a long-lived and healthy society, disaster mitigation, and better energy use.”
According to a Bloomberg press release: “The next generation of deep learning-based AI is computationally intensive and will require large-scale supercomputing resources. Fugaku’s CPU possesses high performance in “convolution operations,” which are central to deep learning, and is connected by a network with superior communication performance, and it will be leveraged for AI and data science research.”
There are several areas where Fugaku can potentially impact, including: (i) accelerated drug development through high-speed high-precision simulations; (ii) early disease detection through medical big data analysis and biological simulations; (iii) weather and disaster predictions, including tornadoes, earthquakes, tsunamis; (iv) development of low-cost organic solar batteries and high-efficiency fuel cells; (v) accelerated device and material creation for next-generation industries; (vi) increased insight into fundamental questions of theoretical science, such as when the universe was created, and similar conjectural processing.