EXECUTIVE SUMMARY:
A Top500 list of supercomputers is produced twice annually by a team of scientists in Germany. For the second year in a row, the top ranked supercomputer is Japan’s Fugaka supercomputer, achieving a world record of 422 petaflops. A petaflop refers to a computer’s capacity to do one quadrillion floating point operations per second.
For those in tech, Nvidia technology powers nearly 70% of supercomputers. This is largely due to the success of its end-to-end HGX AI supercomputing platform, which has improved data analysis and AI workloads. The company is currently pursuing exascale computing developments.
This year’s supercomputer slow down:
The growth curve for modern supercomputers is flattening. While two new systems made the top 10 list, this year’s larger list reflected the smallest number of new entrants since the project’s inception. The decline is attributed to the coronavirus pandemic, which drove investments in commercial high-performance computing systems to an all-time low.
A decline in commercial progress is also noticeable as it relates to device performance. Every 11 years, device performance used to multiply by 1,000. But based on today’s growth rates, it will take 20 years to achieve former levels of growth.
What are the implications when it comes to these extraordinarily powerful computers?
Supercomputers are expected to accelerate developments in precision medicine, to enable scientists to conduct more accurate regional climate simulations, to increase our ability to explore the core forces of the universe, to transform understandings of how to create biofuels, to help evolve higher education and more.
The new supercomputer in the labs of New York University (NYU):
To increase its capacity for cutting-edge research, NYU has invested in a supercomputer built by Lenovo. “High-performance computing, big data and artificial intelligence are critical in so many areas of research throughout higher education and in particular here at NYU,” says Andrew Hamilton, New York University’s President.
For more on this story, visit ZDNet.
