Science & Technology


Since 2013, China has led the world in developing the most powerful supercomputers. Now America is back on top again. Engineers at the US Department of Energy’s Oak Ridge National Lab in Tennessee have just unveiled Summit, a supercomputer with enough processing power to surpass the current record holder, China’s TaihuLight.
At peak performance, Summit is capable, of 200 “petaflops” which is equivalent to a mindblowing 200 million billion calculations a second. To put that in some sort of context, everyone on earth
would have to do a calculation every  second of every day for 305 days to achieve what the new machine can do in the blink of an eye. Summit is 60 percent faster than TaihuLight and almost eight times as fast as Titan” which is also housed at Oak Ridge and held the U.S. supercomputing speed record until Summit’s arrival. Of course, it’s not just about national pride. Supercomputers are already being used to design new aircraft and create new materials. They are used by the military to design weapons and by scientists to conduct research. If the most powerful one is in
the US, American researchers and the country’s armed forces will have an extra edge over their Chinese counterparts. The team at Oak Ridge says Summit is the first supercomputer designed
from the ground up to run AI applications, such as machine learning and neural networks. It has over 27,000 GPU chips from Nvidia, and also includes some of IBM’s Power chips, which the company launched last year specifically for AI workloads. There’s also an ultrafast communications link for transferring data between these machines. Among the AI-related projects that will run on the new supercomputer is one that will process huge volumes of medical reports and images to try to identify possible relationships between genes and cancer. Another will try to identify
genetic traits that could predispose people to opioid addiction and other afflictions. Summit is an important stepping stone to the computing breakthrough: machines capable of an “exaflop”, or
a billion billion calculations a second. The US is looking to build several of these machines, which could cost between $400 million and $600 million each. The aim is to get one or more of these “exascale” computers up and running between 2021 and 2023.

Short link :

Related Stories

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button