NASA Hacks Its Supercomputing Way Through Intel, AMD Parts
NASA today announced that it has run the most powerful supercomputer ever. A supercomputer that can meet all of your current needs, both outside and outside the stratosphere. Just their new supercomputer, Aitken, Not so new (Opens in a new tab).. But, as other great enthusiasts say, spreading upgrades is a way to go, even among well-known rivals such as Intel and AMD.
NASA takes a different approach than most supercomputing clients. Register Cover Instead of ordering a monolithic, self-contained day system like the Exascale class frontier, NASA takes a modular approach similar to modern European supercomputing designs such as LUMI. In addition, a network of additional compute nodes allows you to split costs across your budget and extend the operational life that can be squeezed out of these systems.
“This significant enhancement-a 16% performance improvement over the last expansion and a 49% improvement over last year when the system was ranked 72nd on the Top500 list in June 2021-solves a bigger problem and is important. Means to solve the results faster. NASA’s research projects in aeronautics, space exploration, geosciences, and astronomical physics. ” Michelle Moyer of NASA Ames Research Center said.
NASA’s approach has already been established. The company has learned lessons from this modular approach for over 14 years since it first powered up in 2008 and upgraded its “Pleiades” supercomputer.
And as a testament to how long AMD’s High Performance Computing (HPC) chops lasted, four HPE “Apollo” racks based on AMD’s “Rome” architecture were added to the heart of the system’s cascade lake base. But, of course, when the supercomputer adds another architecture to the portfolio, you can see how much the table has changed.
Aitken’s (Opens in a new tab) The four added racks may sound easy, but keep in mind that the computational density is increasing at an amazing pace. These are enough to power the 308,000 AMD Zen 2, 7nm “Rome” cores distributed across 512 compute nodes. The initial design of the system featured 46,080 Intel Cascade Lake cores (1,152 40-core nodes distributed across four E-Cell systems) provided by HPE.
Due to its high density, the supercomputer fits in a junior high school sporting goods hut and occupies only one acre of land. Nevertheless, the new addition boosts the system’s theoretical peak performance to 13.12 petaflops per second, with 1.27 petabytes of memory supporting all processing growls.
Is this improvement enough to be called the Big Bang of NASA’s supercomputing capabilities? please tell me.