The world's most powerful supercomputer is currently El Capitan with a processing speed of 1.72 ExaFlop/s. It has been used to predict tsunamis in 0.2 seconds to provide early predictions of affected areas. It is the result of a collaboration between AMD, HPE and Lawrence Livermore National Laboratory.
Today, AMD, HPE, Oak Ridge National Laboratory and the US Department of Energy (DOE) announced the construction of the Discovery and Lux supercomputers that will increase the capabilities of high-performance computers in the United States for scientific research purposes.
Discovery is built on the HPE Cray Supercomputing GX5000 system, which was developed specifically for the era of artificial intelligence (AI) supercomputers, using AMD EPYC chips codenamed Venice and AMD Instinct MI430X GPUs.
When it begins operating in 2029, Discovery will be used to study how nuclear energy production is safer, cheaper and more accessible to all. The AI system can also produce digital twins to treat various diseases such as cancer. In the aerospace world, Discovery can shorten the time to produce fuel-efficient aircraft designs.
Lux is an AI cluster developed to solve problems such as nuclear fusion, nuclear fission, materials science and various other scientific studies under the DOE. It is equipped with AMD Instinct MI355X GPUs and AMD EPYC CPUs. This supercomputer will begin operating in 2026.
The Discovery and Lux supercomputers will cost $1 billion (~RM 4.2 billion) to build.
