Locator: 48726NVIDIA.
Supercomputers are tracked here.
Reuters, link here.
From one of the stories:
A new supercomputer named after a winner of the Nobel Prize in chemistry will help power artificial intelligence technology and scientific discoveries from the Lawrence Berkeley National Laboratory, federal officials said Thursday.
U.S. Energy Secretary Chris Wright announced the project Thursday alongside executives from computer maker Dell Technologies and Nvidia CEO Jensen Huang.
The new computing system at the Berkeley Lab will be called Doudna after Berkeley professor and biochemist Jennifer Doudna, who won a Nobel in 2020 for her work on the gene-editing technology CRISPR. It’s due to switch on next year.
“One of the key use cases will be genomics research,” said Dion Harris, a product executive in Nvidia’s AI and high-performance computing division, in an interview. “It was basically just a nod to her contributions to the field.”
Dell is contracted with the energy department to build the computer, the latest to be housed at Berkeley Lab’s National Energy Research Scientific Computing Center. Previous computers there have been named after other Nobel winners: Saul Perlmutter, an astrophysicist, and Gerty Cori, a biochemist.
From ChatGPT:
The upcoming “Doudna” supercomputer, named after Nobel laureate Jennifer Doudna for her pioneering work on CRISPR gene-editing, is set to be deployed in 2026 at the Lawrence Berkeley National Laboratory’s National Energy Research Scientific Computing Center (NERSC) in California. This advanced system is a collaborative effort between the U.S. Department of Energy, Nvidia, and Dell Technologies.
Key Specifications and Features:
Hardware Components:
Equipped with Nvidia’s latest “Vera Rubin” chips, designed to enhance performance in artificial intelligence (AI) and high-performance computing (HPC) tasks.
Integrated into Dell’s liquid-cooled servers, ensuring efficient thermal management and energy consumption.
Research and Application Focus:
Aims to support a wide array of scientific research areas, including genomics, chemistry, physics, and biology.
Will serve approximately 11,000 researchers, facilitating advancements in both fundamental science and national security applications, such as the maintenance of the U.S. nuclear arsenal.
Naming Tradition:
Continues NERSC’s tradition of naming supercomputers after Nobel laureates, following predecessors like “Perlmutter” and “Cori.”
While specific performance metrics, such as processing power and storage capacity, have not been publicly disclosed, the integration of cutting-edge Nvidia chips and Dell’s advanced server technology suggests that Doudna will be among the leading supercomputers upon its deployment.
For comparison, NERSC’s current supercomputer, Perlmutter, features a heterogeneous architecture with Nvidia A100 GPUs and AMD Milan CPUs, delivering significant computational capabilities for a variety of scientific applications.
The Doudna supercomputer is anticipated to play a pivotal role in accelerating scientific discoveries and technological innovations across multiple disciplines.
Cost? From ChatGPT, again:
While the exact cost remains unspecified, we can look to previous NERSC supercomputers for context. For instance, the Perlmutter supercomputer, delivered in 2020, had a total contract value of $146 million, which included multiple years of service and support. Given the advancements in technology and the scale of the Doudna project, it’s reasonable to anticipate that its cost could be in a similar range or potentially higher.
My hunch: 50% higher --> $220 million.
Top 500 supercomputers, link here.
From wiki, Perlmutter supercomputer:
Perlmutter (also known as NERSC-9) is a supercomputer delivered to the National Energy Research Scientific Computing Center of the United States Department of Energy as the successor to Cori.
It is being built by Cray and is based on their Shasta architecture which utilizes Zen 3 based AMD Epyc CPUs ("Milan") and Nvidia Tesla GPUs.
Its intended use-cases are nuclear fusion simulations, climate projections, and material and biological research. Phase 1, completed May 27, 2022, eached 70.9 PFLOPS of processing power.
It is named in honor of Nobel prize winner Saul Perlmutter.
