Overview
- Name: El Capitan
- Location: Lawrence Livermore National Laboratory (LLNL), California
- Cost: $600 million
- Operational Status:
- Construction began: May 2023
- Fully online: November 2024
- Officially dedicated: January 9, 2025
Purpose & Applications
- Primary Use:
- Securing the U.S. nuclear stockpile (replacing underground testing banned in 1992).
- Classified research: weapon design, material discovery, high-energy-density physics, and nuclear data.
- Significance: Supports national security by simulating nuclear weapon performance without physical testing.
Performance & Specifications
- Speed:
- Peak performance: 2.746 exaFLOPS (1 exaFLOP = 1 quintillion calculations per second).
- Benchmark score: 1.742 exaFLOPS (High-Performance Linpack test).
- Global Ranking:
- Third exascale supercomputer worldwide.
- Surpasses Frontier (Oak Ridge National Lab, 1.353 exaFLOPS standard, 2.056 peak).
- Hardware:
- Cores: Over 11 million processing/graphics cores.
- Processing Units: 44,544 AMD MI300A APUs (combining AMD EPYC Genoa CPUs + CDNA3 GPUs).
- Memory: 128 GB high-bandwidth memory (HBM) per APU (shared between CPU/GPU).
Context & Commissioning
- Program: Built under the U.S. Department of Energy’s CORAL-2 initiative.
- Replacement: Succeeds the Sierra supercomputer (2018, ranked 14th in Top500).
Technical Comparisons
- vs. Laptops: Best laptops achieve ~hundreds of gigaFLOPS (1 billion FLOPS).
- vs. Global Competition:
- Frontier (U.S.): Current #2 supercomputer.
- Japan’s Zeta-class: Planned for 2025, aims to be 1,000x faster than today’s top machines.
Implications & Future
- Scientific Impact: Enables breakthroughs in physics, materials science, and weapon safety simulations.
- Ethical Concerns: Raises questions about AI’s role in military applications and classified research.
- Global Race: Highlights U.S. efforts to lead in exascale computing for national security and technological dominance.
Source: Space.com, LLNL, Next Platform
