"** What is the fastest computer in the world?"
1.9s
Add to Favorites
Cozy Upvote
Share
Export
The Fastest Computer in the World (June 2025)
Supercomputer: El Capitan
Operator: Lawrence Livermore National Laboratory (LLNL), United States
Peak Performance: 1.742 exaFLOPS (1 742 petaFLOPS) on the LINPACK benchmark — ≈ 1.742 × 10¹⁸ floating‑point operations per second12
Rank: #1 on the latest TOP 500 list (June 2025)1
Generation: Third‑ever exascale machine (the first built by a U.S. national lab)1
Why El Capitan Holds the Top Spot
| Aspect | Details |
|---|---|
| Location & Facility | Housed in the Livermore Computing Complex at LLNL, California, USA3 |
| Hardware Highlights | Over 7 million CPU cores, AMD EPYC 9004 series processors, and NVIDIA H100 GPUs, interconnected by a high‑speed InfiniBand fabric4 |
| Software Stack | Runs Red Hat Enterprise Linux with a custom exascale‑optimized runtime and scheduling system48 |
| Primary Uses | • National security – nuclear stockpile stewardship • Scientific research – climate modeling, astrophysics, materials science • Advanced simulations for energy, defense, and AI workloads49 |
| Performance Milestones | • First U.S. supercomputer to breach the exascale barrier in production • Sustained LINPACK performance of 1.742 exaFLOPS, surpassing the previous leader by ~15%12 |
How the Ranking Is Determined
- TOP 500 publishes a bi‑annual list of the world’s most powerful supercomputers, ranking them by their LINPACK benchmark score—a standard test of floating‑point performance1.
- The June 2025 edition (the 66th edition) confirmed El Capitan as the clear leader, and the November 2025 update reaffirmed its position5.
What This Means for the Future
El Capitan’s success demonstrates that exascale computing is now a practical reality, paving the way for:
- Faster, more detailed scientific simulations that can address climate change, fusion energy, and biomedical research.
- Development of next‑generation AI models that require massive parallel processing power.
- Continued innovation in hardware‑software co‑design to push performance beyond the exascale barrier.
**
**