Here’s the translation of the provided text into American English:
The U.S. Department of Energy (DOE) and AMD have announced that the supercomputer El Capitan, located at the Lawrence Livermore National Laboratory (LLNL), has reached the top spot on the Top500 list of the world’s fastest supercomputers. With a performance of 1.742 exaflopsAn exaflop is a unit of measurement in supercomputing that …, El Capitan significantly outperforms its predecessor, Frontier, with 45% more power, solidifying its position as a benchmark in the era of exascale supercomputers.
An Unprecedented Achievement in Exascale Computing
El Capitan is a marvel of engineering, featuring a total of 11,039,616 computing cores, powered by 44,544 AMD MI300A processors. These Accelerated Processing Units (APUs) combine CPU cores based on Zen 4 architecture with CDNA 3 GPU cores and high-capacity HBM3 memory, totaling 5.4 petabytes of main memory. Each chip contains 146 billion transistors, making it the largest chip ever produced by AMD.
The system is designed to achieve a theoretical peak performance (Rpeak) of 2.746 exaflops, though its actual measurement under the High-Performance Linpack (HPL) standard is 1.742 exaflops, which equates to performing 1.742 quintillion calculations per second.
El Capitan also ranks as an efficiency model, occupying the 18th spot on the Green500 list, which ranks supercomputers based on their energy efficiency, with a consumption of 35 megawatts under full operation.
Critical Applications for National Security and Research
El Capitan is not only a technological feat but also a crucial tool for the DOE. Its primary mission is to modernize the United States’ nuclear arsenal through simulations of nuclear explosions, eliminating the need for underground testing. Additionally, it will allow for the assessment of aging, safety, and reliability of the nuclear stockpile, as well as the development of new designs for intercontinental ballistic missiles (ICBMs).
The system will also be used for high-performance computing (HPC) workloads and artificial intelligence (AI) applications, enabling an innovative merging of both fields to solve complex problems in science and technology.
A Revolutionary Architecture
El Capitan was built by HPE using the Shasta architecture, which employs liquid-cooled high-density cabinets and a Slingshot-11 interconnection system. This design also powers the DOE’s other two exascale supercomputers: Frontier, the second fastest in the world, and Aurora, which ranks third.
What sets El Capitan apart is its focus on efficiency and integration. Its coherent cache memory architecture minimizes data movement between CPUs and GPUs, reducing energy consumption and improving latency. Additionally, it simplifies the portability of existing code and the development of new applications, making it ideal for demanding workloads.
Comparison with Other Giants on the Top500 List
In the latest ranking of the Top500 list, the top three positions are dominated by systems built by HPE, highlighting their leadership in the supercomputing industry:
- El Capitan (AMD-powered): 1.742 exaflops (Rmax), 2.746 exaflops (Rpeak), 29.581 kW.
- Frontier (AMD-powered): 1.353 exaflops (Rmax), 2.055 exaflops (Rpeak), 22.786 kW.
- Aurora (Intel-powered): 1.012 exaflops (Rmax), 1.980 exaflops (Rpeak), 38.698 kW.
While Aurora remains the fastest AI supercomputer in the world, with 10.6 exaflops on mixed precision workloads (HPL-MxP), its lack of a new measurement for the Top500 suggests ongoing operational issues.
China and Global Competition
Despite El Capitan’s leadership, reports suggest that China possesses several exascale systems of its own, although these have not been included in the Top500 list due to fears of U.S. sanctions. This underscores the global competitiveness in the field of supercomputing and the growing relevance of these systems for technological and strategic leadership.
The Future of Supercomputing
El Capitan is not only the fastest supercomputer in the world but also one of the most cost-effective in its class, according to LLNL. This achievement reflects AMD and HPE’s commitment to advancing exascale computing and lays the groundwork for new innovations in science, artificial intelligence, and national security.
With the official unveiling of the system and its integration into classified networks expected in 2025, El Capitan represents a significant step toward the future of supercomputing and its ability to tackle the most complex challenges of the 21st century.