{
  "id": 14723115,
  "title": "When faster supercomputers aren’t enough",
  "url": "https://urgent.news/2026/10/15/when-faster-supercomputers-arent-enough-14723115",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-15T15:00:00.000Z",
  "source": {
    "name": "The Register",
    "slug": "the-register",
    "url": "https://www.theregister.com/hpc/2026/10/15/partner-content-when-faster-supercomputers-arent-enough/5302562"
  },
  "original_language": "en",
  "account": "When veterans contributed their genetic data to the Department of Veterans Affairs' Million Veteran Program, they envisioned no return from their contribution. Ravi Madduri, a senior scientist at Argonne National Laboratory, acknowledges their selfless act: \"Even after giving everything to the country, they found it in their hearts to give more without expecting anything in return.\" Madduri, who also teaches at the University of Chicago, is committed to fulfilling the original intentions behind the data sharing. He believes others can benefit greatly from the data's potential discoveries.\n\nMadduri has devoted two decades of his career to merging supercomputing with biomedicine, focusing on transforming veteran data into actionable insights for issues ranging from post-traumatic stress disorder to heart disease and cancer. While the supercomputers have generally met their responsibilities, Madduri emphasizes that the hurdles between valuable data and meaningful results are not always resolved by faster machines. Understanding these challenges is crucial for advancing scientific research.\n\nOne major issue Madduri and his team encountered was the inability of the code to handle the vast amounts of genetic data. Any two human genomes differ at millions of specific positions. A genome-wide association study aims to uncover simple links between genetic variants and medical conditions. However, this straightforward question, when considered across all the genetic variants and conditions in a patient's records, becomes an enormous mathematical challenge. The more complex problem lies in determining whether the programming is equipped to handle this.\n\nMadduri's team ran this problem on two supercomputers: Summit at Oak Ridge Leadership Computing Facility and Aurora at Argonne Leadership Computing Facility. On Summit, the team discovered the job would not complete because the data matrix exceeded the memory capacity, causing the analysis to stall. Madduri and his team realized the issue was not merely a speed problem but a coding one. The analysis was initially designed for regular processors, which were not powerful enough to manage the immense computations.\n\nThe solution lay in rewriting the code for graphics processors (GPUs), which are specifically designed to perform calculations involving large grids of numbers – operations that are prevalent in genetic data analysis. By adapting the code for GPUs, a task that would have taken approximately 481,000 CPU-core hours in its original form became feasible. This adaptation is detailed in a separate paper in Bioinformatics, where the team presents the revised version as SAIGE-GPU.",
  "summary": "PARTNER CONTENT: A million reasons why advanced computing matters for genomics",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "The Register Science",
        "title": "When faster supercomputers aren’t enough",
        "url": "https://urgent.news/2026/10/15/when-faster-supercomputers-arent-enough",
        "published": "2026-10-15T15:00:00.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}