# A weather-model test ran faster on NOAA’s new research computer.

NOAA EPIC reports that one named Unified Forecast System regression test ran about 30–40% faster on Ursa than on Hera under the Intel compiler—a development result, not evidence of a more accurate forecast.

Canonical: https://brightaifuture.com/discoveries/noaa-ursa-weather-testing
Format: discovery
Source publication: 2025-08-22
Bright publication: 2026-09-19
Substantive update: None recorded
Evidence and review: Demonstrated; confidence: medium; approved; ai-assisted. AI-assisted editorial comparison with the cited original and supporting public sources, bounded claims and explicit status labels. Bright did not independently audit the underlying records.

## The human problem

Weather-model developers need to run repeated tests quickly enough to find errors and compare changes before software reaches operations.

## The prior constraint

Long model runtimes slow the cycle of testing, diagnosis and improvement, especially as forecast systems become more complex.

## AI’s actual role

The computing environment supports numerical and AI-enabled weather-model development; the documented comparison is a regression-test runtime.

## The documented result

NOAA EPIC reports that one named Unified Forecast System regression test on Ursa completed about 30–40% faster than the same test on Hera when both used the Intel compiler.

## Why it may matter

Faster test cycles can let research teams examine more code changes before operational review, a useful capability even without a claim of better forecasts.

## Limitations

The result concerns one regression test and compiler configuration.

Wall-clock speed does not establish forecast accuracy, warning lead time or public outcomes.

The source is NOAA EPIC’s deployment account, not an independent benchmark.

## Unresolved questions

Do the gains persist across full model suites and production workloads?

How much researcher time and energy use change per completed experiment?

Which improvements reach operational forecasting after validation?

## Provenance and history

{
  "dates": {
    "eventDate": "2025-08-22",
    "publicationDate": "2025-08-22",
    "captureDate": "2026-09-19",
    "lastReviewedDate": "2026-09-19"
  },
  "provenance": {
    "origin": "editorial",
    "externalId": "https://www.epic.noaa.gov/ufs-weather-model-deployment-noaa-ursa/"
  },
  "revisions": [
    {
      "id": "revision:data-centers-noaa-ursa-weather-testing-01",
      "recordedAt": "2026-09-19",
      "summary": "Published this data-center record with its evidence state, measured or reported result, and material limitations kept together.",
      "sourceIds": [
        "source-noaa-ursa"
      ]
    }
  ],
  "corrections": []
}

## Original sources

- [UFS weather model deployment on NOAA Ursa](https://www.epic.noaa.gov/ufs-weather-model-deployment-noaa-ursa/)

## Continue exploring

- [Data Centers](https://brightaifuture.com/data-centers)
- [Planet](https://brightaifuture.com/worlds/planet)
- [Work & learning](https://brightaifuture.com/worlds/work)
- [How much earlier could we understand what is coming?](https://brightaifuture.com/threads/weather)
- [When does a technical result become a public capability?](https://brightaifuture.com/threads/community)