Cosmic ambition at scale: UK’s supercomputer unlocks a 2.5 petabytes universe
Cosmic ambition at scale: UK’s supercomputer unlocks a 2.5 petabytes universe
previous arrow
previous arrow
next arrow
next arrow
Shadow
Featured

Cosmic ambition at scale: UK’s supercomputer unlocks a 2.5 petabytes universe

 
Marking a significant advancement in computational astrophysics, researchers at Durham University have released one of the most extensive cosmological simulation datasets to date. This comprehensive digital reconstruction of the Universe, enabled by high-performance supercomputing, demonstrates the unprecedented scale at which modern astrophysical phenomena can be modeled and analyzed.
 
Central to this achievement is the FLAMINGO project, an international collaborative effort aimed at simulating the evolution of matter across cosmological timescales. The resulting dataset, exceeding 2.5 petabytes, a volume comparable to approximately half a million high-definition films, provides the global scientific community with access to highly detailed virtual universes that trace the formation and evolution of cosmic structures from the post-Big Bang epoch to the present era.

Supercomputing as the engine of discovery

The simulations were executed on the COSMA-8 supercomputer, part of the UK’s DiRAC national high-performance computing infrastructure. This system, purpose-built for data-intensive cosmological workloads, enabled the integration of vast spatial scales with sophisticated physical modeling.
 
Unlike earlier generations of cosmological simulations, which often forced a trade-off between resolution and scale, FLAMINGO bridges both extremes. It simultaneously models:
  • Gigaparsec-scale cosmic volumes, spanning billions of light-years
  • Galaxy formation physics, including gas dynamics, star formation, and feedback processes
  • Dark matter and dark energy evolution, the dominant drivers of cosmic structure
  • Large-scale clustering, producing the filamentary “cosmic web” observed in galaxy surveys
This dual capability reflects a fundamental shift enabled by supercomputing: the convergence of astrophysical detail with cosmological precision.

The computational architecture of a universe

From a technical standpoint, the FLAMINGO simulations represent a triumph of parallel computing and algorithmic design. The underlying software stack, built around advanced cosmological simulation codes such as SWIFT, leverages:
  • Massively parallel processing, distributing billions of computational elements across thousands of cores
  • Hybrid gravity–hydrodynamics solvers, capturing both collisionless dark matter and baryonic physics
  • Time-resolved evolution, tracking the growth of structure across cosmic epochs
  • Petascale I/O pipelines, capable of writing, storing, and indexing multi-petabyte outputs
These simulations follow matter as it collapses under gravity, forming halos, galaxies, and clusters, while simultaneously modeling the energetic feedback from stars and black holes that regulates galaxy growth. The result is a statistically robust, physically grounded synthetic universe.
 
Crucially, the dataset’s size and complexity required not only raw compute power but also innovations in data accessibility. The team developed a web-based platform that allows researchers to query and extract subsets of the data without downloading entire petabyte-scale files, effectively democratizing access to supercomputer-scale science.

A global resource for precision cosmology

The scientific potential of the dataset is vast. Cosmological simulations are essential tools for interpreting observational data from next-generation telescopes and surveys. By comparing simulated universes with real observations, researchers can test competing models of:
  • Dark matter particle properties
  • Dark energy and cosmic acceleration
  • Galaxy formation and evolution
  • Large-scale structure statistics
FLAMINGO’s scale enables percent-level precision cosmology, allowing subtle deviations between theory and observation to be identified and explored.
 
Moreover, the ability to simulate rare, large-scale structures, such as massive galaxy clusters, provides insights that smaller simulations cannot capture. These structures serve as sensitive probes of cosmological parameters and fundamental physics.

Inspiring the next era of computational science

The release of this dataset is more than a technical achievement; it is a statement about the future of science. By making one of the largest supercomputer-generated datasets openly available, the team is lowering the barrier to entry for researchers, students, and institutions worldwide.
 
As noted by project leaders, access to facilities like COSMA-8 is typically limited. By distributing the results of these simulations globally, the project transforms a localized supercomputing capability into a shared scientific resource.
 
This approach reflects a broader trend: supercomputing is no longer just a tool; it is an infrastructure for collaboration and discovery.

Toward an exaflops cosmos

Looking ahead, projects like FLAMINGO foreshadow the coming era of exaflops computing, where simulations will achieve even higher resolution, incorporate additional physical processes, and integrate real-time observational data.
 
For now, Durham’s 2.5 petabytes universe stands as a powerful demonstration of what is possible when computational ambition meets scientific vision. It is a reminder that, in the age of supercomputing, humanity is no longer limited to observing the cosmos; we are beginning to recreate it.
Featured

Intel's Q1 results signal supercomputing surge driving Xeon momentum

In a strong start to 2026, Intel has reported first-quarter financial results that underscore a growing reality across the high-performance computing (HPC) landscape: demand for supercomputing and AI-scale infrastructure is accelerating, and with it, the need for powerful server processors.
 
The company posted revenue of approximately $13.6 billion, exceeding expectations and marking a notable 7% year-over-year increase. Earnings also surpassed forecasts, reflecting renewed strength in Intel’s core data center business.
 
Shares of the U.S. chipmaker jumped 15% in after-hours trading.

Supercomputing demand lifts Xeon sales

At the center of this growth is Intel’s Xeon processor line, long a cornerstone of supercomputing systems and hyperscale data centers. As global investment in AI, simulation, and large-scale modeling intensifies, Xeon-based platforms are seeing renewed demand.
 
Intel’s data center segment delivered particularly strong performance, generating over $5 billion in revenue for the quarter and outperforming analyst expectations. This surge is closely tied to expanding workloads in AI training, scientific simulation, and cloud-scale analytics, domains traditionally dominated by supercomputing infrastructure.
 
Xeon processors remain deeply embedded in HPC ecosystems. Historically, they have powered a majority of the world’s top supercomputers, thanks to their high core counts, memory bandwidth, and compatibility with parallel workloads. As modern systems evolve toward hybrid CPU-GPU architectures, CPUs like Xeon continue to orchestrate workloads, manage data movement, and execute complex simulations.

AI and HPC converge

A key driver behind this momentum is the convergence of AI and traditional supercomputing. Workloads once confined to national labs, climate modeling, molecular dynamics, and astrophysics are now intersecting with enterprise AI applications such as large language models and digital twins.
 
Intel executives emphasized that CPUs remain essential even in GPU-heavy environments. Server processors are critical for feeding data to accelerators, running inference workloads, and maintaining system-level efficiency.
 
This architectural balance is fueling demand not just for accelerators, but for robust general-purpose compute, an area where Xeon continues to play a pivotal role.

Market tailwinds favor HPC growth

Broader industry trends reinforce Intel’s position. The global x86 server market remains dominant, accounting for the vast majority of server shipments and benefiting from the rapid expansion of hyperscale data centers.
 
At the same time, AI infrastructure investments are reaching unprecedented levels, with enterprises and governments alike racing to deploy supercomputing-class systems. These deployments increasingly require dense, scalable CPU architectures capable of handling both traditional HPC and emerging AI workloads.
 
Intel’s continued investment in next-generation Xeon platforms, including upcoming architectures designed for higher core counts and improved memory throughput, positions the company to capitalize on this shift.

Looking ahead: A supercomputing renaissance

While challenges remain, including competition and prior supply constraints, Intel’s latest results suggest a turning point. The company is benefiting from a broader resurgence in compute demand, driven by the same forces that are redefining supercomputing itself.
 
From national laboratories to cloud providers, the appetite for high-performance infrastructure is expanding rapidly. And as that demand grows, so too does the importance of the processors at its foundation.
 
For Intel, the message from Q1 2026 is clear: the supercomputing era is not just alive, it is accelerating, and Xeon is riding the wave.
Featured

Multi-layer simulations reveal the hidden supply chain of solar prominences

A recent study in Nature Astronomy (DOI: 10.1038/s41550-026-02840-7) represents a major leap forward in computational astrophysics, showcasing how state-of-the-art supercomputing is transforming our understanding of solar prominences, one of the Sun’s most mysterious features. By utilizing large-scale, high-resolution simulations, scientists at the Max Planck Institute for Solar System Research have, for the first time, recreated the Sun’s complex multi-layer dynamics from the convection zone up to the corona, uncovering a self-sustaining mechanism that supports these plasma formations.

A multiscale computational challenge

Solar prominences are massive, relatively cool plasma formations (~10,000 K) suspended within the Sun’s much hotter corona (~1 million K). Despite their apparent fragility, they can persist for weeks or months and may ultimately erupt, triggering space weather events that can disrupt Earth’s infrastructure.
 
Modeling such structures presents a formidable computational challenge. The physics spans multiple regimes: magnetohydrodynamics (MHD), radiative transfer, thermal instability, and turbulent plasma flows across spatial scales ranging from sub-surface convection to coronal loops extending tens of thousands of kilometers.
 
The research team addressed this complexity using advanced numerical simulations that integrate:
  • Full-sphere stratified solar models, extending from the convection zone below the photosphere to the corona.
  • Dynamic magnetic field evolution, driven by turbulent plasma flows.
  • Thermal coupling across layers, capturing steep temperature gradients between the chromosphere and corona.
  • Plasma injection and condensation processes, resolved in time-dependent MHD frameworks.
These simulations required high-performance computing (HPC) resources capable of resolving nonlinear interactions across scales while maintaining numerical stability over long simulation times.

Magnetic topology and plasma supply mechanisms

At the core of the study lies a specific magnetic configuration: a double-arched field structure forming a dip in the corona. Within this dip, prominence material accumulates and remains magnetically confined.
 
The simulations reveal a dual supply mechanism:
  1. Chromospheric Injection
    Turbulent magnetic activity in the chromosphere ejects bursts of cool plasma upward. These injections are driven by small-scale magnetic reconnection and wave dynamics.
  2. Coronal Condensation
    Hot coronal plasma flows along magnetic field lines into the dip, where it cools radiatively and condenses into denser material.
Simultaneously, gravitational drainage causes some plasma to fall back toward lower layers. The prominence persists because these losses are continuously offset by the two supply channels, establishing a dynamic equilibrium.
 
This “supply–loss balance” represents a key breakthrough: earlier models typically captured only coronal condensation and neglected the deeper layers of the Sun. By coupling subsurface dynamics with atmospheric processes, the new simulations close a longstanding gap in solar physics.

Supercomputing as the enabling infrastructure

The study’s significance lies not only in its astrophysical findings but in its computational methodology. Achieving a self-consistent, multi-layer solar model required:
  • Massively parallel MHD solvers to handle nonlinear plasma dynamics.
  • Adaptive mesh refinement (AMR) or equivalent resolution strategies to capture fine-scale injection events.
  • Long-duration time integration to observe prominence formation and stability cycles.
  • High-throughput data handling, given the volumetric and temporal scale of simulation outputs.
Such requirements place the work squarely in the domain of modern supercomputing. Without HPC systems capable of petaflops (and increasingly exaflops) performance, resolving the coupled dynamics of magnetic fields and plasma across solar layers would be computationally prohibitive.

Implications for space weather forecasting

Understanding prominence formation is not merely an academic pursuit. Prominence eruptions are closely linked to coronal mass ejections (CMEs), which can trigger geomagnetic storms affecting satellites, power grids, and communications systems.
 
By identifying the underlying supply mechanisms and stability conditions of prominences, the study provides a pathway toward:
  • Improved predictive models of solar eruptions.
  • Better integration of subsurface solar dynamics into space weather simulations.
  • Enhanced coupling between observational data and physics-based HPC models.
As noted by the researchers, a deeper understanding of prominences is “a crucial piece of the puzzle” in forecasting hazardous space weather events.

Toward exaflops solar physics

This study highlights a growing shift in astrophysics: moving away from inference based solely on observations toward data-intensive, physics-driven modeling. High-fidelity simulations of entire stellar subsystems are quickly becoming a hallmark of modern research in the field.
 
Future directions will likely include:
  • Integration with real-time solar observation pipelines.
  • Data assimilation frameworks combining HPC simulations with satellite measurements.
  • Deployment on exaflops architectures to increase spatial resolution and physical realism.
In this context, solar prominences are no longer just spectacular features of our nearest star; they are a proving ground for the next generation of supercomputing-enabled science.
  1. Supercomputers peer into alien worlds, find matter unlike anything on Earth
  2. Riding invisible waves: How open-source code transforms space weather science
POPULAR NOW
MOST READ THIS YEAR
MOST READ