Run the science, not the queue.
HPC clusters, GPU nodes and parallel storage for research groups, institutes and university computing centres.
The constraint
Shared national and campus systems are oversubscribed, and a group can wait weeks for an allocation that ends before the paper does. Research codes also have different appetites: some want GPUs, some want CPU cores and memory bandwidth, and nearly all want fast parallel storage. A cluster built to one generic spec serves none of them well.
Workloads
Molecular dynamics
Run GROMACS, NAMD, AMBER and OpenMM on GPU nodes for long-timescale simulations.
CFD & FEA
Solve mesh-heavy jobs in OpenFOAM, Ansys Fluent and similar codes on dedicated nodes.
Climate, weather & earth system modelling
Run high-resolution models with storage sized for their output.
Quantum chemistry & materials science
Run DFT and post-Hartree-Fock methods in codes such as VASP, Quantum ESPRESSO and CP2K on mixed CPU and GPU nodes.
AI for science
Train and fine-tune research models on the same cluster as the simulations that feed them.
Systems
- GLabs GPU Servers — GPU compute nodes.
- GLabs Vault — parallel scratch, project storage and archive.
- GLabs HPC — institute- and centre-scale systems.
Deployed with
GLIDS for the cluster, scheduler and storage. GLabs AI Stack for AI-for-science work.
Related solutionsHPC Use CasesFoundation Model Pre-Training
FAQs
GPU or CPU nodes for our research codes?
It depends on the code. Molecular dynamics, many CFD solvers and almost all AI run well on GPUs; many quantum chemistry and legacy codes still favour CPU cores and memory bandwidth. We profile your job mix and build a mixed cluster to match.
What storage does a research cluster need?
Usually two tiers: a fast parallel file system for scratch and active projects, and a larger capacity tier for archive. We size both from your data volumes and I/O patterns.
Can we add capacity later?
Yes. Clusters are designed to grow node by node, with headroom in power, cooling and network for future racks.
Send us your job mix. We'll design the cluster around it.
All product and company names mentioned are trademarks of their respective owners.