gpuai.shop
Use case

Scientific and high-performance computing

FP64-capable nodes for simulation, genomics and computational chemistry.

Double precision

Consumer cards are effectively useless for FP64 work - the RTX 4090 runs double precision at 1/64 of its single-precision rate. The datacentre cards do not have that restriction, which makes the A100 and H100 the only sensible choices for anything that needs numerical accuracy over throughput.

Common workloads

  • GROMACS and AMBER molecular dynamics
  • OpenFOAM and Ansys Fluent CFD runs
  • GATK and Parabricks genomics pipelines
  • VASP and Quantum ESPRESSO electronic structure
  • Monte Carlo simulation at scale

Multi-node

MPI over InfiniBand is available on all SXM nodes. The image ships OpenMPI 4.1 with UCX configured for the fabric, so a standard mpirun invocation works without tuning.