Molecular

One image, built forthe card you attach

GROMACS with both MPI and CUDA compiled in.gmx_mpi takes the GPU when there is one, and the CPU when not.

GROMACS 2026.34 cores · 32 GBNVIDIA GPURuns in a shell

hub://gromacs

$ dxflow workflow create --identity gromacs hub://gromacs --start

01Pick a machinecores, memory and a rate per hour

02Create the workflowpulled from the hub, once

03Run it, then stopstopped when the work is done

The application

One image,both ways to run

Built from source with AVX2, its own FFTW, and the usual force fields.

GPU or CPU

mdrun -nb gpu offloads to the card; with no card it runs on a CPU.

MPI ranks

mpirun -np N gmx_mpi mdrun spreads the run over the cores you got.

The usual force fields

The AMBER, CHARMM, GROMOS and OPLS fields, plus free energy, REMD.

How you run it

The container stays up,and you drive it

There is nothing to configure at start. Put the system under /volume and run gmx_mpi against it.

a system, from pdb to trajectory

$ gmx_mpi pdb2gmx -f protein.pdb -o processed.gro -water spce

$ gmx_mpi grompp -f md.mdp -c npt.gro -t npt.cpt -p topol.top -o md.tpr

$ gmx_mpi mdrun -v -deffnm md -nb gpu

Reach for the cardAdd -nb gpu, and -pme gpu or -bonded gpu beside it. nvidia-smi says whether it is there.

Or spread it widempirun -np N in front of mdrun, with -ntomp for the threads inside each rank.

Everything lands on /volumeTrajectories, logs and analysis are written where you put the inputs, and stay after the stop.

The shell

No window,just a prompt

The container stays up, and you run gmx_mpi against /volume.

gmx_mpiCommandon the PATH already

/volumeStorageinputs, trajectories, logs

The container stays upIt idles until you stop it, so the shell is there when you are.

Everything on /volumePut the inputs there; trajectories and logs are written back.

Good to know

One image,both ways of running

MPI and CUDA are built in together, so the same image covers the card and the cores.

It falls back on its own

With a card attached, offload what you want. Without one, gmx_mpi runs on CPU and says nothing about it.

Built from source

Compiled with AVX2 and its own FFTW. GMXRC is already sourced when a shell opens.

The fields and the methods

AMBER, CHARMM, GROMOS and OPLS, with free energy, umbrella sampling and REMD alongside them.

The image

Pulled once,then it stays

GROMACS arrives as one image. This is what comes down the first time, and what the disk should have free for it.

5.4Gamd64compressed, the way the registry counts it

50GOn diskunpacked, with room to work beside it

ghcr.io/dxflow-ai/gromacs:latestPublished from the hub, pulled on the first start and kept for the ones after it.

amd64 onlyOne architecture is published, so pick a machine that matches it.

What it asks for

What it wants,and what it needs

The definition asks for 4 cores and 32 GB. The image comes up on less than that, and a start given --fit trims the ask to whatever the machine actually has.

4 cores · 32 GBAsks forwhat the definition writes down

2 cores · 16 GBRuns onthe least the image comes up on

NVIDIAGPUrequired, and filtered for

The ask is not the floorThe definition writes down what suits the work. The image itself starts on less, which is what the second figure is.

--fit caps it to the hostA start given --fit trims each step to what the machine actually has, for that start alone. The definition is never rewritten.

Machines that fit it

GROMACS asks for 4 cores and 32 GB, with an NVIDIA GPU. Cheapest first.

G4dn 2XLarge

$0.94/ hour8 cores · 32 GB · Tesla T4Start this machine

NC8as T4 v3

$0.94/ hour8 cores · 56 GB · Tesla T4Start this machine

G2 Standard-8

$1.07/ hour8 cores · 32 GB · NVIDIA L4Start this machine

Run GROMACS on your own machinePick a machine that meets it, and it opens about a minute after you ask.

Start a machine Browse workflows