SolverCloud
Open-source solvers in the cloud
The Tools in Your Sandbox
Seven tools come preinstalled on every sandbox. Here's what each one is for, when you'd reach for it, and where to go to learn it properly.
On this page
Solver
Parallel execution
Post-processing
Rendering
Before you dive in
Everything here is already installed — nothing on this page needs you to build or download a solver. Confirm your toolchain with the verification script in the tutorial.
PyFR lives under
/opt/pyfrand MFC under/opt/MFC; the rest are on yourPATH. Each section below repeats this so you know where to look.You have root, so anything missing is one
aptorpipaway — for the life of that VM.
High-order scale-resolving CFD
PyFR solves the compressible and incompressible Navier–Stokes equations using flux reconstruction, a high-order method on unstructured mixed-element meshes. Cases are described in a single INI file that names the mesh, the equation system, the time integrator and any output plugins.
What it's good for
Scale-resolving turbulence — LES, DNS and ILES — where you need to resolve eddies rather than model them. Typical work is external aerodynamics, turbomachinery, wakes and vortex-dominated flow.
Where to start
Read the user guide end to end once, then run one of the shipped examples before touching your own geometry. The example cases are the fastest way to see a correct .ini file.
Learn it from the source
Multiphase compressible flow
MFC is a high-order multiphase compressible flow solver built for problems with strong shocks and sharp material interfaces. It is driven through its own mfc.sh wrapper, which handles building, running and testing a case.
What it's good for
Shock–droplet and shock–bubble interaction, cavitation, bubble collapse, phase change, and other flows where two materials meet across a moving interface.
Where to start
Run ./mfc.sh --help inside /opt/MFC to see the available subcommands, then work through the examples in the repository. The CMF platform publishes over 130 worked MFC cases.
Learn it from the source
Full-wave computational electromagnetics
Palace (PArallel LArge-scale Computational Electromagnetics) solves Maxwell’s equations with finite elements, in the frequency domain, the time domain, and as eigenmode problems. Cases are defined in a JSON configuration file paired with a mesh.
What it's good for
Antennas and radiation patterns, transmission lines and S-parameters, resonators and cavity modes, and superconducting-qubit device design.
Where to start
The documentation’s configuration reference is the important part — a Palace run is almost entirely determined by that JSON file. The CEM platform has worked examples for dipoles, coaxial lines, waveguides and qubit couplers.
Learn it from the source
Run solvers across all your cores
MPI is the message-passing standard every solver here uses for parallelism. In practice you prefix a run with mpirun and a process count, and the solver partitions the problem across those ranks.
What it's good for
Using the VM you are already paying for. A 16 vCPU sandbox running a single-threaded job is wasting fifteen cores.
Where to start
Start with -n set to the number of vCPUs on your sandbox and scale down if memory per rank becomes the limit. More ranks than cores will slow a job down, not speed it up.
Learn it from the source
Inspect and plot solver output
ParaView is the standard open-source post-processor for VTK-format field data. It reads the .vtu/.pvtu output your solver writes and gives you slices, contours, streamlines, integrated quantities and line plots.
What it's good for
Checking whether a run is physically sensible, extracting forces and coefficients, and producing the figures that end up in a report.
Where to start
Run it on the sandbox desktop over Remote Desktop so the data never has to leave the machine. For repeatable figures, use pvpython to script the same view across several runs rather than clicking through the GUI each time.
Learn it from the source
Presentation-quality rendering and geometry work
Blender is a full 3D suite. Alongside ParaView it covers the jobs a post-processor is weak at — photoreal rendering of a result, camera animation, and cleaning up or repairing geometry before meshing.
What it's good for
Turning a correct-but-plain contour plot into something usable in a presentation, and scripted batch rendering of many runs.
Where to start
It runs headless with -b and a Python expression, so renders can be driven from a script on the sandbox without opening the GUI at all — the same pattern the tutorial’s verification script uses to check the install.
Learn it from the source
Try one of them
Reading only gets you so far. Create a sandbox, connect to its desktop, and run one of the shipped example cases — that's the fastest way to see how any of these tools actually behave.
SolverCloud
© 2026 SolverCloud, a NumericalAI product by Empirisch Tech GmbH (empirischtech.at). All rights reserved. |Privacy Policy |Terms of Service |FAQ |Licenses