Solver Unit Draw

A description of the licensing for the solvers.

The following table lists all Altair solvers.
Solver Name
Functionality
AcuSolve
Finite-Element Navier-Stokes CFD Solver (includes AcuFwh, AcuTrace, AcuView)
EDEM
Discrete Element (DEM) Solver for bulk materials
Feko
High-frequency Electromagnetics Solvers (includes WinProp, and newFASANT)
Flux
Low-frequency Electromagnetics and Electromechanical Solvers
HyperForm
One Step Stamping Solver
Manufacturing Solver
Mold-filling, Casting, Extrusion Process Solvers
MotionSolve
Multibody System Solver
nanoFluidX
Smoothed Particle Hydrodynamics (SPH) Solver, GPU-based
OptiStruct
Structural and Multi-Physics Solver and Optimizer
Radioss
Explicit Structural and Multi-Physics Solver
ultraFluidX
Lattice-Boltzmann CFD Solver, GPU-based

In applications where solvers are an integral part of the solution, the unit draw is determined by the GUI application. These applications are Inspire, ElectroFlo, PollEx PCB, Seam, SimLab, and SimSolid. The unit draw is documented in sections Desktop Applications Unit Draw, Inspire Family Unit Draw, and Standalone Applications Unit Draw above.

GPU-based CFD solvers nanoFluidX and ultraFluidX are licensed based on the number of GPUs used per job.

The solvers AcuSolve, EDEM, Feko (including WinProp, and newFASANT), Flux, HyperForm (only InApp), Manufacturing Solver, MotionSolve, OptiStruct, and Radioss are licensed in two ways:
  • In-App Licensing
  • HPC Licensing

Solver In-App Licensing

In-App licensing draws Altair Units per job. The number of Altair Units required to run solver jobs varies depending on the number of the CPU cores and GPU cards, and how many jobs are running concurrently.

Solvers started from inside EDEM, Feko (including WinProp, and newFASANT), Flux, HyperForm, HyperXtrude, HyperMesh, HyperWorks, MotionView, and SimLab use In-App licensing.

GPU acceleration is implemented for AcuSolve, EDEM, Feko, and OptiStruct. For AcuSolve, Feko, and OptiStruct, one GPU card is counted as four additional CPU-cores; for EDEM, one GPU card is counted as 16 CPU cores. For example, if running an OptiStruct job using 4 cores plus 1 GPU, it will be considered as 8 cores.
Number of Cores SimlabFlux2DSolver* AcuSolve, AltairManufacturingSolver, FekoSolver, FluxSolver, HyperFormSolver, MotionSolve, OptiStructFEA, Radioss, SimLabXXXSolver, WinPropSolver EDEMSolver, MotionSolve (w/optimization), OptiStruct
1-4 15 30 50
5-8 20 35 55
9-16 25 40 60
17-32 30 50 70
33-64 40 60 80
65-128 50 70 90
129-256 60 80 100
257-512 70 90 110
513-1024 80 100 120
Each duplication +10 +10 +10

On top of the primary license draw, a decay function is implemented. The decay reduces the total license draw, if multiple jobs of the same solver are running concurrently off the same license server. The decay factor is a multiplier applied to the regular license draw of a solver job. See table below.

Once a job has finished, the next job will backfill into the vacated slot and the same multiplier is applied as for the previous job. Once a job has started, the checked-out unit amount for that job will not change.

In addition to the unit draw of solvers during runtime, AcuPrep, Feko, Flux, OptiStruct, and Radioss Starter check for the existence of a license during initial check and preparation runs. These license checks do not draw any units.
Job Number
Decay Factor
1
1.0
2...10
0.9
11...20
0.8
21...30
0.7
31...40
0.6
41...50
0.5
50+
0.4

*For SimlabFlux2DSolver

Job Number
Decay Factor
1
1.0
2...4
0.1334
5...8
0.0667
9...36
0.05
37...70
0.0475
70+
0.0296
In-App licensed solvers level or stack AUs according to the following rules:
  • At the first invoke of a solver, the AUs level against applications already running on the same machine.
  • At the first invoke of a different solver, the units level.
  • Launching additional jobs of a solver already running, stacks the additional units.
  • When the leveled job finishes, the next invoke levels again.
  • Stacked jobs continue to stack after the leveling job has finished.

For example, if you launch HyperMesh first (21 AUs), and then launch Radioss (30 AUs), 30 AUs will be drawn. If the same user on the same machine adds an OptiStruct job (50 AUs requested), the total units drawn is 50 AUs due to leveling. If you now add a second OptiStruct job, the total units drawn will only be increased to 95 AUs (1.0*50+0.9*50 AUs).

Solver HPC Licensing

The Solver HPC license draws units on a per-CPU core basis. The unit draw is determined by the total number of cores for a solver according to a lookup table.

HPC Licensing applies in all situations where a solver is not started from inside an Altair GUI application and a SolverHPC license feature is present.

The following table applies to the license features: AcuSolve, AltairManufacturingSolver, EDEMSolver, FekoSolver (including WinPropSolver), FluxSolver, MotionSolve, OptiStructFEA, and Radioss. If OptiStruct runs an optimization (OptiStruct license feature), the draw per core is multiplied by 1.5 for a given job. Each license feature follows its own core count. This unit draw stacks.

GPU acceleration is implemented for AcuSolve, EDEM, Feko, and OptiStruct. For AcuSolve, Feko, and OptiStruct, one GPU card is counted as four additional CPU-cores; for EDEM one GPU card is counted as 16 CPU cores.
Number of Cores Draw per Core Cumulative Draw Range*
1-4   30.00
5-36 1.60 31.60 - 81.20
37-70 0.80 82.00 - 108.40
71-200 0.60 109.00 - 186.40
201-400 0.40 186.80 - 266.40
401-800 0.30 266.70 - 386.40
801-1600 0.20 386.60 - 546.40
1601-3200 0.15 546.55 - 786.40
3201-6400 0.10 786.50 - 1106.40
6401-10000 0.05 1106.45 - 1286.40
>10000 (20000) 0.04 >1286.40 (1686.40)

*There may be slight variations due to numerical round-off.

For example, if three simultaneous Radioss jobs are running requiring a total of 48 CPU cores, the total unit draw will be 90.8 (= 30 + 32x1.6 + 12x0.8).

In addition to the unit draw of solvers during runtime, AcuPrep, Feko, OptiStruct, and Radioss Starter check for the existence of a license during initial check and preparation runs. These license checks do not draw any units.

Leveling of interactive applications is not affected. For example, a user on a workstation uses HyperMesh, HyperView (21 AUs leveled) plus two 4-core OptiStructFEA (36 AUs), a (maximum) total of 57 AUs is drawn.

Modified Solver Licensing

A description of modified licensing for the solvers.

The Altair License Manager allows modifying license behavior thru the configuration file altair-serv.cfg. Solver licensing is controlled by the license features SolverHPC and SolverInApp. Default behavior as described is assumed without neither of these features being modified via the configuration file. Should the features be altered in the configuration file or license, the behavior in the following table applies.
SolverHPC SolverInApp Behavior
Yes Yes In-App from GUIs, Elsewhere HPC.
Yes No HPC only.
No Yes In-App where HPC licensing denied.
No No In-App only.

Licensing of GPU-based CFD Solvers

A description of GPU-based licensing for CFD solvers.

The CFD solvers nanoFluidX and ultraFluidX are a direct GPU implementation and draw Altair Units per job based on the number of GPUs.
GPU nanoFluidX ultraFluidX
1 25 50
2 50 100
3-4 100 150
5-8 150 200
9-16 200 250
Each duplication +50 +50