Change language:

Choosing of Frost 3D versions and hardware.

There are 5 versions of Frost 3D. The main difference of each version is the software implementation of mathematical solver used for numerical solving of the heat problem.

As a result, the question “Which version of Frost 3D could fulfill users specific requirements?” often arises.

Below is a list of parameters to help you to choose the most appropriate program version:

1. What are the typical dimensions of the computational domain during simulation (10 x 10 or 100 x 100 meters)?

2. What is the level of computational model detalisation – is there a significant amount of small elements important to the computation?

3. How critical is the computation accuracy – is significant coursing of computational domain possible during discretization?

4. What are the requirements for computation speed – are simulation results required in a matter of hours or can they wait a day or more?

Review of simulation results for the models in Frost 3D will help you choose the right software version. Different hardware setups were used for computations: 1 processor core / 4 processor cores, low-cost video card / powerful Nvidia accelerators.

Note: We remind you that Frost 3D software package only accepts Intel processors and graphics cards from NVIDIA.


Versions of Frost 3D: Computation speed comparison


Bovanenkovo gas field reservoir

Bovanenkovo gas field reservoir

Bovanenkovo gas field reservoir computation results: Coarse discretization

Computational domain: The number of cells: Simulation period:
14.6×11×19.5 meters 0.7 million 5 years
Computational unit:

single core, Intel Core i7

Computational time:

11 hours

Program version:

Single-core CPU Economy

Computational unit:

4 cores, Intel Core i7

Computational time:

6 hours

Program version:

Multicore CPU Standard

Computational unit:

NVIDIA GTX 660

Computational time:

2 hours

Program version:

Multicore GPU

Bovanenkovo gas field reservoir computation results: Fine discretization

Computational domain: The number of cells: Simulation period:
14.6×11×19.5 meters 3.5 million 5 years
Fine discretization
Computational unit:

4 cores, Intel Core i7

Computational time:

33 hours

Program version:

Multicore CPU Standard

Computational unit:

NVIDIA GTX 660

Computational time:

14 hours

Program version:

Multicore GPU


Computer simulation of ground freezing under an oil tank

Computational domain: The number of cells: Simulation period:
90×90×30 meters 7 millions 5 years
Computer simulation of ground freezing under an oil tank
Computational unit:

4 cores, Intel Core i7

Computational time:

11 hours

Program version:

Multicore CPU Unlimited

Computational unit:

NVIDIA GTX 660

Computational time:

4 hours

Program version:

Multicore GPU

Computational unit:

NVIDIA TITAN

Computational time:

27 minutes

Program version:

Multicore GPU

Learn more about computer simulation of ground freezing under an oil tank.


Prediction of ground thaw formations around an oil well

Computational domain: The number of cells: Simulation period:
40×60×200 meters 4 million 2 years
Soil thermal field distribution over 5 years around oil well
Computational unit:

4 cores, Intel Core i7

Computational time:

10 hours 37 minutes

Program version:

Multicore CPU Standard

Computational unit:

NVIDIA TITAN

Computational time:

57 minutes

Program version:

Multicore GPU

Learn more about prediction of ground thaw formations around an oil well.


Artificial ground freezing of underground tunnel

Computational domain: The number of cells: Simulation period:
62×20×18.5 meters 3.9 million 2 years
Computational unit:

4 cores, Intel Core i7

Computational time:

9 hours

Program version:

Multicore CPU Standard

+ Filtration module

Computational unit:

NVIDIA GTX 660

Computational time:

2 hours

Program version:

Multicore GPU

+ Filtration module

Learn more about computer simulation of ground freezing under an oil tank.


Eastern Siberia–Pacific Ocean oil pipeline section

Configuration of mutual arrangement of the pipeline and the ice wedges

ESPO pipeline computation results

Computational domain: The number of cells: Simulation period:
25×25×15 meters 4.8 million 20 years
3D temperature field calculation
Computational unit:

4 cores, Intel Core i5

Computational time:

89 hours

Program version:

Multicore CPU Standard

Computational unit:

NVIDIA GTX 660

Computational time:

34 hours

Program version:

Multicore GPU

Computational unit:

NVIDIA Tesla K20

Computational time:

10 hours

Program version:

Multicore GPU

Learn more about thermal analysis of an oil pipeline on permafrost.


Thermal influence of steel pile wall in TV tower (Gaz-Sale area)

Thermal influence of steel pile wall in TV tower

The main feature of this section is the steel pile wall (width 8 mm). The significant increase in computational time is due to the relatively small size of the elements (~ 1 mm) and high thermal conductivity (time higher than ground).

Thermal influence of a steel pile wall in a TV tower: simulation results

Computational domain: The number of cells: Simulation period:
2×2×12 meters 0.3 million 2 years
Thermal influence of a steel pile wall in a TV tower
Computational unit:

single core, Intel Core i5

Computational time:

146 hours

Program version:

Single-core CPU Economy

Computational unit:

4 cores, Intel Core i5

Computational time:

94 hours

Program version:

Multicore CPU Standard

Computational unit:

NVIDIA GTX 660

Computational time:

55 hours

Program version:

Multicore GPU

Computational unit:

NVIDIA Tesla K20

Computational time:

18 hours

Program version:

Multicore GPU


Efficiency assessment of thermal stabilization pipeline supports «Zapolyare – NPS «Purpe»

Efficiency assessment of thermal stabilization

The main features of this computation are 2 thermoprobes inside of each pile (diameter- 426 mm).

Computation results of the thermal stabilization in pipeline piles «Zapolyare – NPS «Purpe»

Computational domain: The number of cells: Simulation period:
6×8×11 meters 1.5 million 2 years
Computational unit:

4 cores, Intel Core i5

Computational time:

25 hours

Program version:

Multicore CPU Standard

Computational unit:

NVIDIA GTX 660

Computational time:

4 hours

Program version:

Multicore GPU

Computational unit:

NVIDIA Tesla K20

Computational time:

1.4 hours

Program version:

Multicore GPU


Computation of large sites. Ice wall around Fukushima NPP

Computation of large sites

The main features of this computation are:

1) large freezing region (perimeter 1.3 km);

2) high quality discretization (18 millions of cells);

3) a large number of freezing devices (1073 cooling pipes).

Simulation results of an ice wall around Fukushima NPP

Computational domain: The number of cells: Simulation period:
450×210×30 meters 17.8 million 2 years
Simulation results of an ice wall around Fukushima NPP
Computational unit:

4 cores, Intel Core i7

Computational time:

2 hours 51 minutes

Program version:

Multicore CPU Unlimited

Computational unit:

NVIDIA TITAN

Computational time:

7 minutes 26 seconds

Program version:

multi-core GPU

Computational unit:

NVIDIA Tesla K20

Computational time:

5 minutes 43 seconds

Program version:

multi-core GPU

Learn more about simulation of ground freezing around the perimeter of the Fukushima Nuclear Power Plant.


General information regarding time of numerical computation in Frost 3D

Computational speed is influenced by:

1) The size of the computational domain. A large computational domain requires a large number of cells, meaning a significant increase in computational time.

2) The cell sizes in the computational mesh: the computational time is different for 2х2х2 m and 20х20х20 m sites with similar cell quantities; the 2х2х2 m computation site will take longer because of the relatively small cell size.

3) Thermophysical properties of materials (layers). Materials of high thermal conductivity significantly increase computational time.

4) Boundary conditions (speed of environment temperature changes and heat transfer coefficient). The higher speed and amplitude of changes over time resulting in a significant increase in computational time.

5) Cooling devices and working mode. The more cooling devices present and the greater their heat capacity, the longer the time of computation.

The efficiency of computation parallelization (computation speed using multi-core computing systems) is influenced by following factors:

1) The more materials and different boundary conditions present in the computational domain, the less the degree of parallelization of computations (the acceleration of the calculation in the transition from single core to multi-core CPU or GPU).

2) Irregularity of the computational mesh reduces the degree of computation parallelization.