CFD中文网

sw只能这样，不过有个技巧，对icem有用的是面，实体什么没用

@李东岳 和谐版fluent似乎有bug，指定GPU会报错，重装试试

@李东岳 我试一下

显存占用：

第一台设备：NVIDIA DGX-1 8xV100 32GB
CPU:8/40 2x Xeon E5 2698 v4
GPU:Nvidia Tesla V100 x8
超线程：关闭
RAM:500GB
系统：Ubuntu 20.04 LTS
Fluent版本：Fluent 2023R1 Linux64
打开Native GPU 模式
GPU数量：8
迭代步数：20

parallel timer usage

Performance Timer for 20 iterations on 8 compute nodes
  Average wall-clock time per iteration:                1.030 sec
  Global reductions per iteration:                          0 ops
  Global reductions time per iteration:                 0.000 sec (0.0%)
  Message count per iteration:                              0 messages
  Data transfer per iteration:                          0.000 MB
  LE solves per iteration:                                  0 solves
  LE wall-clock time per iteration:                     0.000 sec (0.0%)
  LE global solves per iteration:                           0 solves
  LE global wall-clock time per iteration:              0.000 sec (0.0%)
  LE global matrix maximum size:                            0
  AMG cycles per iteration:                             0.000 cycles
  Relaxation sweeps per iteration:                          0 sweeps
  Relaxation exchanges per iteration:                       0 exchanges
  LE early protections (stall) per iteration:           0.000 times
  LE early protections (divergence) per iteration:      0.000 times
  Total SVARS touched:                                    366

  Total wall-clock time:                               20.607 sec


Simulation wall-clock time for 20 iterations        173.04942 sec

@李东岳 V100已经出来快5年了，现在国内主流是A800和H800，前两天联想代理说OEM的GPU服务器8卡A800 90万

@李东岳 根据用户手册，GPU求解器对湍流模型、边界条件等有较多的限制
34.9.1. Models

The following models are supported by the Fluent GPU Solver:

Thermal energy and CHT. See Modeling Thermal Energy.

Species transport without reactions. See Species Transport Without Reactions.

The GPU solver supports the following turbulence models:

Laminar

Standard and realizable k-epsilon

k-omega GEKO and k-omega SST

Large Eddy Simulation (LES)

For the LES turbulence model, the following sub-grid scale models are supported:

Smagorinsky-Lilly

WALE

Stress-Blended Eddy Simulation (SBES)

34.9.2. Material Properties

Material properties defined with the following settings are supported by the Fluent GPU Solver:

constant

ideal-gas

incompressible-ideal-gas

boussinesq

polynomial

piecewise-linear

piecewise-polynomial

sutherland viscosity law

34.9.3. Solver Settings

This section decribes the general solver settings and solution methods that are supported by the Fluent GPU Solver.

The following general solver settings are supported:

Pressure-Based solver

Absolute velocity formulation

Steady and Transient flows

For instructions on specifying the solver settings above see General Task Page.

The following solution methods are supported:

SIMPLE and SIMPLEC pressure-velocity coupling schemes. See Pressure-Based Solver Settings for details.

First-order and second-order spatial discretization schemes. See Choosing the Spatial Discretization Scheme for details.

The Bounded Central Differencing spatial discretization scheme is supported and becomes available when Large Eddy Simulation (LES), Stress-Blended Eddy Simulation (SBES), or URANS turbulence models are enabled. For more information see Modeling Turbulence.

First order and second order implicit transient formulations are supported for incompressible transient cases.

34.9.4. Cell Zone and Boundary Conditions

This section outlines the cell zone and boundary conditions supported by the Fluent GPU Solver.

The following boundary condition settings are supported:

Wall

Stationary Wall and Moving Wall, including Rotational wall motion. See Wall Boundary Conditions.

Thermal conditions. See Thermal Boundary Conditions at Walls.

Translational Periodic boundary condition with or without a non-conformal mesh interface. See Periodic Boundary Conditions for details.

Inlet

Velocity-inlet (incompressible). See Defining the Velocity.

Pressure-inlet. See Pressure Inlet Boundary Conditions.

Mass-flow-inlet (incompressible). See Mass-Flow Inlet Boundary Conditions.

Outlet

Pressure-outlet. See Pressure Outlet Boundary Conditions

Mass-flow-outlet (incompressible). See Mass-Flow Outlet Boundary Conditions

Symmetry. See Symmetry Boundary Conditions.

Porous Jump. See Porous Jump Boundary Conditions.

Intake Fan. See Intake Fan Boundary Conditions.

Outlet Vent. See Outlet Vent Boundary Conditions.

For the supported boundary conditions listed above, boundary condition settings can be defined as constant or with a profile, as outlined in Profiles.

The Fluent GPU Solver supports fluid cell zones with the following settings:

Constant Source Terms. See Defining Mass, Momentum, Energy, and Other Sources.

Porous Zone. See Porous Media Conditions.

@百善孝为先0626
【ANSYS 2023R1最新中英版安装教程！】 https://www.bilibili.com/video/BV153411m7sj/?share_source=copy_web&vd_source=77b01b023ae092eeabee9ce43ae2853f
视频评论区

显存占用

测试设备NVIDIA DGX-1 8xV100 32GB
CPU:8/40 2x Xeon E5 2698 v4
GPU:Nvidia Tesla V100 x8
超线程：关闭
RAM:500GB
系统：Ubuntu 20.04 LTS
Fluent版本：Fluent 2023R1 Linux64
打开Native GPU 模式
GPU数量：8
迭代步数：40

Performance Timer for 40 iterations on 8 compute nodes
  Average wall-clock time per iteration:                0.748 sec
  Global reductions per iteration:                          0 ops
  Global reductions time per iteration:                 0.000 sec (0.0%)
  Message count per iteration:                              0 messages
  Data transfer per iteration:                          0.000 MB
  LE solves per iteration:                                  0 solves
  LE wall-clock time per iteration:                     0.000 sec (0.0%)
  LE global solves per iteration:                           0 solves
  LE global wall-clock time per iteration:              0.000 sec (0.0%)
  LE global matrix maximum size:                            0
  AMG cycles per iteration:                             0.000 cycles
  Relaxation sweeps per iteration:                          0 sweeps
  Relaxation exchanges per iteration:                       0 exchanges
  LE early protections (stall) per iteration:           0.000 times
  LE early protections (divergence) per iteration:      0.000 times
  Total SVARS touched:                                    366

  Total wall-clock time:                               29.934 sec

Simulation wall-clock time for 40 iterations 56.57831 sec

算例文件下载地址 https://pan.baidu.com/s/1Y3TEk31at763xwD7pyCsTQ?pwd=v3tv

@百善孝为先0626 Linux的还是windows的

@百善孝为先0626 破解的，我们被实体清单了，安世亚太说，卖给我们的话，美国政府会罚款

@李东岳 感谢李老师

3200万网格算例在哪里下载，最近fluent2023支持了GPU计算，我打算在DGX 8X V100 32GB 上试一试

@13673619804 你这个像破解不完整，重装试试

楼主链接挂了，补一下

@13673619804 来个截图吧，不止这一句报错吧

@13673619804 细说一下：系统；系统版本；fluent版本