[iancmlositano]

Hi guys! Greatly need a BIG hand here. I have never successfully setup parallel computing with GPGPUs on ANSYS Fluent. I have these to note:

1. All NVIDIA with CUDA cores is detected by ANSYS Fluent 18.0 AMG solver and very much compatible;
2. GPU parallel computing allows up to five (5) coupled equations; and
3. CFD pressure coupled solver works fine with GPU computing.

I am running a sliding mesh VAWT simulation with a mesh of approximately 2 million cells on a 3.41 GHz 4-core PC with 32 Gb RAM. Without the GPU setting, the simulation works well although at a very low computing turnaround - around 6 to 7 minutes per time step. Setting the GPU up does not run the simulation. Instead I get this error report in the Fluent command line every time:

Thrust failure: out of memory
File and line number are not available for this exception.
AMGX ERROR: file amg\amg_gpu.c line 1027
AMGX ERROR: Thrust failure.
The fl process could not be started.
MPI Application rank 0 exited before MPI_Finalize() with status 1

Maybe someone out there has made this work. Will appreciate help making me understand the situation or best make GPU computing work for my case.

Thanks!

[LuckyTran]

The error is here:

Originally Posted by iancmlositano

Thrust failure: out of memory

Your simulation simply doesn't fit on the GPU memory. Find ways to make it smaller.

[iancmlositano]

Originally Posted by LuckyTran

The error is here:

Your simulation simply doesn't fit on the GPU memory. Find ways to make it smaller.

I see. Thank you very much!