Learning Forum

[mechE]

Hello, 

I am trying to start Fluent 2022 R2 in gpgpu mode (by using the native GPU solver option on the Fluent start screen). However, I see the following error and Fluent stops responding as soon as it starts - 

GPU Solver (Beta) enabled

999999 (C:\ADOagent\_work\2\daily_dev22.2\ansys_inc\v222\fluent\fluent\multiport\src\mpsystem.c@1236): mpt_read: failed: errno = 10054

999999: mpt_read: error: read failed trying to read 4 bytes: Invalid argument

===================================================================================

= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES

= RANK 0 PID 7336 RUNNING AT ANSYS-CL-36

= EXIT STATUS: -1073740791 (c0000409)

===================================================================================

I have the following GPU on my machine - 

CUDA visible GPUs on ANSYS-CL-36

CUDA runtime version 6000

Driver version 9020

Number of GPUs 1

0. Quadro K620

3 SMs

1.124 GHz

2.14748 GBytes

Any ideas on why this doesn't work? Would appreciate any help on this!

[Rob]

It is a beta feature.... Check the gpu is on the supported list, and that you are attached to the Uni licence server. 

[mechE]

Thank you for the response Rob. Based on the Fluent 2022 R2 release notes my gpu should be supported - 

"Fluent supports NVIDIA's CUDA-enabled workstation, data center and server cards." 

Also, I am trying this on campus, so I am connected to the license server for sure. I can start and run Fluent in regular parallel CPU mode without any trouble.

Any other ideas on things to try?

 

 

[Rob]

There's the usual "update the drivers" and also, run "nvidia-smi" from the DOS window. I assume you're not trying to autoload a case, and have 3ddp selected?