Jobs suddenly remain queued indefinitely

[parsley]

Hi,

thank you for the reply. I think this may indeed be the issue. The simulation seems to start running fine when I use the Software kernel. I then tried running the Dipole Antenna example with aXware and it also failed, while it works without GPU.
While the log shows my GPU:

OpenGL view 4.6.0 NVIDIA 536.67[Quadro GV100/PCIe/SSE2], backwards compatible, 8 multi-samples
GPU: Quadro GV100
Vulkan view 0.1.3.242[Vulkan - Quadro GV100 (2249244672)], 8 multi-samples
13:57:58] Checking out license feature 'MODELING', expiring the 1-nov-2025, version 1.3, (1).
produced by Sim4Life version 8.2.0.16876
to local Ares successfully established.
14:07:39] Checking out license feature 'POSTPRO', expiring the 1-nov-2025, version 8.2, (1),

It is not showing up as in the S4L documentation when I open Ares, see attachment.

In case it helps, I also ran the Acceleware Diagnostic Tool and it looked fine (see attachment).

[parsley]

Hello again,

after doing some further testing I found that my GPU is found by Ares for an older Version of S4L (V.6.2.1.4972) but not for newer ones. In the newer versions, I now also get an error message upon startup:

Error : [2025-May-19 11:25:27] No such feature exists. (-5,412) - [(license path)]

I ran the AxDiagnostic tools also for V.6.2.1.4972 and only found a driver conflict for V.6.2.1.4972 as one would expect (the image from above was for V8.2.2.18061).

I'd really appreciate any advice or suggestions!

[brown]

Hi @parsley ,

Is it possible that you selected the 'Install Deprecated Drivers' option when installing V8.2.2.18061? Doing so with your GPU (Quadro GV100) could have introduced an incorrect driver version. If that is the case, you would have to uninstall that version of Sim4Life, then reinstall while ensuring that you do not check the deprecated drivers option.

[parsley]

Hi @brown ,
thanks for the reply. I have tried both options and to be sure I just reinstalled without deprecated drivers one more time, but to no avail. I have also set AX_USE_UNSUPPORTED_CARDS=TRUE as stated in the manual. I also ran AxDiagnostics again and I'm attaching the detailed results in case it helps.

[brown]

Hi @parsley ,

Could you also share the location of nvidia-smi.exe on your PC? This is used by the software to get GPU information.

[parsley]

Hi,
yes, see below. I've also included some other files that ChatGPT suggested may be relevant in case it helps.

[brown]

Thanks @parsley .

If Ares cannot locate your GPU, it could be due to the installation location of nvidia-smi.

Could you please try backing up the file ExtractCudaInfo.py installed at <S4L install>\scripts\tools, and then try again while replacing it with the attached version instead?

ExtractCudaInfo.py

[parsley]

Unfortunately this did not work, also not after rebooting.

[brown]

@parsley , could you please contact s4l-support@zmt.swiss to organize a call to address your issue?

[parsley]

After a call with S4L support, here is the solution for future reference:
My machine had two GPUs connected, one of them old and deactivated. However, the script ExtractCudaInfo.py detected both of these in the line 'num_gpus = int(xml_root.find("attached_gpus").text)' and then got stuck trying to find information on the deactivated GPU. A quick fix for the issue is therefore to manually set the correct number, in my case 'num_gpus=1', in the script.