GPU usage issues - Is nVidia K2100M too old?

[merkato]

Trying to use nVidia Quadro K2100M as an GPU for node-odmat Ubuntu 22.04. Installed all needed things (nvidia-driver-470, nvidia-docker2 and dependencies). As an result can get nvidia-smi in docker container:

sudo docker run --rm --gpus all nvidia/cuda:11.0.3-base-ubuntu20.04 nvidia-smi
Wed Oct 19 06:20:42 2022       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.141.03   Driver Version: 470.141.03   CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro K2100M       Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   40C    P8    N/A /  N/A |    276MiB /  2000MiB |     15%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

Started WebODM with

./webodm.sh start --gpu

And getting this:

* 06:14:02 [App ] Command line: DensifyPointCloud /var/www/data/8f5a74e9-9107-44ea-8e39-fb085a140cb5/opensfm/undistorted/openmvs/scene.mvs --resolution-level 1 --dense-config-file /var/www/data/8f5a74e9-9107-44ea-8e39-fb085a140cb5/opensfm/undistorted/openmvs/Densify.ini --min-resolution 1024 --max-resolution 2048 --max-threads 8 --number-views-fuse 2 --sub-resolution-levels 2 -w /var/www/data/8f5a74e9-9107-44ea-8e39-fb085a140cb5/opensfm/undistorted/openmvs/depthmaps -v 0 --geometric-iters 0 06:14:03 [App ] The camera directions mean is unbalanced; the scene will be considered unbounded (no ROI) 
06:14:12 [App ] Preparing images for dense reconstruction completed: 181 images (9s557ms)
06:14:13 [App ] Selecting images for dense reconstruction completed: 181 images (570ms) 
[WARNING] OpenMVS failed with GPU, is your graphics card driver up to date? Falling back to CPU. 
[INFO] running "/code/SuperBuild/install/bin/OpenMVS/DensifyPointCloud" "/var/www/data/8f5a74e9-9107-44ea-8e39-fb085a140cb5/opensfm/undistorted/openmvs/scene.mvs" --resolution-level 1 --dense-config-file "/var/www/data/8f5a74e9-9107-44ea-8e39-fb085a140cb5/opensfm/undistorted/openmvs/Densify.ini" --min-resolution 1024 --max-resolution 2048 --max-threads 8 --number-views-fuse 2 --sub-resolution-levels 2 -w "/var/www/data/8f5a74e9-9107-44ea-8e39-fb085a140cb5/opensfm/undistorted/openmvs/depthmaps" -v 0 --cuda-device -2 --geometric-iters 0

Worth to mention that openmvs showed up for few seconds in nvtop. Is this GPU already too old?

[APOS80]

I think there need to be a pretty resent Cuda driver but I’m not sure.

I got the 3060 12gb that works but that’s a pretty new gadget

[merkato]

Ok.

Looking for info:

• What minimal version of nvidia-driver for Ubuntu is needed?
• Which CUDA Toolkit version?

[shiva]

Running on Ubuntu 20.04.5 LTS, using a NVIDIA Geforce GTX 1660 Super.
Driver is nvidia-driver-515 with nvidia-cuda-toolkit in version 10.1.
Both packets from official Ubuntu repositories.

The card you mention is pretty old. A modestly recent 8 core CPU would probably be faster and
the 2Gbyte of VRAM will limit the image size it can process.