I was revisiting some of the threads here and wanted to add that I posted this question to ChatGPT and it produced a solution that you may want to try.

Have you removed the 'house' Ubuntu drivers and installed the Nvidia Drivers?

If not, then do that :
sudo vim /etc/modprobe.d/blacklist-nouveau.conf

Then add this to that file.
blacklist nouveau
options nouveau modeset=0

Regenerate the kernel initramfs
sudo update-initramfs -u

Reboot the machine, then install the Nvidia drivers.

Nvidia Driver Page
https://www.nvidia.com/en-us/drivers/unix/

The OS should be smart enough to 'share the load' between applications.

Phil

I can't do it at the moment.

So if you stop all other services....does it work then?

if the video card is used by any process, the platform will not start. I still have receivers on another service and an additional service for decoding with GPU support will be launched

I do not know..

But I wonder, why such a energy-consumption system to power Phantom?

There is no need to use that much GPU.

I'm trying to understand what you are looking for as typical a 0-30MHz websdr only needs an Intel 6600T or better and the GPU inside the CPU is more then capable of processing all.

I tried an nVidia card (old one) but never got it working via Cuda.

nvidia-smi
Sun Jun 15 14:56:40 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 575.51.03              Driver Version: 575.51.03      CUDA Version: 12.9    |
|-----------------------------------------+------------------------+----------------------+
| 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  NVIDIA GeForce GTX 1650        Off |  00000000:04:00.0 Off |                  N/A |
| 49%  54C    P3            32W /  90W |    446MiB /  4096MiB |    50%      Default |
|                                        |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
|  1  NVIDIA GeForce GTX 1650        Off |  00000000:06:00.0  On |                  N/A |
|  0%  43C    P8              8W /  75W |      12MiB /  4096MiB |      0%      Default |
|                                        |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU  GI  CI              PID  Type  Process name                        GPU Memory |
|        ID  ID                                                              Usage      |
|=========================================================================================|
|    0  N/A  N/A          79438      C  ./build/spectrumserver                  442MiB |
+-----------------------------------------------------------------------------------------+

or the platform itself finds a free card ?

Did you install OpenCL, Cuda or such?

As you need to have it installed and WORKING before compiling.

Then enable it in the TOML and it should use the GPU.

nVidia chips can also do OpenCL, las I heard.

However, I never got Cuda to work. As such I use Intel with onboard GPU, more then fast enough via OpenCL.

how to configure to use one of two video cards on nvidia chip ?

And they managed again....different line....so I adopted the iptables:

iptables -D INPUT -m string --algo kmp --string "%3C%" -j DROP
iptables -D INPUT -m string --algo kmp --string "device.rsp" -j DROP
iptables -A INPUT -m string --algo kmp --string "%3C%" -j DROP
iptables -A INPUT -m string --algo kmp --string "device.rsp" -j DROP
As they attack the device.rsp all the time, I added that too.

Also replaced algo with KMP as KMP can scan over multiple TCP-packets, where BM may miss it.

Hopefully it's over now.