-
Notifications
You must be signed in to change notification settings - Fork 19
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to get GPU acceleration working? (Debian 12.2, Gnome, Wayland, X11, Nvidia P1000 GPU, Zbook Studio G5) #48
Comments
Thanks for the report. Could you please run the app with
When app starts, you should see log Edit: sorry, I closed this issue by mistake |
Ok, here we go, I just post everything, there are quite a few error messages. There is one more strange thing: sometimes, on rare occasions, the GPU acceleration just works for some time - then ist crashes, and it is not accessible any more. The same applies after a reboot - the GPU accelaration is gone, even wehn it worked just fine for a few hours. I dont know what triggres it... `flatpak run net.mkiol.SpeechNote --verbose ` |
In theory your video card is supported and should work fine. For some reason CUDA reports an "unknown" error (
It looks that your NVIDIA driver is not the latest one. My system has "driver=12020". I don't think this resolves the problem but you may try to update the driver on your machine.
You may try as well "beta" version from Please let me know if any of this helped. |
Ok, ust one strange thing: Now, after an Uptime of roughly 19 hours, the GPU is suddenly recognized by Speech Note. I can use it freely now. I did nothing to fix it in the meantime, I don't get it... |
The Issue persists in the beta build as well. |
Ok, I found the bug! After a fresh boot, Speech Note does not recognize the GPU - unless the GPU is used by an app. Hoe it helps somehow. Here, the terminal output after using Steam: |
I've never encountered anything like this 🤔 . My first thought was that your GPU uses some form of power control and disables "heavy features" in non-gaming mode. On laptop this kind of power optimization makes sense.... but the truth is that I have no idea. |
ok, lets leave it at that. I have my workaround. Some kind of power saving while the GPU is in auto mode in the Zbook-BIOS - this is my guess too. Greetings, and thanks for your work, Tomasz |
I think this is spot on - this technology is called NVIDIA PRIME. The device uses a weaker integrated Intel GPU by default and starts the more powerful NVIDIA card when you run games. @tkonicz you should be able to modify that behavior in the NVIDIA Settings application. For example, to turn the NVIDIA GPU permanently on (that will cost you battery life). Or get it to start whenever you launch Speech Note. |
This might be a solution for PRIME switching: You need to set
and than just start the app normally with:
This trick resolved this problem. |
Hi, just a quick question: it this a permanent change, remaining even after a reset? And, will now every flatpak app use NVIDIA, or just SpeechNote? |
reference: https://docs.flatpak.org/en/latest/flatpak-command-reference.html#flatpak-override It will affect only
|
Nope, it does not work here. Speechnote did not display at all after applying the override. Had to reset it. |
Can you please check what outputs:
? |
If the problem still occurs, please re-open. |
Hello,
I got the "Speech Note" Flatpak working on my Debian 12 system (Zbook Studio G5). I can use Whisper in offline mode here. After downloading Whisper (large and/or medium), the speech recognition is quite good, but very slow (50 sec.). GPU acceleration would help, so I installed the Nivida drivers for my P1000. They work just fine with games, eg., but not with "Speech Note" and Whisper. Any ideas how to fix this? How do I get my Nvidia card to accelerate the speech recognition of whisper on Debian 12? Maybe this is a bug?
My Nvidia Driver Version: 525.125.06
I have already libcudart11.0 and nvidia-cuda-toolkit installed.
I tried both Wayland and X11.
My Card, the P1000, seems to supports CUDA 6.1 - this should be enough?
Terminal output, when starting Speech Note:
The text was updated successfully, but these errors were encountered: