GrayJay FlatHub worked on Kubuntu 25.04 frozen screen on Kubuntu 26.04

[JoshiFresa]

I don't think Nvidia video drivers are broken, as I mentioned ,you need to run nvidia-settings with sudo privileges for all the settings and features.
I think we are going into the weeds unnecessarily, apologies for dragging things down there.
I have no experience with Cuda, that itself could well be broken, if the installation method you used specifically for Cuda does not yet support this very new kernel, and *buntu still being in a pre-release state.
So, what does it show now?
We really need someone with Nvidia experience here. I don't know where to go or what to ask, in terms of data gathering.

Good to hear I can't damage my Nvidia card, some people mentioned such things so it ringed possible lol. About Cuda I had installed Cuda in the working installation but took it away in the new kernel now. I will install it although now I tried to do a good form installation and failed again based on this page: https://linuxcapable.com/install-nvi...-ubuntu-linux/

nvidia -smi now says "command not found" again

I guess this new kernel is still not ready to receive some Nvidia cards since also Nvidia only support up to Ubuntu 24.04 if I remember correctly, so it's about being patient or go back to Kubuntu 24.04. Maybe installing Cuda may help and miracles often happen to me but thanks for all the support. I learned a lot from you. There's a spritual aspect in the way people connect in the web. You interact with someone you don't know but the ego is not so important, it's like a big plasma of collective open source identity with many eyes and ears that moves in waves.

[claydoh]

I tried to do a good form installation and failed again based on this page: https://linuxcapable.com/install-nvi...-ubuntu-linux/

But read the full page mentioning 26.04 and cuda. It is accurate.

nvidia -smi now says "command not found" again

nvidia -sm- or nvidia-smi (no space)? The latter is the correct command.

[JoshiFresa]

Yes I was aware of the 26.04 disclaimer. Yet I tried my chance just in case it may work with an older package. But so far everything failed.

Oh and that nvidia-smi, I was writing it wrong all the time and it explains it all. Maybe I had success sooner but could not appreciate it because of me writing the wrong command I will erase, purge, clean everything as in this great tutorial again and try one more time without expectations and if nothing works I'll either have to reinstall 24.04 or wait for Nvidia and Kubuntu to upgrade it.

[JoshiFresa]

I did the whole uninstall/install of Nvidia again and this time it worked! I guess the whole time I was typing nvidia-smi wrong. I did not even install the cuda toolset and yet my AI app ComfyUI recognizes the Nvidia card and CUDA.

But.. If I try to check my CUDA status with "nvcc --version | grep "release" | awk '{print $6}' | cut -c2-" or simply "nvcc --version" I get this answer:

Command 'nvcc' not found, but can be installed with:
sudo apt install nvidia-cuda-toolkit

I guess CUDA can work even if the Cuda-toolkit is not installed (??)

Then when I tried to install CUDA with the tutorial in https://linuxcapable.com/how-to-inst...-ubuntu-linux/ with the command "sudo apt install cuda-toolkit nvidia-open -y" I get this answer:

Error: Unable to locate package cuda-toolkit
Error: Unable to locate package nvidia-open

So instead of trying different things I prefer to submit it here. Also if I understand well the CUDA-toolkit is a proprietary file from Nvidia but Ubuntu should recognize. In my case it doesn't when I do "sudo apt update", probably because it's Kernel 26.04. Not sure if I should install the Kuda-toolkit from a file. Many questions.

[Snowhog]

I tried to install CUDA with the tutorial in https://linuxcapable.com/how-to-inst...-ubuntu-linux/

That tutorial is for 24.04 and 22.04. Why would you think it would work on 26.04?

[JoshiFresa]

Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_mxfp8', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4']}
Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']}
Found comfy_kitchen backend cuda: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
Checkpoint files will always be loaded safely.
Total VRAM 3770 MB, total RAM 31822 MB
pytorch version: 2.9.1+cu130
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 3050 Laptop GPU : cudaMallocAsync
Using async weight offloading with 2 streams
Enabled pinned memory 28639.0
working around nvidia conv3d memory bug.
Using pytorch attention
aimdo: src/control.c:68:INFO:comfy-aimdo inited for GPU: NVIDIA GeForce RTX 3050 Laptop GPU (VRAM: 3770 MB)
DynamicVRAM support detected and enabled
Python version: 3.12.10 (main, Apr 7 2026, 23:43:48) [GCC 15.2.0]
ComfyUI version: 0.18.1
comfy-aimdo version: 0.2.12
comfy-kitchen version: 0.2.8
ComfyUI frontend version: 1.42.8
[Prompt Server] web root: /home/ME/Desktop/ComfyUI-Easy-Install/ComfyUI-Easy-Install/python_embeded/lib/python3.12/site-packages/comfyui_frontend_package/static
Asset seeder disabled
[FishAudioS2] Models folder registered: /home/ME/Desktop/ComfyUI-Easy-Install/ComfyUI-Easy-Install/ComfyUI/models/fishaudioS2
[FishAudioS2] Registered 4 nodes (v0.4.5): Fish S2 TTS, Fish S2 Voice Clone TTS, Fish S2 Multi-Speaker TTS, Fish S2 Multi-Speaker Split TTS


THIS IS THE LOG OF COMFYUI, AI TOOL. IT FINDS CUDA BUT IT DOES NOT SEEM TO RATE IT. CUDA=0. YET IT WORKS PERFECTLY.