mirror of
https://github.com/Comfy-Org/ComfyUI_Nvidia_VFX_Nodes.git
synced 2026-11-04 23:37:30 +00:00
[GH-ISSUE #8] NvVFX_Load failed: The requested feature or capability was not found (code -14) #6
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
Comfy-Org/ComfyUI_Nvidia_VFX_Nodes#6
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @SurpassHR on GitHub (Mar 12, 2026).
Original GitHub issue: https://github.com/Comfy-Org/ComfyUI_Nvidia_VFX_Nodes/issues/8
Got error on my RTX 4070S.
ComfyUI Version: c2d9a4fe3cd4b3b3534b2d4037043ce9064c1e5f
@RookieZoe commented on GitHub (Mar 12, 2026):
Same issue here (RTX 4090):
@iCELAKE01 commented on GitHub (Mar 12, 2026):
Experiencing the same on 3070, tried reinstalling but to no avail
@caissonfiv commented on GitHub (Mar 12, 2026):
same，4090D
@jonnie commented on GitHub (Mar 13, 2026):
I have exactly the same problem on my RTX 5090 running on Ubuntu. I tried running the example code (from here: https://pypi.org/project/nvidia-vfx) outside of ComfyUI and that too has the same problem. I've checked that my machine, drivers, etc. meet the requirements. Unfortunately, nvidia-vfx seems to be closed source so cannot investigate why this is happening.
@iCELAKE01 commented on GitHub (Mar 13, 2026):
there is no reason for it to be a wsl2/ubuntu issue right?
@jorgemcgomes commented on GitHub (Mar 13, 2026):
I don't think the problem is related to Comfy at all.
I'm getting the same error running the official example in Google Colab:
https://colab.research.google.com/drive/1mF-d2PcptP59Si6azOEGAbX25jLHPNgL?usp=sharing
Tried Nvidia L4, A100, and G4 (RTX PRO 6000), all fail with this same error.
There is probably an issue with some dependency, but it's not clear at all what it might be since the error is cryptic.
Also probably not the best place to post this, but Nvidia-VFX isn't open source so...
@RookieZoe commented on GitHub (Mar 20, 2026):
Update:
I Just wipe NVIDIA Driver out，and reinstall Driver and CUDA to latest version:
and then, Rebuilt the comfyui runtime environment, and use
comfy --here install --restorereinstall comfyui，then, handle the problematic dependencies.After these tedious steps, the error disappeared and it is now running normally.
@rahdirigs commented on GitHub (Mar 20, 2026):
If its the same error code (-14) that you are getting, then I don't think the message is as cryptic as the other code (-12). The error message (for code -14) states the device lacks the capability, so my first hunch would be to check the driver versions. On the pypi project it states that:
is required. In my case I had 580.75, so I upgraded to 580.105, and the error went away. Since you are using colab, I am not sure if upgrading drivers is an option.
@SurpassHR commented on GitHub (Mar 20, 2026):
That makes sense, but newer drivers are not stable at all, I'm tired of testing NV drivers on linux, so just let it be, the upscaling is fast but not good enough to make me reinstall all of these.
@jorgemcgomes commented on GitHub (Mar 20, 2026):
In the Colab environment:
Which matches the project requirements. So who knows. The requirements might be wrong, or it might be an issue somewhere else.
@rahdirigs commented on GitHub (Mar 20, 2026):
a) its never going to work on T4 on Colab, since its Tesla architecture
b) you can try running
!ldconfig -p | grep nvidiaand verify if you see anything. TL;DR it needs the nvidia specific linkers and not just CUDA / TensorRT. if you don't see anything then the (-14) error code is valid. if you see some entries, you need to look for specificallylibnvidia-ngx.sowithout this linker, even if everything else is present, you will run into -12 error code.@SurpassHR commented on GitHub (Mar 21, 2026):
The Tesla T4 is based on the Turing arch, not the Tesla arch. I think the real reason is that @jorgemcgomes needs to use a driver version equal to or newer than 580.83, 580.82.xx is not included in 580.82+.