[GH-ISSUE #9] NVIDIA DGX Spark installation fails because hard dependency nvidia-vfx has no linux_aarch64 wheel #5

Open
opened 2026-10-26 22:27:48 +00:00 by bsakaguchi · 3 comments
Owner

Originally created by @SpecterHi on GitHub (Mar 12, 2026).
Original GitHub issue: https://github.com/Comfy-Org/ComfyUI_Nvidia_VFX_Nodes/issues/9

ARM64/aarch64 installation fails because hard dependency nvidia-vfx has no linux_aarch64 wheel

Body

Summary

I am trying to install Nvidia_RTX_Nodes_ComfyUI on Linux ARM64 (aarch64), but installation fails because the project has a hard dependency on nvidia-vfx, and nvidia-vfx currently does not provide a linux_aarch64 wheel.

From what I can see, this is not a ComfyUI Manager issue. It is a platform compatibility issue caused by the required dependency.

Environment

  • OS: Linux
  • Architecture: aarch64
  • Python: 3.12.3
  • CUDA: 13.0
  • NVIDIA Driver: 580.126
  • ComfyUI: 0.16.1

Reproduction

Install the node, or directly run:

pip install --extra-index-url https://pypi.nvidia.com/ nvidia-vfx

Actual result

Installation fails with:

RuntimeError: Didn't find wheel for nvidia-vfx 0.1.0.1

The debug output shows that the package is checking only these wheel targets:

  • cp310 ... manylinux ... x86_64
  • cp311 ... manylinux ... x86_64
  • cp312-abi3 ... manylinux ... x86_64
  • Windows amd64

There does not appear to be any linux_aarch64 wheel.

Additional findings

This repository appears to require nvidia-vfx as a hard dependency:

  • requirements.txt includes nvidia-vfx
  • pyproject.toml includes nvidia-vfx
  • the node imports nvvfx directly at startup

Also, the published nvidia-vfx source package is a wheel-stub package that downloads a prebuilt wheel from https://pypi.nvidia.com/, so this does not look like something users can build locally on ARM from source.

Expected result

One of the following would help:

  1. Provide linux_aarch64 support for nvidia-vfx, if ARM64 is intended to be supported.
  2. Explicitly document that Nvidia_RTX_Nodes_ComfyUI is currently x86_64-only.
  3. Fail with a clearer platform error message in the README or installer path.

Request

Could you confirm whether ARM64/aarch64 support is planned?

If not, please consider documenting that this node currently requires x86_64 Linux or Windows, so ARM users do not spend time debugging an installation path that cannot succeed.

Originally created by @SpecterHi on GitHub (Mar 12, 2026). Original GitHub issue: https://github.com/Comfy-Org/ComfyUI_Nvidia_VFX_Nodes/issues/9 ARM64/aarch64 installation fails because hard dependency `nvidia-vfx` has no `linux_aarch64` wheel # Body ## Summary I am trying to install `Nvidia_RTX_Nodes_ComfyUI` on Linux ARM64 (`aarch64`), but installation fails because the project has a hard dependency on `nvidia-vfx`, and `nvidia-vfx` currently does not provide a `linux_aarch64` wheel. From what I can see, this is not a ComfyUI Manager issue. It is a platform compatibility issue caused by the required dependency. ## Environment - OS: Linux - Architecture: `aarch64` - Python: `3.12.3` - CUDA: `13.0` - NVIDIA Driver: `580.126` - ComfyUI: `0.16.1` ## Reproduction Install the node, or directly run: ```bash pip install --extra-index-url https://pypi.nvidia.com/ nvidia-vfx ``` ## Actual result Installation fails with: ```text RuntimeError: Didn't find wheel for nvidia-vfx 0.1.0.1 ``` The debug output shows that the package is checking only these wheel targets: - `cp310 ... manylinux ... x86_64` - `cp311 ... manylinux ... x86_64` - `cp312-abi3 ... manylinux ... x86_64` - Windows `amd64` There does not appear to be any `linux_aarch64` wheel. ## Additional findings This repository appears to require `nvidia-vfx` as a hard dependency: - `requirements.txt` includes `nvidia-vfx` - `pyproject.toml` includes `nvidia-vfx` - the node imports `nvvfx` directly at startup Also, the published `nvidia-vfx` source package is a `wheel-stub` package that downloads a prebuilt wheel from `https://pypi.nvidia.com/`, so this does not look like something users can build locally on ARM from source. ## Expected result One of the following would help: 1. Provide `linux_aarch64` support for `nvidia-vfx`, if ARM64 is intended to be supported. 2. Explicitly document that `Nvidia_RTX_Nodes_ComfyUI` is currently `x86_64`-only. 3. Fail with a clearer platform error message in the README or installer path. ## Request Could you confirm whether ARM64/aarch64 support is planned? If not, please consider documenting that this node currently requires `x86_64` Linux or Windows, so ARM users do not spend time debugging an installation path that cannot succeed.
Author
Owner

@jovan2009 commented on GitHub (Mar 12, 2026):

These are all the wheels that nvidia provides at this time
https://pypi.nvidia.com/nvidia-vfx/
Image
If none of them work on your system probably it's only nvidia that can do something about it. Being version 0.1.0.1 I guess it is still pretty early in the development cycle.

<!-- gh-comment-id:4045478799 --> @jovan2009 commented on GitHub (Mar 12, 2026): These are all the wheels that nvidia provides at this time https://pypi.nvidia.com/nvidia-vfx/ <img width="604" height="184" alt="Image" src="https://github.com/user-attachments/assets/0fd5e842-99a2-4ee6-b9bc-cc1099d4217b" /> If none of them work on your system probably it's only nvidia that can do something about it. Being version `0.1.0.1` I guess it is still pretty early in the development cycle.
Author
Owner

@PhonicUK commented on GitHub (Mar 16, 2026):

+1 here, took me a while to even digest that the issue was down to NVIDIA not shipping ARM64 binaries, so not usable on DGX Spark / DGX Station

<!-- gh-comment-id:4070831537 --> @PhonicUK commented on GitHub (Mar 16, 2026): +1 here, took me a while to even digest that the issue was down to NVIDIA not shipping ARM64 binaries, so not usable on DGX Spark / DGX Station
Author
Owner

@os-sos commented on GitHub (Jul 22, 2026):

Looking forward to a solution to this. Hopefully they catch up to the DGX Spark hardware support. It's a pretty big community.

<!-- gh-comment-id:5045694016 --> @os-sos commented on GitHub (Jul 22, 2026): Looking forward to a solution to this. Hopefully they catch up to the DGX Spark hardware support. It's a pretty big community.
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
Comfy-Org/ComfyUI_Nvidia_VFX_Nodes#5
No description provided.