RTX PRO 6000 Blackwell Workstation Edition - Driver Support

I just received a PRO 6000 Blackwell Workstation Edition - Driver Support today. Unfortunately, I can’t get to a clean nvidia-smi result, it reports “No devices were found”

I tried to reproduce from a completely clean slate, and got to the same unfortunate result:

  1. Install Ubuntu 24.04.2 LTS
  2. Install Cuda 12.9 from nVidia’s website
  3. Install “cuda-drivers”, which installs nVidia 575.51.03-0ubuntu1
  4. Run nvidia-smi, still the same “No devices were found”

Here’s lspci -v

01:00.0 VGA compatible controller: NVIDIA Corporation Device 2bb1 (rev a1) (prog-if 00 [VGA controller])
Subsystem: NVIDIA Corporation Device 204b
Flags: bus master, fast devsel, latency 0, IRQ 277, IOMMU group 13
Memory at f0000000 (32-bit, non-prefetchable) [size=64M]
Memory at c000000000 (64-bit, prefetchable) [size=128G]
Memory at e000000000 (64-bit, prefetchable) [size=32M]
I/O ports at f000 [size=128]
Expansion ROM at f4000000 [disabled] [size=512K]
Capabilities:
Kernel driver in use: nvidia
Kernel modules: nvidiafb, nouveau, nvidia_drm, nvidia

The card works as expected on Windows.

I just happened to try and run the open driver, and now it works fine. What could be the difference?

From Blackwell, only the open source kernel module can be used. Incidentally the GSP cannot be deactivated (for the same reason).

1 Like

@zebcom did you find the soltuion? I have 4 of those and no luck, i am getting cuda initialization error on everything. Pytorch do not see the GPUs.

Nvidia-smi sees them and i tried cuda 12.6 up to latest 12.9 and nothing.

I have a 5080, it is @nicholaus.shupe who has a Pro 6000. I only explained that you need the open source kernel module with Blackwell, the closed source driver does not work with these GPUs. But it seems your problem is different? Do you use a graphical environment?

Did you guys tried the server driver variant ? Because workstation GPUs need server driver to work properly. Like below :