Best branch for Nvidia hybrid laptop?

General
1

Hello everyone,

I have a Yoga Pro 9I (2024 model 16IMH9) which has an Nvidia 4060 dGPU alongside the Intel Arc graphics onboard the 185H Meteor Lake CPU.

Unfortunately the Nvidia card support still has a couple of super-annoying issues, namely:

  • The Gtk4 rendered in Wayland is broken and you need to force ngl…
  • Whenever I connect a monitor to its USB-C ports (both the Thunderbold and the 3.2gen2 one) it later crashes Wayland, OR if it does not crash Wayland, it later hangs by never waking up from sleep (which also occurs when using X11).

So currently I am on the bluefin-dx-nvidia:stable branch, using X11 and I make sure to never connect external displays (which is quite annoying) in order to avoid the instability that arises from that…

I was wondering:

  1. Does anyone have a HYBRID LAPTOP (not desktop) that “just works” with Nvidia?
  2. What branch are you using?
  3. Is there an a nvidia branch that has their open source version of the driver? How do I switch to that? In fact, how can I tell if the active driver is the proprietary version, or if it is the newer OSS version?
  4. How can I tell if a connected monitor uses Displaylink and could this be related or is it an nvidia issue?

I would really like to at least get rid of the external display issue, as it is a major usability hurdle…

2

I have a laptop with Nvidia 3050 dGPU and Intel Iris Xe onboard. I’m using the aurora-dx-nvidia:latest image with zero issues. I connect/disconnect external displays via HDMI daily.

13th Gen Intel(R) Core(TM) i9-13900H (20) @ 5.40 GHz
NVIDIA GeForce RTX 3050 4GB Laptop GPU [Discrete]
Intel Iris Xe Graphics @ 1.50 GHz [Integrated]
3

Ah, good old HDMI. I do have a port but I never use it.

Do you know if nvidia:latest is using the new OSS driver from nvidia?

4

NVIDIA-SMI 560.35.03 Driver Version: 560.35.03 CUDA Version: 12.6

5

So, running nvidia-smi give me the same:

> nvidia-smi 
Mon Oct 14 15:52:46 2024       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.35.03              Driver Version: 560.35.03      CUDA Version: 12.6     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 4060 ...    On  |   00000000:01:00.0 Off |                  N/A |
| N/A   44C    P8              1W /   55W |     207MiB /   8188MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
...

My question is, how do I know if this is the open source driver? I believe Nvidia has started using open source kernel drivers and recommends using those going forward?

I know I am not using the old nouveau driver via lsmod:

 ~> lsmod | grep -E "nouveau|nvidia"
nvidia_wmi_ec_backlight    12288  0
nvidia_drm            135168  3
nvidia_modeset       1650688  4 nvidia_drm
video                  81920  5 nvidia_wmi_ec_backlight,ideapad_laptop,xe,i915,nvidia_modeset
wmi                    32768  4 video,nvidia_wmi_ec_backlight,wmi_bmof,ideapad_laptop
nvidia_uvm           6848512  4
nvidia              72577024  81 nvidia_uvm,nvidia_modeset

What I don’t know is whether the above nvidia_... modules are from the old closed source driver, or if they are from the newer open source “nvidia” driver.

I have found that there is an issue merged feat: build nvidia open source kernel module by p5 · Pull Request #220 · ublue-os/akmods · GitHub about this, but not sure how you actually use the new “open” drivers.