I've have a G14 with RTX2060 and I'm very happy. Battery life is good (4-6h work with IDE etc) and webcams work as intended. DRM is mostly a non-issue, at least not more than on any other Linux install.
The nvidia driver is a bit of a pain to set up, especially if you're on wayland or if you want to use displays via USB C (reverse prime does not work yet due to an nvidia driver bug, so you need to use the nvidia gpu as primary when docked). Other than that, everything works and the community (arch wiki and rog-core) provide good support for getting everything up and running. A bit of configuration and a somewhat recent kernel will be needed (it's pretty new hardware after all), but it's not that hard if you either know Linux or are willing to spend some time. I'm running this config for 8 months now, so the situation from a fresh install might be even better.
Similar to the sister comment, I came from an XPS 13, but I'm happy. The laptop is a bit heavier, but in exchange you get a lot more power, far more RAM (up to 48GB), more ports and, subjectively, a dar better keyboard. Initially I wanted to stay with an XPS, but now I'm happy I made the jump.
EDIT: One unbelievably good point I initially forgot: You can run a VM with GPU passthrough on the laptop. If you need Windows with graphics performance, especially on the go, this is an incredible advantage.
>EDIT: One unbelievably good point I initially forgot: You can run a VM with GPU passthrough on the laptop. If you need Windows with graphics performance, especially on the go, this is an incredible advantage.
Can you speak more to this? Does Pop_OS! configure this (more-or-less) out of the box? Would it work with any Nvidia card or are there restrictions?
I'm interested in such a setup and have done research from afar. I've developed on Linux for years, but have heard GPU passthrough can be difficult unless you wrangle drivers a bit.
EDIT: I'd also like to use it for CUDA in Linux. Do you have to reboot when enabling GPU passthrough (not a dealbreaker, just curious)?
Sure! Regards working out of the box, you'll probably have no luck with that. VFIO is still somewhat of a niche and VFIO on a laptop is a niche of a niche, so you'll need to expect some command line usage and a bit of config file editing. This being said, getting it to work is not too hard and, if you've done VFIO in the past, you should feel right at home.
For the initial setup, you can mostly follow the guide on the arch wiki [0]. You might need to adapt it a bit on ubuntu(-based) systems, but the major points are the same. The only difference to the usual setup is that you need to include an APCI-table [3] to get the laptop driver to work in the VM. This reddit thread [1] describes it quite well. The author of that thread also has a repo [2] where he published his files and on which I based my scripts on. Note that you need the usual CPUID work-around (see the "Error 43" section on the arch vfio page). Also, you'll need an USB C to HDMI/DisplayPort/... cable and a monitor for the initial setup, after that you can use a dummy plug [5] and looking glass (also described on the arch wiki) to use pass through without external monitor. The setup is also very stable for me, with the exception of the looking glass part - if you absolutely depend on the setup, having a monitor handy will save you a bit of headache, but you can always revert to a VM display in the worst case.
Regarding GPU support: It should support all GPU configurations, but you should check reddit etc to be sure. If you're talking about other laptops, the situation is a bit different. The big advantage of the G14 is (1) that the GPU has its own IOMMU group and is therefore easy to isolate and (2) that the display ports are well split between the two GPUs (the HDMI out and the internal display are connected to the AMD iGPU and the USB C DP-channel is connected to the RTX). This is rather rare, but makes setting this up (comparatively) easy. Note that the USB channel of the USB C out is not routed via the GPU, so if you're connected to a docking station, its display ports will be used by the VM, but the USB ports are on the host, still. This, however, comes with the drawback that you can not pass through an USB controller, as the groups for those do not work out. The ACS override patch (see arch wiki) might fix this, but unless you need to use audio interfaces, the "normal" USB redirector will suffice anyway.
Rebooting is semi-required. In order to change from VM to Linux and back, you'll need to [un-]bind the nvidia driver and switch to vfio-pci. Going from Linux to VM is only possible when nothing uses the nvidia driver, which either needs a restart of your display server or only using the iGPU for all apps (quite easy on sway/wayland, haven't tried on X). The reverse should work as well - the author of the reddit threads above has scripts to do so, but this did not work out for me. Leaving the vfio-pci driver for the GPU is not an option either, as this will prevent it from going into the low power state and therefore heat up the laptop and use up the battery. Because of this, I usually simply reboot. In summary, if you don't mind rebooting, having the GPU bound to vfio-pci via kernel parameter on a secondary boot option is by far the easier way :) But it should be possible without, it'll just be a bit of work.
To work with CUDA, you'll need to have the nvidia driver loaded, but you'll want to have that anyway to enable low-power states, as mentioned above. Using only the iGPU for display is possible in combination with CUDA, you'll simply need to stop your CUDA-apps before unbinding the driver. That's actually the setup I use, too :)
I have the Dell G5 4800H (AMD iGPU+AMD dGPU), there are a couple of graphic driver issues under Linux. I'm using the USB-C connection which is connected to the iGPU.
* Locks up on suspend/resume
* Locks up when inserting an external USB-C monitor
* Requires a recent kernel to work properly (5.8 or later)
* Locks up without amdgpu.runpm=0 kernel command line
* dGPU can't be turned off, so increases power consumption
* Locks up with vsync on with the dGPU
* Kernel warning on startup, but doesn't affect stability
Besides these issues (which can be worked around), it's very solid.
It's worked great so far except for plugging into Dell's USB-c dock in the morning. This is a guessing game of which (of 3) monitors are going to work, often requiring opening/closing the lid a time or two or turning on/off screens.
Battery life is decent (VERY good if you turn down brightness and ssh to your work) and it's blazing fast. It does get egg cooking hot when unplugged and running the GPU (games).
Running Win10 currently, but trying to figure out which Linux is going to run with 3 monitors without doing hours of research. Suggestions welcome. Guess I'll try PopOS next.
I have a Dell monitor which has a built-in USB-C hub. I experience similar issues when plugging in my MacBook Pro - the monitor part works fine, but the keyboard/mouse usually takes a few attempts at plugging in ("docking"). Might not be your laptop so much as Dell's USB-C dock acting up.
I use it with Fedora happily, and for work I have to use Zoom daily. I did have to replace a fan with one from Aliexpress due to rattle, but haven’t had issues since.
- What kind of external webcam do you use? Does it have any problem with any video-conferencing software?
- Do you access any streaming service (with DRM)? Any issues?
- Any issues with drivers, returning from sleep/hybernation?
Thanks!