I have a 2013-era laptop with a discrete NVIDIA GPU. The most graphically intensive tasks I will be doing on it are running a desktop window manager and watching online videos.
I have previously run Ubuntu on this laptop with the NVIDIA drivers, and that involved manually signing my kernel and the DKMS NVIDIA modules after every update. I would like to avoid this process this time around if possible, even if that involves not using the NVIDIA GPU at all.
What should I set up for the least amount of trouble? If I don’t install the proprietary NVIDIA drivers at all, will this impact the laptop’s battery life (by virtue of the GPU not powering down)?
It has been reported that using the discrete nvidia gpu as primary for the laptop does shorten battery life. Mine, with routine use, still gets at least 3 hours on the battery with using nvidia drivers and with the nvidia gpu set as primary so it is used constantly. Obviously your use determines how much you want the nvidia gpu active. It would be up to you and your usage demands if a slightly shortened battery life is acceptable for the convenience of better overall graphics.
With fedora, following the instructions of the Howto/NVIDIA - RPM Fusion
site to install and enable the nvidia drivers would work very well. Fedora uses a tool called akmods that automatically rebuilds the driver modules as needed each time the kernel is updated and as long as the rpmfusion repo is enabled each driver update will also be installed and compiled when you do an update.
Installing the nvidia drivers does not automatically force the nvidia gpu to be active all the time so that should not impact battery life unless you choose to make it active full time.
Thank you for your response.
The solution I went with was, well, not installing any drivers! It turns out the open-source Mesa+Nouveau stack, which was installed out-of-the-box, is exactly what I need. The Intel GPU is used by default, and the NVIDIA card is recognised and powered off; the Nouveau Optmus page shows how you can check this:
$ sudo cat /sys/kernel/debug/vgaswitcheroo/switch
In this case,
DIS: DynOff means that the dGPU will only be used when required and that it is currently powered off. Running graphical applications with
DRI_PRIME=1 offloads processing to the NVIDIA GPU, using the Nouveau drivers.
Since I do not anticipate requiring any more graphical power than the iGPU provides, I don’t plan to install the proprietary NVIDIA drivers, which means no third-party modules in the kernel, and hence no signing required. This is really the best set-up I could have hoped for, and I’m confident it will cause the least amount of problems.