My advice is to generally opt for integrated on mobile, unless you absolutely need them. I did on my last computer (training ml models can often be sped up with Cuda cores), but the trade off was it breaking three times when updating my Nvidia drivers (had to chroot in an manually update, huge pain to deal with), so I specifically went away from Nvidia drivers on my latest laptop.
My advice is to generally opt for integrated on mobile, unless you absolutely need them. I did on my last computer (training ml models can often be sped up with Cuda cores), but the trade off was it breaking three times when updating my Nvidia drivers (had to chroot in an manually update, huge pain to deal with), so I specifically went away from Nvidia drivers on my latest laptop.