First post here. I’ve read enough to make my eyeballs bleed and, at this point, can only hope that this question isn’t dumb enough to merit banning from the forum…
Is it possible to run CUDA on Gnome Wayland, provided that you are only using the NVIDIA GPU for computation (i.e. not graphics)?
A few more details about my setup: I’m on a Dell Precision 5530 (effectively the same as the XPS 15 9570). This laptop has hybrid graphics with two cards: 1) an integrated Intel GPU (UHD 630) and 2) an NVIDIA Quadro P2000. I’m running Arch (Antergos) with Linux kernel 4.19.4. I’m using Wayland as my display server (as opposed to Xorg) for various reasons that I can go into, but I think are ultimately uninteresting for the purposes of answering the question.
The key point is that I don’t need the NVIDIA card to do any heavy graphics lifting. I’m not using this laptop for gaming, CAD software or anything like that. However, I do want to use it for computational purposes. Running TensorFlow and training deep learning models through CUDA and so forth.
There are a few threads on this forum that come close to providing an answer, e.g.:
However, they don’t touch on the Wayland question and also seem tailored to Desktop machines (as opposed to laptops with an existing Linux distro installed). It would be great to get some expert advice before I risk borking my system
Thanks in advance.
P.S. I’ve already described by setup above, but here are the top few lines of “lshw -short” for posterity. The last two lines are the GPUs.
H/W path Device Class Description ================================================== system Computer /0 bus Motherboard /0/0 memory 31GiB System memory /0/1 processor Intel(R) Core(TM) i9-8950HK CPU @ 2.90GHz /0/100 bridge 8th Gen Core Processor Host Bridge/DRAM R /0/100/1 bridge Xeon E3-1200 v5/E3-1500 v5/6th Gen Core P /0/100/1/0 display GP107GLM [Quadro P2000 Mobile] /0/100/2 display UHD Graphics 630 (Mobile)