CentOS

It takes about 40 seconds to boot to the login, half of which comes before the CentOS loading bar, and includes things such as "Warning: AMD Processor - this hardware has not undergone upstream testing" (suggests to consult FAQ on wiki which doesn't say anything relevant I could find) and "fast TBC calibration failed", among other things. Not sure how to get the text it outputs here, it only appears on screen for a very short time, I had to take a picture. How can I troubleshoot?

CPU is Ryzen Threadripper 1900X.

EDIT: and the mobo is an ASRock Fatal1ty X399.

You can create a graphic whic you can analyze in a browser .
Just run And then open the file with a browser.
Every row indicates a service and it's duration of startup.

Also you can run

Thanks hunter86_bg, that helps. Kernel takes less than a second, while initrd takes 3 seconds, doesn't seem that significant... oh, NetworkManager-wait-online.service takes 6.1s. That's a bit much. But even including EVERYTHING on that list, it's total less than 14 seconds... so while there's room for some improvement, there's definitely a different culprit here.

EDIT: Captured a video. The longest time is spent on a blank screen with a blinking cursor, but here it is anyway. https://www.youtube.com/watch?v=aa8VMu-dN2Q

Have you tried the kernels (i think it was called kernel-ml) from elrepo?

The longest time is spent on a blank screen with a blinking cursor

Did you remove rhgb quiet from the kernel command line so that you can actually see what's going on?

TrevorH wrote: ↑

2018/09/15 09:21:13

The longest time is spent on a blank screen with a blinking cursor

Did you remove rhgb quiet from the kernel command line so that you can actually see what's going on?

Thanks, forgot about that. Now I have. Still a long period with a blinking cursor, though, and no warnings immediately after that as far as I can tell.

https://youtu.be/ZJwfkT7oYek

Try appending edd=off to the kernel command line. If that works you could also try edd=skipmbr and see if that works too. If off does not work then I would not expect skipmbr to either.

That didn't seem to work, had the black screen with the cursor for about 17 seconds after reboot. Oddly though I still saw the loading bar, shouldn't that not be showing up since I removed the "rhgb" option? I did run grub2-mkconfig...

EDIT: just now I hit E at the boot screen, and there they were again, rhgb and quiet. So I deleted them again, this time typing in "edd=off" in their place. While grub2-mkconfig might not what I need to be using (maybe it's EFI? I'll have to figure that out), it did give me about 17 seconds of blank-with-cursor again.

grub2-mkconfig is completely unnecessary on CentOS. It is worth updating both /etc/default/grub and /boot/grub2/grub.cfg (or its UEFI equivalent) just for completeness sake but nothing will ever run grub2-mkconfig except a human. New kernel installs update grub.cfg by calling /sbin/new-kernel-pkg which in turns calls grubby which copies the previous grub entry and amends it to become the new one. All automatic.

If edd=off did not work then I'm afraid I'm about out of ideas.

By the way can you provide the .
I used to have a similar issue due to 'resume=' stanza.