Which Linux Distribution Boots The Fastest? An 11-Way Linux Comparison

[DanL]

openSUSE use by default wickd which is slow like a hell. Replace wickd with network manager. After that the boot will be really fast.

Yeah, wickd didn't even work when I tried opensuse TW recently. Same settings worked on network manager.

On my main "daily driver" distro (Debian sid), I even got rid of network manager and systemd.networking service because they were the worst offenders of boot time. My network setup on this desktop is simple, so a few text configuration files are no problem to deal with.

[leipero]

This might give some idea: https://www.freedesktop.org/wiki/Sof...aderInterface/

Again, I'd say this test is yet again totally wrong and misleading. The firmware + loader time isn't distro specific. Of course you get shitty boot time if the boot loader menu stays active for 10 seconds on distro X and 1 seconds on distro Y. Also if the UEFI spends time training RAM modules on one occasion and got fixed the next day, the first tested distro would get bad results. I'm really disappointed with this.

I'm not getting this, what do you mean by it? There's a reason why I ask, because recently when moving PC, one module didn't had good contact (or whatever) in the slot, and I did end up with very slow boot time, after re-sitting it and cleaning it worked as before. However, few months ago, system would boot in ~10 seconds, now it need 15, so maybe something fishy is going on with motherboard (or RAM slots), that would probably go under firmware load, but I can't say for sure because I did not try it at the time it was slow.

[Guest]

"The reported firmware time didn't yield too much of a difference."
Eh, the results differ betwen 9522 and 11645 ms. that's over 2 seconds. How does the distro have any effect on this? Isn't this 100% UEFI, not Linux?

This includes bootloader which is distro dependant (GRUB/GRUB2, syslinux or whatever else). I supposed the point was that bootloader should come up in constant time (should be the case with testing, not so sure how true is it over time & updates to bootloader script & kernel entries).

That time includes everything prior to kernel taking over.

openSUSE use by default wickd which is slow like a hell. Replace wickd with network manager. After that the boot will be really fast.

Startup finished in 3.064 (firmware) + 7ms (loader) + 1.085s (kernel) + 983ms (userspace) = 5.139s

From UEFI to working X (firefox autostarts after X in just under a second) with my own kernel with builtin initrd which loads kernel modules (nvidia,vboxdrv) on 2x NVMe boot drives in raid0. Basically just applied basic rules of this (https://www.youtube.com/watch?v=KTRA1PRJWH8) and stripped down systemd (only systemd, systemd-journalctl, systemd-udevd, systemd-networkd services) + light DE (awesome wm) and EFI stub bootloader. Without slow external HDD present that can be shaved by another ~1.3s.

[bug77]

Why no Neon? Those guy do a really good job, you could have given them some exposure...

[Michael]

Why no Neon? Those guy do a really good job, you could have given them some exposure...

Only so much time in a day, especially with this being lots of manual time installing each distro, etc, and just trying to test the major ones that came to mind... Next time.

[caligula]

This includes bootloader which is distro dependant (GRUB/GRUB2, syslinux or whatever else).

Some distros support multible bootloaders. You can even install all distros under the same bootloader. The bootloaders have different settings. Some distros make it wait for several seconds while others boot immediately. The extra waiting for Enter gives a wrong impression of slowness. The test methodology is broken. If some distro waits for a keypress for 60 seconds, in this test it would lose.

I supposed the point was that bootloader should come up in constant time (should be the case with testing, not so sure how true is it over time & updates to bootloader script & kernel entries).

That time includes everything prior to kernel taking over.

Startup finished in 3.064 (firmware) + 7ms (loader) + 1.085s (kernel) + 983ms (userspace) = 5.139s

AFAIK the time spent on bootloader goes under '(loader)'. The '(firmware)' part is 100% UEFI BIOS. Correct me if I'm wrong here.

[linuxforall]

Not bad for old distro like the Ubuntu 16.04

[caligula]

Not bad for old distro like the Ubuntu 16.04

What does this even mean? Ubuntu switched to systemd in 16.04. There's really nothing new related to the boot process in more recent versions. Oftentimes more recent operating systems boot slower as more bloat creeps in with each release. Furthermore, operating system images don't become slower as a result of aging. The package checksums guarantee this.

[Kano]

Basically you can optimize the kernel config that you don't need an initrd and boot straight via EFI without GRUB. This way you can gain 5 seconds. Maybe some distros already use systemd-boot - formerly known as gummiboot which is pretty fast too. New Ubuntu runs GDM and not lightdm, might be slower too. Basically the DE should be the same for all tests otherwise you compare apples with oranges.

[Mike Frett]

Typo:

That's a feature, not a bug.