Fedora vs. openSUSE vs. Manjaro vs. Debian vs. Ubuntu vs. Mint Linux Benchmarks

[monraaf]

Regarding the bad results of openSUSE:

The article says, all distributions (except Manjaro) were using deadline or CFQ. This is not true.
According to Richard Brown (openSUSE Chairman) the beta used NOOP on SSDs. The final version will use deadline. That should make it better in most cases.

Source:
https://www.reddit.com/r/openSUSE/co...s_from/cwkah18

[Cape]

That is what their installer does man and if you have never install a recent opensuse distro please don't say such things... I am using opensuse since it was called suse 8.2 and believe me I know how it works.
I still remember losing grub in 13.1 version and cannot enter any OS when install with btrfs root and it was a known bug then and since that I never used btrfs again afraid not to mess my system with that thing one more time.

Ok, that has nothing to do with btrfs, but rather with a bad implementation of it. As for the many partition problem: if suse enables snapshot/backup for every partition than you double the number of partitions. Again its a choice made by the distro who tries to be safer on data integrity at the cost of performance and simplicity (a debateable choice for a desktop usecase...). I'm sure that even in Suse you can customize your disk layout during install and make a simpler /,/home,/boot layout with btrfs and get better performance and a simpler system.

Anyway, to me this story just highlights one more time that major distros are overall bad. I find it ridiculous that developers release new software in a hurry but then a distro let months pass before it can reach the end user and, many times, there are outstanding bugs caused by maintainers which delays the adoption of new features even more.
At least, arch Linux solves one of this problem with his release model. Yet a standardized communication model between developers and releasers/distros is in my opinion the only way out of this mess and possibly the door for Linux on desktop and offices.

[jegjessing]

It might be slow, but at least it does not destroy my machine.
This is the 3. time in a row where I try to install Fedora and it stops after wiping my partitions by crashing.

So no partition and back to opensuse.

OpenSUSE is slower but far more stable, as far as my experience goes.

[elav]

Correct me if I am wrong, but it looks like there were several different desktop environments were used, but they did not seem to greatly impact the test results. Am I correct in this?

I agree!

[geearf]

Should I even bother looking past the first page?
"Out of the box performance" >> Proceeds to use a beta version of every distro except Linux Mint and Manjaro. In the latter, it's not that hard to use one command to switch to the unstable repos after your install:
sudo pacman-mirrors -gb unstable

And BOOM, mesa 11.

As for Linux Mint, it's as simple as checking a box in your update manager to get to the "beta" repository...

But, we all know how you feel about Manjaro and Linux Mint, so I'm just falling on deaf ears.

Yeah I did not get the point of using different versions of software to test the distributions against each other.

[ObiWan]

OpenSuse is the only one using NOOP on the SSD, and afaik NOOP slows down the entire system with heavy disk usage.

So part of the problem could be the scheduler.

[djdoo]

You have quite a messed up idea of btrfs. I have my root and home in btrfs on a normal HD (no need for dozen of partitions... where did you get that?) and i can't complain about performance. Do you want a good reason for choosing btrf over ext4? Here you go: online, super easy defrag. And don't tell me that linux partitions don't need defragging because i my 10 years linux exp I always experienced a sudden decrease of performance after about 2 years, and defragging always solved.

That is what their installer does man and if you have never install a recent opensuse distro please don't say such things... I am using opensuse since it was called suse 8.2 and believe me I know how it works.
I still remember losing grub in 13.1 version and cannot enter any OS when install with btrfs root and it was a known bug then and since that I never used btrfs again afraid not to mess my system with that thing one more time.

[sysrich]

I know this is a somewhat blasphemous opinion for this site but there is more to life than Benchmarks..

BTRFS makes perfect sense as the default root filesystem for openSUSE and SLE when you realise that we have snapper, which provides full, system wide, snapshoting, rollback, and boot to snapshot - no more sysadmins or rogue software breaking things in a way that cant be undone

Also, at openSUSE/SUSE, we take time to care about other factors in addition to performance, such as reliability and consistency - last time I checked we're the only distribution to have Write Barriers enabled by default, to ensure that your data is actually written to your disks in the correct order in a consistent manner. That comes with a bit of a performance cost sometimes, but I'd rather be sure my data is actually ON my disk, than hope it is..

That said, I still think there's a little tweaking we can do to openSUSE Leap before we release it in November. It's still in Beta, we got time

[Cape]

I still cannot understand why they chose btrfs as the default FS for root, in all benchmarks it is behind ext4 and xfs in performance and it needs a dozen of partitions in order to run a system that a newbie will be really frighten when see the tree at the installation procedure and likely will not proceed.

You have quite a messed up idea of btrfs. I have my root and home in btrfs on a normal HD (no need for dozen of partitions... where did you get that?) and i can't complain about performance. Do you want a good reason for choosing btrf over ext4? Here you go: online, super easy defrag. And don't tell me that linux partitions don't need defragging because i my 10 years linux exp I always experienced a sudden decrease of performance after about 2 years, and defragging always solved.

[sysrich]

Easy explanation for the poor performance of both OpenSUSE & Manjaro:

(Soft) REAL-TIME Linux Kernel [which obviously doesn't make any sense on a desktop system, just like openSUSE & Manjaro don't make any sense!]

Type "uname -a" and look for "PREEMPT", which means that the Linux kernel itself freezes ITSELF for userspace apps, which of course leads to worse performance and increased latency on the kernel side!

It's not quite THAT simple..but, yes, there are certain areas where a PREEMPT kernel isn't a smart idea

Which is one of the reasons openSUSE is thinking of dropping it in their default kernel (called kernel-desktop)

HOWEVER, openSUSE Leap already has a non-PREEMPT kernel, called kernel-default (it's the default in SLE)

I'd love to see these benchmarks re-run with that kernel, could someone make it happen please?