2017: Should Linux Benchmarking Still Be Mostly Done With Ubuntu?

[eydee]

Ubuntu is fine for testing. While it doesn't have the latest and greatest components, it's very easy to install, maintain and change through PPAs. Which means less work, more data on the site and less begging for donations.

[AndyChow]

Ubuntu is fine and you're doing a great job. I'd like to see a Clear linux comparison between Intel and AMD hardware, if it even runs on AMD. That said, I've really been impressed about the consistently great work you do. I will become a premium member as soon as my situation allows it.

[labyrinth153]

As much as I dislike ubuntu, it is the standard end user Linux distribution. Though from my point of view, Red Hat is the standard.

[dungeon]

According to first 100 votes, Debian based and Arch based votes represent majority... 43% are Debian based, 30% are Arch based, with 8% does not care and 19% something lese...

Basically Michael should do 4 out 10 tests on Ubuntu, then 3 out of 10 of Arch, 2 out of 10 something else... and remaining 1 out 10 something else, even non Linux

If he wants to cover majority that should be: 2 should be on Ubuntu, 1 of Mint, 1 on Debian, 1 on Arch, 1 of Antergos, 1 on Manjaro, 1 of Fedora, 1 of OpenSUSE and on every 10 tests one should be unusual/random choosen.

Mixed diversified wreck as Linux always was... it is impossible one to cover them all

[Dukenukemx]

Currently, Mint and Debian are the top two distros according to DistroWatch.com, with Manjoro at #3. Ubuntu is #4, but considering how many distros are based on Ubuntu, it makes sense to continue to use it. If there was a distro to switch to, it would be Debian, but not really worth it.

[dungeon]

Currently, Mint and Debian are the top two distros according to DistroWatch.com, with Manjoro at #3. Ubuntu is #4, but considering how many distros are based on Ubuntu, it makes sense to continue to use it. If there was a distro to switch to, it would be Debian, but not really worth it.

Well, distrowatch is only one part of the info... see for example w3techs's there about web servers statistic:

https://w3techs.com/technologies/det...-linux/all/all

There for example Arch isn't existent at all... Holy Trinity - Ubuntu, Debian and CentOS leading a thing

There CentOS was first until year 2011, then Debain was before and after at the top for about next 5 years and since 2016. Ubuntu became first.

On distrowatch it is more boring if we look at yearly stats, as in last 15 years only 3 distros were at top:

2002-2004 Mandrake
2005-2010 Ubuntu
2011-today Mint

But funny thing about distrowatch is that Ubuntu spins stats are separated, if not Ubuntu in whole would be first there too . As if we count every spin too, Ubuntu MATE + Lubuntu + Xubuntu + Kubuntu + Ubuntu Budgie + Ubuntu Studio + Ubuntu DP + Ubuntu Kylin = 2561 points

Together with Ubuntu Unity's 1452 points that is 4013. so to me in whole it is first by far So it just looks like that Mint is first there

Don't have time but if i start counting first 100 and add all derivatives to base, i am afraid i will got again Ubuntu base the first and Debian base the second... basically same thing like on web servers per w3techs

I think why for some people Ubuntu somewhat suck at a year when Debian Stable should be there is exactly because of that, since one year you have Ubuntu LTS and another one goes Debian Stable and again... so it make sense to use Ubuntu LTS for benchmarks at the year it is released and Debian Stable on the year when it is released.... otherwise Debian Sid because that rolls and both are maded from that

[dimko]

I'd like to see Gentoo benchmarks but I understand it takes a lot of time to setup a Gentoo system. It'd be nice to see exactly the performance increase you get when every single package on your system is compiled with -march=native -O2, and there's quite a few different ways to configure a kernel that effect performance. I could test this myself but I guess I'm lazy.

Its not just compilation, keeping up to date is a bitch too.
Sources: personal experience.

[jabl]

Yeah. The rollingness gives the most up-tp-date results.

The rolling-ness also makes it difficult to reproduce benchmark results, and to pinpoint why something changed (did the benchmark result for some software I'm interested in change due to a change in the software itself, or due to the 475 package updates since I last ran the benchmark?).

[dungeon]

The rolling-ness also makes it difficult to reproduce benchmark results, and to pinpoint why something changed (did the benchmark result for some software I'm interested in change due to a change in the software itself, or due to the 475 package updates since I last ran the benchmark?).

Why difficult, Debian has wayback machine:

http://snapshot.debian.org/

With that you can revert anything and go back even 12 years ago if you want

[RickXy]

For Linux benchmarking, try to stay close to Stock. Debian would be a distribution, which comes closeby to Stock. For Ubuntu benchmarking, use Ubuntu