Is Antergos Arch Linux Really Faster Than Ubuntu, Fedora?
Phoronix: Is Antergos Arch Linux Really Faster Than Ubuntu, Fedora?
Frequently within the Phoronix Forums it is requested to do benchmarks with Arch Linux since its users tend to be adamant that it's the fastest Linux distribution. In the past I've run benchmarks of the Arch-based Manjaro to look for speed differences as an easy and quick to deploy variant. Today the latest Arch Linux variant I am benchmarking is Antegros Linux.
Don't know why some vocal Arch users claim the system is faster than other distros, e.g. Ubuntu. I use Arch myself and the main advantage is mostly being cutting edge (or bleeding edge if you choose so) with little effort, plus the possibility to stay it without reinstalling, even if it is just every year. Arch devs don't have some magic fairy dust that makes linux faster. The reason I would have requested Arch is because that way you know the applications used in the benchmarks are up-to-date and vanilla.
So I totally get the "defaults" approach to benchmarking. Being able to reproduce the results is, in fact, the cornerstone of benchmarking. But the issue here is that Arch is not about "defaults" its about doing it yourself and adding all the tweaks and mods you want. This is why Phoronix went with Antergos, to get some defaults. But I think the best way to go about it would be if somebody (somewhat knowledgeable) took his offer and posted the tweaks they would do in detail.
At the end of the day though, I use Arch because of the documentation, Aur, package management, and rolling release. I don't claim any "magic" speed improvement. However, the nature of Arch makes it easy to strip down (or simply not install in the first place) a lot of the overhead most other distros install to appeal to as many people as possible.
Although I was initially concerned with the choice of default IO scheduler, I agree with your reasoning, and I think a defaults test is appropriate. More useful than comparing different distros would be to use Antergos to test various package versions. I have not used Antergos (I do use Arch), but I assume that it probably doesn't introduce things like GVFS by default, and via the Arch repos you could control the environment more effectively to emulate what various types of users might be running, like different desktop environments and setups (gaming/desktop/server), to get better benchmarks.
Also, what caused it to not use deadline by default for the SSD?
I see what you're saying but what you don't realize is the reason they're faster is because by default, they don't come with half the bloat and crap you don't need. Using a pre-made desktop distro for Arch won't give you any better results than Ubuntu or Fedora because in the end they're all running roughly the same things. Start with a command line and only install what you use, and then you get a performance boost. Also, there's a certain point where your hardware is good enough that optimizations don't matter anymore. Distros like Arch and Gentoo are appealing on low-end systems where every byte counts.
Originally Posted by Laser
Back in the days when I knew significantly less about Linux, I set up my server using the desktop version of Ubuntu (Hardy Heron). Then, when support for that was coming to an end, I tried using the inline upgrade (to Lucid Lynx I think it was), and it worked in a fashion, but it was far from painless. Afterwards, I took some time to install Arch without a GUI, and I've had the same installation since, frequently up-to-date and absolutely no need to worry about when support is going to end.
I agree with the other posters, the speed is nice, but it's the ease at which you keep up-to-date which is the kicker. Aur and documentation is also top-notch. If, when you google something Linux related, and nine times out of ten you get the Arch Wiki page, it helps that you're running Arch and you don't have to account for a different distro.
Thanks for looking into this Michael, it's much appreciated.
the bolded comment reeks of "i don't know a thing about performance/system tuning, so instead, i am just going to talk crap"... No bro, you squeeze more performance by tuning your system properly. Installing 'only-what-u-need' may cut down on cruft / save some disk space - but that is about it. There is a reason that Archwiki provides performance tips in XYZ wiki page and also has this page; https://wiki.archlinux.org/index.php...ng_performance [that's why it's also funny that michael is asking for docs..lol.] ... There is also a reason why people like Redhat/SUSE make comprehensive performance and/or system tuning guides as well; for example; http://activedoc.opensuse.org/book/o...d-tuning-guide ... I always tune my systems, and sometimes the benefits are quite substantial [everything from boot time, to scheduler[s] performance, improving user-space apps performance/responsiveness [for example, firefox by using 'profile-sync-daemon'], etc, etc...
Originally Posted by schmidtbag
And i cal total BS on your last 2 comments; There will NEVER be a point where H/W is good enough that optimizations don't matter anymore && Arch/Gentoo are for people who want to build their own systems from the ground up / for people who want the benefits of ports-like systems, amongst other things... Arch, gentoo, etc are just as appealing for high-end systems as low-end systems... you have no idea what you are talking about.
And then you end up with a somewhat broken system. Don't get me wrong, there are many arch users that know their shit, but I've seen many halfassed and broken installations just because one doesn't understand everything, because "pacman -S kde-meta" is not l33t enough.
Originally Posted by schmidtbag
I've also seen my fair share of broken Ubuntu installations (oh god, in my school they used to have an half-assed Ubuntu-server used as a desktop distro. Also every PC was at best created from a clone, and at worst recreated from scrap, making every PC have a different problem. Guess why everyone hated Linux in my school), but Arch and Gentoo make it easier.
Anyway, Arch uses the same package as Ubuntu & Fedora, there's no black magic behind. So it's going to be just as fast, given the same version. Obviously Arch get better over time, since it gets the recent packages while Ubuntu/Fedora are stuck on an older one.
Same for Gentoo, as usually flags and stuff like that are 100% insignificant outside of micro-benchmarks that have no value in real world.
At some point we need to stop the pissing matches about which distro is best. No single distro is perfect for everyone. If it was we wouldn't have so many choices. I use Gentoo but I don't go around telling people my system is better because I build everything for my exact hardware. I use it because it's best for me. Ubuntu is good for people who want minimal interaction with maintenance. Arch is good rolling release distro that keeps you on the bleeding edge. Hell my employer uses CentOS. To each their own, use what works best for you or the task at hand. Compilers are getting better every year so the speed difference in a lot of these test won't even be visible to the majority of people.
I would like phoronix to repeat twice a year this tests
I recently tested at Half Life lost coast with my recent NVIDIA GTX 750 Ti Manjaro and gave me 150 fps and Ubuntu gave me 100 fps.
Of course with same kernels and almost same packages running it is almost the same BUT
As you show, a Fedora not updated is far slower than a normal Antergos always updated
I still have Xubuntu + Manjaro and I prefer Manjaro, and I will switch to any other that i would feel better with and it is not imperceptible magic, perhaps the benchmark tests do not show enough the differences for good or for bad.