Is Arch Linux Really Faster Than Ubuntu?

[phoronix]

Phoronix: Is Arch Linux Really Faster Than Ubuntu?

Often when we are preparing for cross-distribution comparisons or benchmarks of different operating systems (like our recent Mac OS X 10.6 vs. Windows 7 vs. Ubuntu 10.04 benchmarks) we are often asked to include Arch Linux in the mix. This is usually on the basis of including a rolling-release distribution to provide a performance look at a constantly evolving distribution with many of the most recent open-source packages rather than a traditional distribution with packages that may be months older. Many of those requesting Arch be included in our testing mix also claim that Arch performs significantly faster than Ubuntu and our usual test candidates. The main reason we do not deliver many benchmarks of Arch, Gentoo, or other distributions that use a rolling release approach is that they are not very reproducible with their results since their packages are frequently changing and there are more end-user customizations going on compared to most other distributions. However, to test the performance claims of Arch versus others, we have compared the performance of the newest Arch 2010.05 media against Ubuntu Linux.

http://www.phoronix.com/vr.php?view=14946

[ethana2]

Seriously?! Ubuntu doesn't put compiz on hold when you're playing a 3d game? That is horrible. When I'm playing a game, I expect all my gpu power to be used for it.

What, do I have to prepend
metacity --replace &
to every 3d game in my menu? Inelegant, poorly thought out.

My first instinct was to request tests with ubuntu running metacity to make it fair, but if that's not thought out better, they deserve to take the hit.

Arch is superior to ubuntu for 3d gaming in terms of performance. How unfortunate.

[RealNC]

That's a Compiz problem, not an Ubuntu one. If you run Compiz on Arch, same thing will happen.

[grigi]

The reason this is an invalid comparison is because Rolling release distributions encourage much more customization.
That in turn encourages a more minimalist approach.
So the system is "lighter", and a lighter system gives the appearance of speed.

There is no real performance difference in separate apps/benchmarks, is because there isn't. But the system may feel faster because it is lighter.

Also, for my one dev environment, getting Ubuntu's build tools in proper order takes longer than installing Gentoo. Different use-cases require different distributions.

Honestly there is no such thin as a "standard recommended installation" for a rolling-release system.

Over the many years I played with lots of distributions, the only place I found a significant difference, was with a custom-built installation of Gentoo on a Atom system (because the Intel Atom is an in-order machine, and setting the compiler to produce optimal code for that really helps).

[Ragas]

Over the many years I played with lots of distributions, the only place I found a significant difference, was with a custom-built installation of Gentoo on a Atom system (because the Intel Atom is an in-order machine, and setting the compiler to produce optimal code for that really helps).

How did you do that and what were the key steps? I have a very slow netbook too and I want to optimize it.
The things that hold me back from it is that I don't think it's too clever compiling all stuff on such machine (wich of course could be worked around) and eventually my girlfriend would kill me, if nothing worked on her computer, because I simply didn't configure it till then.

[ethana2]

You can't blame upstream. If the compiz devs won't pull the patches Ubuntu needs to behave correctly, I expect Canonical to make their own bzr branch of compiz and ship it with Ubuntu instead.

[yotambien]

The reason this is an invalid comparison is because Rolling release distributions encourage much more customization.
That in turn encourages a more minimalist approach.
So the system is "lighter", and a lighter system gives the appearance of speed.

There is no real performance difference in separate apps/benchmarks, is because there isn't. But the system may feel faster because it is lighter.

Agreed. If I remove the seats from my car it will weigh less and be faster. Not very comfortable though.

With World of Padman was the first test where there is a large difference between Arch and Ubuntu. However, the reason for this is likely that Arch Linux does not use Compiz by default when an accelerated OpenGL driver is available, which is in contrast to Ubuntu. This is simply due to a difference in how each OS is by default. If Arch Linux has Compiz running or Metacity was running instead of Compiz on Ubuntu, we would have likely seen similar numbers.

If you say that much I think you should back it up. It's probably true, but just an extra test with Compiz running in Arch would give you the actual numbers. Good job, though.

[V!NCENT]

Seriously?! Ubuntu doesn't put compiz on hold when you're playing a 3d game? That is horrible. When I'm playing a game, I expect all my gpu power to be used for it.

You'll find that, as a gamer, you want to be able to Alt-Tab out of your game if you are getting pwned (read = dead, wait for the next round), or not and go afk, so you can quickly check what server IP's are, or in a MMORPG you might want to check some world map with a lot of useful info. Maybe you are playing a game and you want to see how to max out your kills so what strats to use and whatever, leaderboards...

In Windows your GUI has to be somewhat reloaded and causes the game app to pauze and when getting back into the game you'll get a black screen/freeze for a couple of seconds. This is irritating as hell.

So I for one will go ahead and call it a very welcome gaming feature. There should be a GUI checkbox option created somewhere called "DIsable Compiz when running fullscreen OpenGL" though for people with a 10yo GPU.

Arch is superior to ubuntu for 3d gaming in terms of performance. How unfortunate.

How fanbias'd...

[Michael]

For those of you not aware of the Compiz hit, sure, tomorrow or Friday I'll post some benchmarks showing its cost with ATI/NVIDIA/Intel hardware and the different drivers.

[V!NCENT]

For those of you not aware of the Compiz hit, sure, tomorrow or Friday I'll post some benchmarks showing its cost with ATI/NVIDIA/Intel hardware and the different drivers.

That would be nice

Maybe later if you have the time I would also like to see a Kwin Vs. Mutter V.s Compiz Vs. Metacity.
I want to get to know the avarage impact of compositing in general versus oldskool framebuffer blitting