Power & Memory Usage Of GNOME, KDE, LXDE & Xfce

[Guest]

Ramzswap

To help with your swap problems, change the swappiness of the kernel. The default is very high; it will start swapping even with plenty of RAM left. The default is "60". The range is 0 to 100. 0: never use swap, 100: always use swap even when there's no memory usage to speak of. It works a bit like a percentage, the higher it is, the sooner swap will be used. Some people reduce it to zero, but I don't recommend that since values under 20 can impact performance negatively. I use 30. Put this in /etc/sysctl.conf:

vm.swappiness = 30

Well, I would suggest to try ramzswap available in the newest kernel. I tried it and it seems like a fantastic solution (for my purposes). The only problem is that it is non stable yet, so it freezed my computer when I did "swapoff -a" for example.

Effectively, what will ramzswap get you is the following: Let M be the amount of memory on your machine. Ramzswap will effectively change that into something like (M+S/2), where S is the size of swapped pages.

After the next reboot, swap should only get used when memory is really about to run out.

I don't understand why the reboot ...

[KAMiKAZOW]

Packages may be still being built but 3.5.10 was the last maintenance release and hasn't been maintained for over a year now.

Please, if you don't know sh*t, don't pretend to know. The latest KDE 3.5.10 bugfix is mere 6 days old: http://websvn.kde.org/branches/KDE/
The 3.5 branch isn't as busily maintained as 4.4, but fixes are still going in.

I'm quite disappointed in many aspects of KDE though with regard to robustness. There doesn't seem to be much in the way of fault tolerance in quite a bit of it. If KDE is so sensitive that absolutely every aspect of a KDE deployment must be incomprehensibly accurate with no room for any error, otherwise completely random behavior is observed without clear error reporting then I think the KDE team need to work on that. Only some parts offer decent enough fault reporting that an end user can easily find a solution to a given situation.

I don't think I understand you here. Are you saying that it's KDE's fault that Kubuntu is less stable than good distros like openSUSE?
As the Kubuntu project openly stated by itself, the QA is currently lacking esp. regarding backporting features.
If you take features from an alpha-quality dev branch, then you have to make sure to update it with bugfixes. If you can't, then better stop backporting alltogether.

And while you have a bit of a point here, there's no viable or palatable solution for Kubuntu right now. I would argue it made sense to ship the Ubuntu/Gnome based package management with Kubuntu until a viable solution was developed, but if they were to do something so bold as that you'd have KDE zealots crying foul because as everyone knows, you can't have Gnome software on a KDE desktop or the world implodes.

That's completely false. The Kubuntu team itself decided to be Qt-only and not to let any GTK-based app in by default. That has nothing to do with "KDE zealots".
Other distributions have no problem shipping GTK applications in the default install. That includes the respected openSUSE that ships with Firefox as default browser and IIRC Gimp is also installed by default.

Canonical is obviously has no interest to put actual development resources into a KDE-based package manager, even though the foundations are there. KPackageKit and Shaman are there. Canonical just needs to put actual manpower into improving one of them.

And that's the main difference between Canonical on one side and Novell and Red Hat on the other side: If Novell and Red Hat need something, they just develop it (eg. Novell wrote the new KNetworkManager and Red Hat does lots of driver development). Canonical OTOH just takes and only seldom gives back (in all fairness: It has improved).
Novell had no problems rewriting YaST in Qt4.

these days Ubuntu (and therefore any Buntu) seems to be the biggest blip on the radar for projects not in a standard repo though.

No,it's not. You don't seem to know the big community that just centered around openSUSE's Build Service.

[RealNC]

I don't understand why the reboot ...

Because running sysctl to change the swappiness will not affect the stuff already swapped out.

[BlackStar]

I never could understand that. I keep installing version after version of openSUSE (since 10) only to delete it after being completely fed up each and every time. The biggest issue is that they bastardize the KDE UI to fit some kind of ugly, perverse design

openSUSE 11.2's Plasma theme is just a slightly modified Air theme. It's 98% the same as vanilla KDE SC.
openSUSE 11.2's KDE themes were modified by KDE's Oxygen artist team itself.

I'm talking about the UI, not the theme. OpenSUSE does many braindead things, like shuffling KDE configuration options around, throwing tiny bitchy icons into the mix and using that awful shade of green. Add the god-awful package manager on top and the general slowness and openSUSE just plainly kills the KDE experience.

I actually thought all this was KDE's fault at first, before I tried Arch and Kubuntu. Nope, KDE is pretty usable there - the blame falls squarely on openSUSE's shoulders.

(K)Ubuntu isn't the only distro with many apps in the repos. Fedora, Debian, Arch, and openSUSE all have someone to package an app for you into the repo.

That's right, but these days Ubuntu (and therefore any Buntu) seems to be the biggest blip on the radar for projects not in a standard repo though. So thats a valid reason to select one of the Buntus for a distro, and the best KDE based Buntu is Kubuntu.

It's more than that. Banshee released some (opt-in) project statistics recently and *buntu(*) users where an order of magnitude more than users of other distros. The difference is staggering: we are talking about 383 installations of the *buntu package vs 33 source installations vs 27 installations of the openSUSE package.

Given that *buntu is the most popular distro by far, I think it makes perfect sense for Michael to focus benchmarks on that. This way, benchmarks will actually reflect potential user experience much better. Even so, Michael does test other distros from time to time, so it doesn't really make sense to gripe for favoring Ubuntu.

((*) doesn't differenatiate between flavors of *buntu, but it's safe to assume most are using vanilla Ubuntu, given Banshee's Gnome affiliation.)

[bytor4232]

What method of measuring memory usage are you using? There's a ton of different ways this gets reported, and IIRC most of them are misleading.

Also, I agree that Ubuntu + KDE and Ubuntu + XFCE isn't nearly as interesting as testing Kubuntu or Xubuntu directly. I'm not actually sure there's any difference, but I could easily see the former including a bunch of extra software running in the background which wouldn't share memory as well.

I'd be very surprised if KDE didn't still use the most memory though. It's got so much more functionality built into it that it's kind of a no-brainer - a tie in memory usage would be correctly interpreted as gnome having serious memory problems.

http://leanubuntu.blogspot.com/2009/06/lxde-vs-xfce4.html

http://leanubuntu.blogspot.com/2010/02/gnome-as-lean-ubuntu-choice.html

I didn't test KDE/Kubuntu. But yes, there is a lot more going on in the background with Xubuntu than Ubuntu(cli)+xfce4, same with Ubuntu vs Ubuntu(cli)+GNOME-Core

[mugginz]

I don't think I understand you here. Are you saying that it's KDE's fault that Kubuntu is less stable than good distros like openSUSE?
As the Kubuntu project openly stated by itself, the QA is currently lacking esp. regarding backporting features.
If you take features from an alpha-quality dev branch, then you have to make sure to update it with bugfixes. If you can't, then better stop backporting alltogether.

Well here it is. If you're saying that OpenSUSE is demonstrably more stable than Kubuntu, then I call Bullsh*t!

As I've said, one distro may have brokeness in one place while another distro may be broken in another. OpenSUSE is FAR from perfect, neither is Kubuntu, and yes I do place some blame on upstream for this. If you're going to try to suggest for even a moment that vanilla KDE has been perfect at any point in the 4.0 to 4.4.1 release time line then I think you've got a tough sell on your hands.

People want to actually blame KDE brokeness of Kubuntu. If OpenSUSE had fixed every single KDE bug, and had upstreamed them, and then Kubuntu didn't take those patches on board then you may have a point. If Kubuntu ships vanilla KDE and there are bugs, its KDEs fault. If OpenSUSE have themselves fixed the brokeness and have this in shipping packages, and Kubuntu only ships vanilla KDE then it's valid to say KDE sucks without OpenSUSE patches, it's not valid to say Kubuntu magically broke KDE because they didn't fix it themselves.

It's only fair to lay the blame where it deserves to be laid. I might add that no, OpenSUSE haven't fixed all of KDEs bugs. Lets hope someone eventually does though.

That's completely false. The Kubuntu team itself decided to be Qt-only and not to let any GTK-based app in by default. That has nothing to do with "KDE zealots".
Other distributions have no problem shipping GTK applications in the default install. That includes the respected openSUSE that ships with Firefox as default browser and IIRC Gimp is also installed by default.

Kubuntu also ships with some Gnome apps, so OpenSUSE is no orphan there. It's which Gnome parts a KDE distro ships that's the sticking point. If you seriously think for one moment there will be no uproar if Kubuntu ships with Gnome based Samba, Bluetooth and package management then I think you're mistaken.

I would prefer that they did ship them in place of KDEs, but because it's just so easy to add them it's no real big deal to anyone unless someone wants something to complain about. When there's a solution just a few mouse clicks away for someone, but instead they prefer to stay with what's broken and simply wine then I have little respect for them.

Canonical is obviously has no interest to put actual development resources into a KDE-based package manager, even though the foundations are there. KPackageKit and Shaman are there. Canonical just needs to put actual manpower into improving one of them.

But isn't it part of that project TimeLord that you mentioned to infact work on KPackageKit fixes?

Also there's been much discussion on whether or not people consider that Canonical puts a satisfactory amount of man power into Linux. Or at least do they put enough manpower in to justify them being allowed to ship Linux, etc. Remember, money and manpower don't come from the magical Linux fairy, but instead from the community and from private companies. Canonical are a young company, one that still isn't cash flow positive, but have a vision for the Linux desktop that they are executing on. These things take time and their major focus is unapologetically Ubuntu and Gnome first.

They are ramping up with more resources being spent on Kubuntu from what I understand though and this is a good thing. That being said, and even though they spend more time on Ubuntu than Kubuntu, when doing usability comparisons between OpenSUSE and Kubuntu I don't find it at all to be the case that OpenSUSE has a premium experience relative to Kubuntu.

And that's the main difference between Canonical on one side and Novell and Red Hat on the other side: If Novell and Red Hat need something, they just develop it (eg. Novell wrote the new KNetworkManager and Red Hat does lots of driver development). Canonical OTOH just takes and only seldom gives back (in all fairness: It has improved).
Novell had no problems rewriting YaST in Qt4.

I'm confident that any fair evaluation of this subject will reveal that Canonical are doing what they can in this area. Cashflow is extremely important to consider when you're asking someone to find money to pay devs. Canonical can't simply keep asking Mr. Shuttleworth for more money all the time. As Canonical's source of income grows, so does their contribution. How long have RedHat and SUSE been on the scene for now relative to Canonical, and what are their relative incomes? This info is available on line for those who are interested.

No,it's not. You don't seem to know the big community that just centered around openSUSE's Build Service.

Well have a good hard look at the user bases of the various distros, and consider which open source baskets people are putting their proprietary eggs in. Which is the most prominent client for Skype? Which distros did the prebuilt Boxee binaries get built for in the early days? Which distro will Lukewarm Media be sure to support before all others when packaging their up and coming title Primal Carnage?

If you google for a Linux program, go to their site, and look for a pre-built binary for that particular project, how often do you find it to be the case that there's no Ubuntu build but there is for the other distros? I would hazard a guess at not many.

[thefirstm]

That original news article was flamebait. Someone should have seen that coming.

[mugginz]

That original news article was flamebait. Someone should have seen that coming.

Well any KDE vs Gnome story is guaranteed to bring out the trolls. That's why it's helpful to have well designed and supportable methodology in order for the story to provide accuracy, and to then leave the inaccurate commentary and various other flameage to the public to make.

[ChemicalBrother]

So, I'm using Arch Linux + KDEmod (Chakra Linux). KDE 4.4.1 on my EEEPC 1000H and htop says: 221 MB RAM usage.

This is such a big difference to the phoronix results... I don't get it.

[V!NCENT]

So, I'm using Arch Linux + KDEmod (Chakra Linux). KDE 4.4.1 on my EEEPC 1000H and htop says: 221 MB RAM usage.

This is such a big difference to the phoronix results... I don't get it.

Aside from the fact that you should at all times have preload installed on a laptop with massive batterylife...

Ubuntu packs everything. It doesn't matter if you're having just an Intel card or something... there are kernel modules running you'll never use.

This makes up for a great user experience at first sight, but that also means it will search for a USB bluetooth device every 5 freaking seconds, even if you don't even own one...

Other than stuff like that, everybody knows why they use a specific DE/WM... Enlighment users, Windowmaker users, Openbox users, Xfce users, KDE users, Lxde users... Only Gnome users don't.

Yes... I favor any UI over Gnome.

I'm currently actually running Gnome because I wanted to test Fedora once again and the Fedora KDE4 experience was horrible so I choice Gnome.

It boggles my mind why anyone would prefer Gnome to Xfce... ever...

Gnome supposed to have the 'awesome' clearlooks theme that beats every theme in usability, yet they forgot the usablility of shades of grey mixed with other collors... It's truly sad for Gnome that KDE SC 4.4.2's theme is now 'more usable' than Gnome's. So I installed Gnome-do for the docky, changed the upper menu bar thing to gray and made it transparent, put a KDE4 wallpaper (in Gnome you can't set the shade of the color of the wallpaper to gray for example because Gnome doesn't implement features that people like to have) in place and changed the blue to light gray-ish blue and finally my eyes agreed to the look of my desktop:

http://img534.imageshack.us/img534/351/screenshotue.png