Announcement

Collapse
No announcement yet.

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

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Originally posted by Linuxhippy View Post
    Thanks a lot for running those tests - I always had the feeling KDE4 was a memory hog - using more than twice the amount of memory right after startup than KDE-3.5.

    Well, hopefully they'll work on it - at the rate they add new features I doubt there's a lot time left tuning if, at least I haven't seen major speed/rousource optimizations since KDE-4.0

    Thanks again, Clemens
    It is apparently Kubuntu that is the memory hog, not KDE 4.4.1.

    I'm running Arch Linux with KDE 4.4.1, and I can start KDE using justt 260 MB of memory. With a qt-based browser (Arora) and file manager (Dolphin), editor (kwrite) and console (konsole) running, it uses 350 Mbytes. This is less than GNOME. I have no idea how Phoronix and Kubuntu manged to use up as much memory as they did.

    BTW, KDE 4.4.1 running under Arch Linux, with composited desktop using the open source ATI 3d graphics drivers, is faster than GNOME on the same hardware.

    Comment


    • Originally posted by hal2k1 View Post
      It is apparently Kubuntu that is the memory hog, not KDE 4.4.1.

      I'm running Arch Linux with KDE 4.4.1, and I can start KDE using justt 260 MB of memory. With a qt-based browser (Arora) and file manager (Dolphin), editor (kwrite) and console (konsole) running, it uses 350 Mbytes. This is less than GNOME. I have no idea how Phoronix and Kubuntu manged to use up as much memory as they did.
      Nice try, but 350MB is almost exactly twice the memory usage I get on Arch/Gnome using the (pretty heavy) New Wave theme, Compiz, Gnome Do, Firefox and Nautilus: 174MB.



      KDE does use more memory than Gnome. It still fits in a 512MB machine, which means that noone really cares.

      Comment


      • Originally posted by susikala View Post
        Replies to this thread have been mostly an endless string of "protect my favorite DE" bullocks. I can sum them up to:

        1. kde pulling in lots of unrelated dependencies, because
        2. kubuntu is a horrible implementation of kde, and
        3. you're not using vanilla kernels or archlinux/gentoo

        i.e. "you SUCK".

        The thing is, no one cares about archlinux and/or gentoo and your -Os flags (I have yet to witness the advantage of using compilation flags, never changed for me an INCH of how fast stuff runs. I was just staring annoyed at my screen looking for some lameass TEXT EDITOR to compile. That's when I got back to debian). People use debian, ubuntu, fedora and suse and co., if distrowatch's to be believed. It's not a matter of the distribution, it's the fact KDE is just more bloated than the rest, in general. Which is bloated enough (even XFCE feels slow nowadays).

        I've tried many distributions, even those that offer KDE as their main DE. KDE 3X, KDE4X, the thing was ALWAYS sluggish and annoying to use, when you want to get down to being productive. If your DE needs a gaming computer to feel snappy, I guess you're doing something wrong.

        Honestly, I couldn't care less about GNOME, but pointing the blame at the distribution, the implementation, or QT, is just puerile. KDE is crap, don't use it. These tests show it clearly.
        I'm running Arch Linux, which according to distrowatch is far more popular than Kubuntu.

        On Arch Linux running KDE 4.4.1, I am running a faster and more powerful desktop that uses less memory than Phoronix is reporting for Ubuntu/GNOME.

        OpenSuse and Mandriva are more popular KDE distributions again than Arch, but Arch is a rolling release and is therefore more up to date and cutting edge. I cannot speak for how light-weight or otherwise Mandriva or OpenSuse are, but Arch can easily run a full KDE desktop using less memory than Phoronix has published for running Ubuntu/GNOME.

        BTW, using Arch and pacman is virtually indistinguishable, from a GUI-ness point of view, to Mandriva, OpenSuse, Kubuntu or Ubuntu/GNOME with apt-get, urpmi or yast.

        Comment


        • Originally posted by SkyHiRider
          I'd say memory is not an issue today - it's relatively cheap and there's more the enough of it. I have 4 gigs of ram and usually don't even use half of it.

          Of course it is an issue. From Fedora's smolt:

          (everything in Mb)
          Code:
                      M < 256      1.5%
            256 < M < 512      12.3%
            512 < M < 1024    24.7%
          1024 < M < 2047    30.0%
          2047 < M < 4095    25.6%
          4095 < M < 8191     4.7%
                      M > 8191     1.1%
          1 Gb RAM nowadays is already limited if you actually use your computer. ~39% users have only up to that amount. The next against the wall will be users with up to 2 Gb, who alone are a third of the total. Add to this that the ~29% of the machines are laptops--a proportion I'd only expect to increase with time--whose memory can't normally be upgraded.

          Also, it's not the extra 30 odd Mb KDE may claim on a clean reboot, but the overall memory usage what counts (no, that wasn't tested, I'm thinking about the importance of physical memory in general). Nowadays some applications are very memory hungry, and yes it is a real problem. I rather have my memory going to the applications than to the window manager or whatever.

          Comment


          • Originally posted by BlackStar View Post
            Nice try, but 350MB is almost exactly twice the memory usage I get on Arch/Gnome using the (pretty heavy) New Wave theme, Compiz, Gnome Do, Firefox and Nautilus: 174MB.

            ...

            KDE does use more memory than Gnome. It still fits in a 512MB machine, which means that noone really cares.
            Fair enough. I used htop as you did and had a look at the biggest memory user, and found to my surprise that it was Arora, which was using the same amount (4.2% of memory) for each tab opened. By closing Arora, I got it down to 265MB, which doesn't beat your figure, but nevertheless it is almost down to half of what the Phoronix tests claim that Kubuntu uses.

            I guess the critical thing is that if you have 512Mbytes or more, any desktop (other than perhaps Kubuntu) will run fine. If you have only 256Mbytes, no variant of Ubuntu will run well, and you will have to use something like Arch, perhaps with LXDE but it appears even GNOME might be OK.

            A 256 Mbyte machine is getting to be a pretty old machine these days. Even the cheap netbooks come with 1 Gbyte.

            Comment


            • Originally posted by hal2k1 View Post
              I guess the critical thing is that if you have 512Mbytes or more, any desktop (other than perhaps Kubuntu) will run fine.......
              Kubuntu is fine in 512M, but I wouldn't ever recommend someone run it on a 256M machine. I generally recommend one of the Buntus where possible and depending of user needs, but when you're talking about a 256M machine then Arch looks like a much better choice.

              I consider a 256M machine a special case when it comes to OS selection. Ease of use and third party package availability rate very highly and most other considerations much lower without good reason to consider them as primary concerns.

              Still don't know why Michaels tests showed such a big difference between Gnome and KDE.

              Just for laughs I'm running some tests on a Celeron 1.2G (PIII), 40G, and 512M just to see what I find there.

              Comment


              • (In my last post I said that it was my last post in this thread. Now it seems, I lied. But things have improved since then, a little bit.)

                Originally posted by hal2k1 View Post
                It is apparently Kubuntu that is the memory hog, not KDE 4.4.1.
                I think, the main problem is the methodology of the measurement, i.e. what is in fact being measured and how.

                Originally posted by hal2k1 View Post
                With a qt-based browser (Arora) and file manager (Dolphin), editor (kwrite) and console (konsole) running, it uses 350 Mbytes. This is less than GNOME. I have no idea how Phoronix and Kubuntu manged to use up as much memory as they did.
                Well, unlike you, I do know how Phoronix arrived at those numbers: the Phoronix numbers include the memory used by the disk cache. Try running phoronix-test-suite on your computer and see for yourself. So, for example if an application read some configuration data, or a PNG image, from disk during startup, then that disk data will still be in cache and gets included in the Phoronix results. Who knows how many megabytes got included like this.

                Maybe a better method would be:

                echo 3 > /proc/sys/vm/drop_caches
                cat /proc/meminfo | grep MemTotal\|MemFree

                But by the far best solution would be to have a special page on Phoronix dedicated to methods used in obtaining the measurements. Such as "How does Phoronix measure memory consumption? What does the resulting number include, what it doesn't, etc.". THAT is what I am missing here. While you can get a precise idea about what and how is being measured by examining the phoronix-test-suite source code, it is not a good solution for a benchmarking site.

                Message to those who are using htop: those numbers are NOT comparable to Phoronix numbers. The htop numbers do not include cache and buffers. In my (first, I think) post I clearly mentioned that those 190MB I reported for KDE 4.4.1 do NOT include cache.



                Then, there is also the problem of not having information about processes running on the computer during the benchmarking ... but let's just stop here.

                Comment


                • Originally posted by << ⚛ >> View Post
                  (In my last post I said that it was my last post in this thread. Now it seems, I lied. But things have improved since then, a little bit.)
                  So what exactly should we be apologising for, what is it that offends your sensibilities so heavily one wonders.

                  Originally posted by << ⚛ >> View Post
                  Well, unlike you, I do know how Phoronix arrived at those numbers: the Phoronix numbers include the memory used by the disk cache.
                  Well if he's mixing together the memory usage of code+content with the amount the system has decided to cache due to what has been read off the disc then that seems a little inconsistent with the stated metrics being measured.

                  Comment


                  • It's a pity KDE 3.5.10 hasn't been tested.

                    Comment


                    • Originally posted by birdie View Post
                      It's a pity KDE 3.5.10 hasn't been tested.
                      Does *buntu still ship a 3.5 flavor?

                      Comment

                      Working...
                      X