Announcement

Collapse
No announcement yet.

KDE4 memory usage vs KDE3 => benchmark ?

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

  • #16
    Code:
    free -m
                 total       used       free     shared    buffers     cached
    Mem:           496        484         12          0         12        191
    -/+ buffers/cache:        279        217
    Swap:         1027          4       1022
    That's memory usage of kubuntu lucid on old thinkpad T40. One of the biggest memory hog is chromium browser. after closing that:

    Code:
     free -m
                 total       used       free     shared    buffers     cached
    Mem:           496        372        124          0         10        174
    -/+ buffers/cache:        187        309
    Swap:         1027         11       1016
    So memory usage drops from 279MB to 187MB only closing browser with one tab open.

    Comment


    • #17
      Originally posted by cb88 View Post
      I run KDE4 on my grandparents computer which they use for email internet etc... and KDE4 is definitly less optimal than KDE3

      It shouldn't require more than a 2Ghz celeron and 512Mb ram to run the desktop itself snappily (which it doesn't)


      I still run KDE3 on my laptop from the kde-sunset overlay on gentoo just because I like it that much more interface wise... some of the utilities are more polished in KDE4 but meh... I dislike the whole UI rendering and caching mentality they are using it makes it feel sluggish IMO compared to KDE3 it could be some other completion that causes it but that is what comes to mind.

      If you think KDE4 uses less ram I don't know what you must have been smoking... its bigger and slower. The exception to that might be quad core desktops with Nvidia graphics there I can see how it might be faster.
      and you turned off composite and nepomuk/strigi of course ...?

      Comment


      • #18
        Originally posted by birdie View Post
        I'm using Fedora 13, but both KDE3 and KDE4 are compiled and installed manually.

        I hate Fedora's default GCC flags, they are insane.
        Last time I tried Fedora had Gnome parts running in KDE (afaik there was also Metacity installed for some reason), so memory usage was higher - KDE libs + Gnome libs etc.

        Comment


        • #19
          Originally posted by kraftman View Post
          Last time I tried Fedora had Gnome parts running in KDE (afaik there was also Metacity installed for some reason), so memory usage was higher - KDE libs + Gnome libs etc.
          I don't run Fedora per se.

          And trust me I don't have any pieces of Gnome here.

          Code:
          [birdie@localhost ~]$ rpm -qa | grep -i gnome | wc
                0       0       0

          Comment


          • #20
            Originally posted by energyman View Post
            congratulation, you just measured crap.
            I hope you can put your money where your mouth is.

            Comment


            • #21
              @energyman composite on indexing off... it feels faster anyway with composite on either way window drawing is laggy I have tried both composite on and off.

              It is an ancient i810 video chip though compiz is about all its good at... maybe google earth.

              Comment


              • #22
                Originally posted by birdie View Post
                OK, here are the real world results:

                Memory usage before and after the start of:

                Code:
                KDE3:
                             total       used       free     shared    buffers     cached
                Mem:       3942172     383688    3558484          0       3928     101748
                Mem:       3942172     576068    3366104          0      30376     199300
                
                Total: 187MB
                
                KDE4
                
                             total       used       free     shared    buffers     cached
                Mem:       3942172     386824    3555348          0       4660     10475
                Mem:       3942172    1073012    2869160          0      24796     605136
                
                Total: 670MB
                Both environments were run with an empty, new profiles, and four applications which I just ran and kept open:

                Akregator, Konsole, Konqueror and Kolorpaint.

                Now, don't tell me KDE4 doesn't need a huge optimization.
                It's a lot smaller if you exclude the buffers/cache (which you'll get if you read the second line of the "free" command):

                Code:
                383688 - 3928 - 101748 = 278012
                576068 - 30376 - 199300 = 346392
                
                Difference: 68380 = 66MB
                
                386824 - 4660 - 10475 = 371689
                1073012 - 24796 - 605136 = 443080
                
                Difference: 71391 = 69MB
                If you really want to compare memory including cache, run this command to flush the disk cache before logging in:

                Code:
                su -c "echo 1 > /proc/sys/vm/drop_caches"
                After logging in, use each app for a few minutes before checking the stats so that it has a chance to reload whatever it actually needs.

                Another thing to keep in mind is that if you run KDM 3/4, some of the KDE libraries may already be loaded which will throw off the before/after logging in stats. Either way, it's not like KDE4 is using 400MB more than KDE3.

                Comment


                • #23
                  yeah, you have to reboot between that 'measurements' and empty profile means that nepomuk and strigi is on ...

                  Comment


                  • #24
                    thyrr

                    If you had bothered to read my message carefully you would have noticed that I did

                    echo 3 > /proc/sys/vm/drop_caches

                    which is far more effective.

                    And cache memory occupied by any of these environments DOES count, e.g. KDE 4 has a very big icons cache in /var/tmp/$USER-kde4/ which is almost all the time kept in memory by Linux kernel.

                    Comment


                    • #25
                      Why the heck no one dare make the same measurements and only criticizes me? And I tell you, rebooting is NOT necessary. That's stupid, since Linux allows to clean ALL caches.

                      Comment


                      • #26
                        KDE system monitor tells me that bare naked kde with all other parts of OS takes 350 mb and with the apps you mentioned around 407 mb.

                        kubuntu 64 bit kwin on, nepomunk off.

                        Comment


                        • #27
                          val-gaav you too measured crap. Why does nobody read the link I posted?

                          Comment


                          • #28
                            Actually I know it's not accurate ... but AFAIK not accurate means in this case that it's higher then it really is...

                            I really don't care enough about it to spend time on doing it the right way ... just that the value seems a lot lower then previous "crap" tests imply.

                            Comment


                            • #29
                              If that number includes shared memory, a higher amount could simply mean that the newer version uses more shared libraries (which might be loaded anyhow for different purposes), not that its own memory usage is higher.

                              Comment


                              • #30
                                Akondi and etc

                                Disabling Akondi, Desktop Effects, nepomuk will bring the memory usage down. The odd thing is I use to run KDE on a P2 350MHZ with 128 Megabytes of ram. KDE 3.3?

                                The real problem is Xorg 7.4 - 7.5 and Linux 2.6.22 - Linux 2.6.31. These things are the ones eating the ram up.

                                It's as if there were some secret instructions that executed in the older Xorg and Kernel that get turned off in the newer one's. CISC -> RISC.

                                As a test I took a copy of Slackware 12.1 which ships with I believe the 2.6.24.1 kernel and installed it to a Pentium 1 / 233mmx computer with 64 Megabytes of memory. It was swapping out at the command prompt with no Xorg loaded.

                                The 2.6 kernel's memory management is a hog. I haven't really read into any modern memeory management algorithms since I read the Dinosaur book. That's probably linux 2.2?

                                So I can't really be sure if Xorg has much to do with it at all.

                                Nobody cares about old junk lying around.

                                Comment

                                Working...
                                X