KDE 4.6.5 > StarCraft 2

I normally enjoy StarCraft 2 in the evening a bit or like now, if it is raining and I have no fun in any more real work.

But just now and some days ago, it has again proven, KDE > StarCraft 2.

Just again, StarCraft 2 (played with wine, thanks a lot wine team, you do a great job!) got sluggish and then terminated.
A quick dmesg call shows the cause.
Like last time, the OOM killer got me.

Now, first I thought: What, this stupid game uses more than my 4 GB of RAM? (I have no swap, have only one ssd)
But now, the reason is much easier: KDE has beaten StarCraft 2…. (even if we call it KDE SC, it still not loves SC 2 :)

Looking at my process table, is is just clear: StarCraft 2 is the biggest memory user, that is just normal, around 2.5 GB, at peak.
But no reason to crash, or? With 4 GB.
But wait, I let my kontact run the background, big failure.

My stats here:

Kontact and co, just only taken the > 50 MB RSS offenders, there are lot of other smaller processes running around:

cullmann 1452 2.1 10.7 1243552 436228 ? Sl 11:05 9:33 /usr/bin/kontact
cullmann 1132 0.5 9.6 813136 391996 ? S 10:38 2:33 /usr/bin/akonadi_nepomuk_email_feeder –identifier akonadi_nepomuk_email_feeder
cullmann 1127 2.6 1.3 550532 54860 ? Sl 10:38 12:13 /usr/bin/akonadi_imap_resource –identifier akonadi_imap_resource_0
cullmann 1060 2.0 2.8 481124 114288 ? Sl 10:38 9:34 /usr/bin/mysqld –defaults-file=/home/cullmann/.local/share/akonadi//mysql.conf –datadir=/h

Summa around 1 GB, with the smaller ones.

+ my friend the semantic desktop :/
cullmann 1104 41.6 2.1 168752 88320 ? SNl 10:38 193:59 /usr/bin/virtuoso-t +foreground +configfile /tmp/virtuoso_hX1095.ini +wait

The planet is no place for bugs, I know, but anyway, this is no bug report. What should I report at all? That just using KDE + Kontact eats >> 1 GB of my RAM and virtuoso around 40% of my CPU? And yes, it even uses that with Kontact shutdown and all other apps closed… And no, I have no indexing active, that OOM’d me last time, even before Kontact + Akonadi on :/

I kind of think, we need to improve in that area, really.
Beside, in the old days, if Kontact did take too much RAM, I just closed it, now, still the akonadi parts + mysql stay alive, with around 1/2 GB memory usage :/
Guess need to kill them hard before starting to play :)

A lot of persons at work switched back to GNOME or whatever, because we can’t get rid of these problems. Either kdeinit eats our CPU (with Ubuntu Natty) or nepomuk/strigi/akonadi and Co. eat the memory + fill our NFS homes :/ And no, we don’t do bug reports for this, as these are not reproducable problems. (Beside the memory usage I guess and that virtuoso runs amok, but that are known issues if you look at our bug tracker)

Good side of things: In that light, Kate’s memory usage is kind of non-existing in comparison :P

I will continue to use KDE for sure and try to be more clever and shut down Kontact and co before playing (and turn of nepomuk completly, even if I get here and then a error message because of this) or I will just buy more RAM, 4GB to ease the pain should be not the biggest issue. Happy enough, at work I have anyway 8GB or more in every workstation because of our static analysis software needing it ;)

73 thoughts on “KDE 4.6.5 > StarCraft 2

  1. > And no, we don’t do bug reports for this, as these are not reproducable problems

    I guess you should nevertheless, so this is known, other people could add there “Me too” and perhaps somebody else manages to reproduce it, by finding a common pattern.

    1. That rarely happens. I used to report most bugs I found right from KDE 3.x all the way somewhere till 4.4. After that I just stopped. I find and file bugs in my free time and that time is very limited.

  2. I just upgraded all the systems in my lab to Natty and noticed that KDE performance was significantly worse (unusable almost). I hadn’t heard about kdeinit taking so much CPU until you posted — I will have to look at top and see if that’s whats happening to us. Is this a known problem? Are there any workaround?


  3. I have this too since 4.7ß, reproduced more than once…
    Not sppeaking about plasma-desktop going above 80MB ram (never exceeded 40 before) nor kwin…
    Strange things with this 4.7 release…

  4. Yeah, kde4 is fucking very nice. Just a plain desktop environment which requires 8Gb of ram just to be able to run. Good job everyone. I do remember times when kde3 was happily running on my old desktop with 512 of ram and it was really fast consuming no more then a half of the memory. More and more I think about moving to some distro which can provide me with working kde3+amarok1.4. I really miss those days.

  5. Hm. First off, I’m using master / trunk of everything KDE. I also see KMail being a bit slower (e.g. listing folders) and using more memory than before, but the numbers are nowhere near yours. On my old 1 GB RAM laptop there is very little swapping when using it.
    To shut down Akonadi if you need to do that, run “akonadictl stop”. When you run “akonadictl start” afterwards in a console, you can see Akonadi debug output which might give you some hints.
    AFAIK Akonadi uses Nepomuk to implement message searching (which is very fast now), and when all messages are indexed it should stop using CPU. Especially if you disable automatic interval mail checking.
    It helps to check “always use this sorting” somewhere in KMail’s View menu to speed up folder listing. I think this means that most of the message threading information is cached on disk instead of recomputed each time.
    And yes, you should file bug reports. I need to do that, too.
    There are too many ways of using KMail to test them all among the KDEPIM developers (the software works for them), so they really need useful bug reports in order to fix things for the rest of us.

    1. I use 4.6.5 release, and nepomuk had time to index now since more than 8 hours, with 40%, still no change. I normally never do any kmail config, beside set imap account, smtp account, sort by date, flat and be done. Should the imap resource not shutdown itself if nobody uses it?

      1. I’d wait for 4.8 before re-enabling strigi/nepomuk/akonadi again… 4.7.3 is useable but not great, I hope 4.8 will fix many performance and memory issues. Kmail 2.x probably still won’t be like the 1.x series but then again, that one had years to mature :D

  6. Unfortunately, there is a design mistake with Nepomuk. If an indexer leaks memory (there are about 20 of them), then Virtuoso appears as the culprit, and people simply blames Nepomuk itself. I saw that with bangarangnepomukwriter, and the solution is to restart Nepomuk to free all leaked memory. Unfortunately (II), the command needed to quit Nepomuk is needlessly long and complex. The simplest version of it is:

    qdbus org.kde.NepomukServer /nepomukserver quit

    If you run that, you’ll shutdown your Nepomuk Server, freeing all leaked memory, and, if you restart your server with nepomukserver &, memory usage should be less.

    1. Thats nice, but thats not the thing I want to keep in mind to be able to play StarCraft. Could nepomuk not just restart itself then every hour or so to avoid this, if this is a known issue?

    2. “kquitapp NepomukServer” should work. But thank you for the idea, before I tried to kill all nepomukservices separately, and that was not that easy.

  7. Before nepomuk and akonadi, when you needed information you fetched from the hard disk, what’ would be the difference now?
    It shouldn’t consume so much ram, when it’s not needed it should go to cache memory, where anything can replace it. And when opening kmail, it should fetch relevant data from there. Plasma calendar uses akonadi too, but it shouldn’t take more than a few MB to keep contacts of current and next moth in ram shouldn’t it?
    I think akonadi/nepomuk should be more intelligent about what they store on ram or on hdd, to minimize this kind of things, if not using kmail there’s no need to have mails in ram, if not using rss there’s no need to keep rss in ram, etc, etc.
    About virtuoso continuously indexing things, I don’t really know why it happens when file indexing is not active, is it indexing the akonadi database for searching? does anyone know what is it really doing?

      1. If it is indexing mails, and as far as I remember in kmail 1 I could search mails, fast, and when I closed kmail it stopped everything, does that mean that it is doing a lot more of indexing? as before all the proccess fitted in the time I used kmail, now it doesn’t, that must mean 1 of 2 things, it is doing more work, or it is doing it inefficiently. That or mails are indexed only when kmail is not open or something like that.

        1. First of all, the old KMail did do less than the new Akonadi. For example, Nepomuk does a full mail index, the old KMail only did headers. Remember how long it took to search the CONTENT of mails as opposed to the quick filter in a folder?

          Second, I’m sure KMail2 simply is far less efficient still. It’s backend (akonadi) is quite different from the completely optimized KMail1 and will take time to mature.

          Third, Akonadi & Nepomuk are more flexible and more componentized – I bet that while thats great from a reliability and maintenance pov, it probably means they will always be a bit slower.

  8. People report this kind of problems since … well … err … maybe KDE 4.0 and nothing happened, so why should this change now.
    I am pessimistic on this, as usual then someone says “hey, its your fault, switch this config or that” or “just run this command every hour and everything is fine” or “Hey 4 GB more RAM cost only … ” or “…”

    1. I would not be that pesimistic. A lot of stuff improved since 4.0. Kate was a pain in 4.0, too. It improved a lot. KWin is really a nice windows manager, never got problems in the 4.x time and it got better each release. Same for konsole :)

      1. Yes, there are of course improvements. Performance got better, bugs got less and so on. But as far as i remember there were complaints about a more and more “bloated” KDE, and that complaints usually got answered like i said.
        And your post all in all is about a “bloated” KDE.
        I really like KDE, and will stick with it, even if there are some things i dont like.

        1. no, his post is about a bloated nepomuk/virtuoso/kontact.

          rather unfair and inaccurate to then lump every piece of software developed by any KDE team into that category.

          (interestingly, the specific problem Christoph notes in this blog entry are evidently specific to individual installations. it’s either a specific sort of configuration or combination of data that triggers it, as some see these problems while many others don’t. it would be interesting to see what it is that triggers this set of problems identified. some have noted that nepomuk indexers may be culprits here, and then i wonder “which one(s)”?)

          1. Actually, this happens in both Ubuntu and ArchLinux standard installations without any options set and strigi turned off.
            I doubt that this is a settings specific problem.
            But as this is more a post from the user field of view, I won’t be really able to debug this given I did just get that with a release build from packages.

          2. So true. People tend to forget that kde 3.0 were not much better then 4.0, and as 3.x improved into solid and loved piece of software, so do 4.x . If we could leave akonadi/nepomuk/strigi outside, 4.x would already be wonderful. But aaron, please note that with recent libs and with kmail2 out, akonadi & co. is no longer “outside”. It is stuffed down our throats – and choke.

          3. > rather unfair and inaccurate to then lump every piece of software developed by any KDE team into that category.

            He is criticizing a software chosen by the KDE team to be in the default install, so IMHO it is fair to criticize “KDE”.

          4. For you as developer who has more insight it may sound unfair. For me as user, who dont know why kded4 is using 100% CPU (i read it may be one plugin, but dont know which one), it is KDEs fault.
            If dolphin uses all RAM and CPU while hovering over an mp4 video (I think this problem is gone mostly) for me as User it is KDE (think it was strigi or something, dont know). As developer you may know where the problem is, i dont know this, so for me it is KDE.
            So you are right that it is inaccurate, but as user i only can describe the symptoms. If I’d know where the problem is i would of course fix it myself, but i dont know.

          5. It is a combination of the number of mails you have (certainly an issue, KMail2 doesn’t seem to scale as well as KMail1) and some other config things. There certainly are ‘good’ setups as well as bad…

  9. A lot improved, but Nepomuk & Strigi aren’t useable when leaking memory and using too much cpu horse power are still the major drawbacks of both technologies. I really don’t get the problem here, because the “Gnome world” with beagled, trackerd and so on never were so problematic as nepomuk & strigi are right now. I know that nepomuk has semantic search enabled. But in the end these are just a few more entries in a search index so no big black magic.
    Either this issues will be addressed or people will disable it and don’t use it. (with kmail 2 now out and akonadi urging you to enable nepomuk [it will pop up dozens of messages to inform you that nepomuk isn’t enabled and that you should enable it] it will be hard [if you want to use kmail2])

  10. That’s why even with 4GB of RAM I’m still running good old mellow KDE 3.5.10.

    I don’t need eye candy, I don’t need plasmoids, I don’t need translucency everywhere, but I do enjoy speed and low memory footprint. Also I need decent applications, like KSensors (and don’t remind me of KDE’s plasmoid – it only shows temperatures). A lot of KDE developers are so busy adding features, they’ve somehow forgotten about optimization and bug fixing. There are now 10 years old bugs filed against _core_ KDE components, and applications like KMail, but why should one fix bugs/implement features required by _your_ users (most hated/severe bugs at https://bugs.kde.org/), when we can add new shiny features.

    Funnily KDE devs think about tailoring KDE to smartphones and tablets, but have they every thought that most “smart” devices have less than 1GB RAM and only a part of it is available for user applications? Ah, and MeeGo is almost dead. How have I overlooked that?

    I’m sorry for ranting, this post is just a fine excuse to rant again :)

    P.S. Luckily for KDE developers MS is moving in the same direction, so you’ve got your excuse. Win7 64 (with superfetch disabled) can easily consume 800-1GB of RAM just for itself with zero user applications running.

    1. I would not say KDE developers don’t care for the bugs. I care for them, and other Kate and KDE devs, too. It is sometimes just very hard to fix some bugs, as they would need larger changes for which nobody really has the time. (at least that is sometimes the case for Kate bugs, like: support that I can load 10 GB files, sorry, that work won’t be done by me)

      In the above shown case of akonadi + nepomuk, I guess some polish is missing. Perhaps its because this is the first release with this pair for Kontact and co. and I hope we improve, as always. Unluckily, I have no time to help :(

      1. I cannot disagree with you, but such intrusive technologies that akonadi + nepomuk bring should have never been released in the alpha state and by reading numerous posts all around the net, I get an impression that core KDE 4 components are still not stable, mature and generally bug free.

        I’m sorry but that’s unacceptable and it’s all saddening me. This is one of the reasons I’m not going to install Linux on any of my friends computers in the nearest future (even though I’ve been running it exclusively for more than ten years – but I’m not an average PC user at all). I simply don’t want people to suddenly start hating me :)

        P.S. I’ve never had any problems with Kate though. :)

  11. My question would be on the design of entire semantic solution.

    Why should all resources be running all the time? It must be possible to have a single, low footprint, process to keep checking for changes on remote servers. Like just run simple request, and check if it hadn’t changed since last time. Only then there are some changes – fire up the resource, grab updates and kill them all. After all, all data is stored in a single database, that can be accessed no matter if resource is up or down.

    Take an example of SquirelMail – the very old php based webmail client. It manages to render a page (including fetching list of messages) in less than a half of second with like 16-20 mb of ram usage. Checking email each minute would be like 0.2MB average usage. So even if user would use 20 resources, having Akonadi written in php style and code (and php is not the fastest thing in the world), it would be like10MB average RAM usage…

  12. KDE is so great overall, it’s really sad to see, several years after the release of 4.0, that Nepomuk/Strigi/Akonadi are still not that usable. I always disable Nepomuk & Strigi, because it always eats *tons* of memory and slows down the system. This is pretty unfortunate as it was very promising and I’m sure a huuuge amount of great work was done by the team…

    But when those are disabled, KDE 4.x is pretty pleasant to use !

  13. I have found Nepomuk quite usable as long as I keep it out of my dev folder (the one that contains everything I have in subversion or git). I haven’t used Kontact in years so I can’t say much about akonadi.

  14. The solution is that Semantics in the fileystem should not be programmed in userspace. This is where a million things can go wrong and another million constantly-on performance heavy dependencenies and monitors are running.

    IMHO a separate project needs to replace nepomuk core which implements semantic data as an extension to any widely used core and proven linux filesystem (preferably) in collaboration with its upstream.

      1. No, it would make it much better in my opinion. I mean only the semantic subsystem. You’re talking about strigi the indexer and soprano. What the constant activity is doing I assume is scanning and writing semantic data to an ever growing larger userspace database which, as the database grows larger, the write operations and data read lookup time increase. Whereas kernel space each semantic write is as effectively neglible as a simple file data write. Same for the read, and all without an ever growing costly userspace database. By the way, this database system is what is targeted by some as part of “bloat”, or at least performance costly operators (ie, sesame2, virtuoso, and [x] in the future). In filesystem space, the costly sesame2/virtuoso/x databasing can disappear forever (and filesystem write/read performance is only as costly as the underlying architecture, ext4, btrfs, etc). Nepomuk’s core theory is very low level (associating arbitrary data with grouped file block data), and filesystem programmers have already perfected file metadata. In my mind, a technology that attempts to be a core part of the filesystem and which can have such drastic performance effects (vis-a-vis current filesystem-to-userspace workarounds) should start in the core of the filesystem after all. Just my opinion.

  15. Same here, to play sc2 I have to stop akonadi, nepomuk, desktop effects and twinview. Unfortunatelly nepomuk seams to rescan everything after enabling it again :(

    1. KDE needs a daemon for scheduling cpu/ram/hdd intensive tasks. In general, no two hdd intensive tasks should be running at the same time, or big performance penalty kicks in as half of the time is wasted whine HDD head keeps moving back and forward.

      Think of starting a firefox with any daemon in the background; copying files to USB key while KTorrent checks files after download; Searching for files, while email is being fetched and indexed. Or worst situation – when all services runs at the same time

      Would it be rocket science to have all intensive tasks to be prioritized and grant permission only when all hardware components is at least 70-80% idle OR task is to be executed on explicit user request.

      1. For god sake, no.
        Such multitasking things are *completely* the job of the kernel scheduler and kernelspace in general. All that has to be done (and should be done) in userspace is defining processes priorities (‘nice’ values).
        Please update/tune your kernel and/or your hardware if you’re not satisfied with your machine multitasking performance.

  16. And with all these wonderful things that happen Aaron still insist that this joke called semantic desktop is the core of KDE and you should feel bad if you don’t want it. How nice :}

    1. While I can understand some of the frustration, anonymous comments like this (attacking specific developers) are just rude and do not contribute at all in solving the issue. It’s just a matter of implementing it correctly, so it does not get in the user’s way. This goal is obviously not completely achieved. It needs time, and testing. Clearly, it’s not perfect, yet. But it improves over time.

      1. How many more years? 3-5? Semantic desktop is pain. This should be a separate project, not a core component. Because of all this confusion with semantic desktop KDE has a bad press. Number of users over time will decrease because of this. Maybe is time to admit, that this overwhelmed the developers.
        KDE is very good, without semantic desktop, will be perfect.

      2. ‘anonymous comments like this (attacking specific developers)’

        well, Christoph, your blog entry led the way. too often our community encourages a culture of sloppy negativity performed in unuseful contexts. you noted in the blog entry itself that t wasn’t the best place for such a report and that it wasn’t even a useful bug report in the first place due to vagueness. harmless among friends, but over the years we’ve taught others through such actions that the inappropriate is indeed acceptable. if a core developer does it, of course others who follow the project will as well.

        btw, you can turn off nepomuk in the control panel if it gives you problems. i used to have to do this on my day-to-day system up until about 9 months ago or so when improvements to it made it finally work well enough for always-on running. you lose full text search in kontact and a few other such nice features, but if it really gets in your way, then it can be a worthwhile trade-off.

        btw, one reason akonadi may stay alive when kontact goes away is the calendar integration of events in the clock plasmoid(s). go into the configuration of the clock(s) you have running and turn off “display events” under the calendar tab. in recent versions this prevents all touching of akonadi on clock startup (where before it would still connect to it then drop the connection immediately after; obviously wasteful and now fixed)

        1. Well, I just showed the current state, like I see it on my machine. Its the same with both ArchLinux or Ubuntu, if I use KDE 4.6.5.

          If you feel there is no place for any critic on the planet, that’s sad.

          I know I can turn of nepomuk, but that leads to an notification on each kontact start that I miss nepomuk. Beside, IMHO KDE should be ok to use for me without any configuration. I installed it from scratch and have no custom .kde settings beside disabling Strigi, after it swapped me away last week.

          If you read the comments, you will see, that it is no standalone view of myself and others have the same problem. What’s your problem with voicing it?

        2. Well, I’m not saying you are a bad man or anything, it’s just you don’t want to admit that many KDE users do not need said semantic desktop despite multiple examples that contradict your viewpoint.

  17. “A lot of persons at work switched back to GNOME or whatever, because we can’t get rid of these problems.”

    I can’t imagine someone switching to disaster like gnome3 :P There are huge memory leaks in gnome’s hell and its overall concept is terribly broken. As for KDE I’d like to see akonadi and co. disabled by default.

  18. No, it’s not a matter of implementing it correctly.
    A project this large should be built around a modular structure of components with the right degree of isolation. You should be able to EASILY and independently disable anyone of them, impacting just related features and nothing else.
    Kontact should work (within its limits) without akonadi, and akonadi without nepomuk, and nepomuk without soprano or strigi, and the desktop without all of this.
    If this is not the case, then even the best implementation cannot help.

    It’s not a matter of integration… integration, done the right way, empowers, done the wrong way, imprisons.
    Judge by yourself which one is true, here.

  19. The concept of priorities/scheduling should be implemented:

    resource hungry apps/process should be run with least priorites unless it is really required,

    for example, updatedb with ; nice -n20 updatedb -v ; #takes longer but it does not hamper or slow my system down.

    Strigi or whatever resource consuming process must be run with least priority; like “nice -n 20” (the least)

    But I can understand the conundrum of kde developers, they are using computers for developement hence they have big super-fast RAM, Harddisks, Processors; Obviously, they can’t and won’t want to see the efficiency of their applications; because if it works “just fine” for them, they’d say “you lusers are liars, we have everything working here nicely, you morons should get a better (read developer’s) system not an ancient (read 2-3 years old) desktop or laptop”; cause this is free software; and if you can’t code it… bear with it… lusers!”

    It’s a sorry state…

    I upgraded my nvidia 8400gs to 9800gt for KDE. pentium 4 512mb to core2duo 2.4 with 4gb ram; still i find it sluggish;

  20. Unfortunately I have the same problem. My 4 gb ram were enough for quite a long time but nowadays often I have to wait ages for swaping when a process runs “bersek” again. Really drives me mad when I actually want to do some coding but have to track down stuff.

    After all it gets frustrating. KDE is simply not a system I can suggest anybody to use even though I use it for ages and started coding on some parts of it a few years ago –> actually KDE was the reason for me to dig more into coding in the first place.
    During the whole 4.X cycle there was always something that was highly unstable or a resource hog.
    I never had that many crashes using any other system. Even Windows 95 ran more stable. With 4.6 I have at least a Dolphin crash per day and I know this is worked around in 4.7 though it is still unfortunate.

    And yeah I myself too caused some crashes in the programs I work on so in some cases I can point in my own direction.
    Nonetheless when I just want to use things it gets annoying, no matter if I am to blame for a specfic issue. ;)

  21. Haha, funny that I”m not the only one with this problem. After importing my Kontact 1 mails (which took absurdly long btw), Kontact2 + all that semantic mumbojumbo used about 1GB of RAM, moreover just browsing my mail maxed out one of my cores.
    Well, I just disabled Nepomuk, Virtuoso, Strigi etc. (I never found any positive use in them, and I really tried…) and switched to Google Mail for my daily Email tasks.
    I don”t want to bash KDE btw, I think it rocks (and I still remember all the tears that were shed over KDE 4.0 :D )

  22. I think it’s wrong to mix things up.

    First of all, to claim that it is not possible to use KDE4 productively on < 4 GB RAM or that one cannot use its effects with some pretty old nvidia 7600 and needs a 9xxx is plain bs. Honestly, if one starts a comment with that claim – you cannot expect to be taken seriously. And I guess these people don't since they have a very frustrating life which causes this kind of attitude.

    The main problem I see is that those people createan atmosphere which makes it kind of impossible to point out or accept problems without offending anybody. Saying that akonadi/nepomuk et al do have memory issues is a fact since it happens on a lot of computers. In some cases it might be some config – yet denying the fact that they eat lots of memory on many systems is silly and appears to be some defense mechanism or whatever.

    I have not read yet that memory usage was measured completely wrong and that in fact akonadi only uses some kb. Same killer phrase kind of argument. If some process shows 1GB memory consumption in top or whatever tool there is something wrong – no matter whether its some percentage more or less.

    So I would like to see that people stop claiming that KDE4 is unusable because of some issues. That's just not true. And I would like to see that some devs/users stop denying that there are serious issues that do lead to a massive memory consumption in case of akonadi/nepomuk or annoyances in other areas. If one works on apps/services that are always running it's only natural that those apps get more attention and thus their bugs are more annoying than bugs in apps that one starts and quits after usage. That's why plasma and its issues will always be more "popular" than some grave bug in application xy.

  23. Same here at work, people stopped using KDE due to the amount of ram all those services take.
    I’m a big fan of KDE, but as a developer I can’t use a desktop that requires MySql to run.

    ps: please kill Kickoff

    1. May I ask, why you want to kill Kickoff (the menu starter, right?)?^^
      I rarely use the menu starter, but when I have to use it, I think it works quite well. :-)

      (if it is, because you do not like how it works, you can switch to a classic menu starter, which AFAIK is also part of the default Plasma widgets :-))

    2. Why can you not use a desktop that requires embedded mysql? Please give some objective arguments. “I don’t like” is not objective. Claiming that this make it impossible to use on hardware with < 2 GB would be a lie. So I'm eager to hear the objective reasoning.

      From my point of view I can say that mysql never ever got in my way or wasted hundreds of MB RAM so that I had to kill it. I did however notice sqlite's inferior speed when it is used within e.g. digikam (e.g. writing thumbnails). Since there is no perfect fit for all goals, i.e. minimal memory footprint, highest speed, common within distros one does have to make a choice. sqlite is slow, not a good choice for something that acts as service/central storage/cache. So the choice includes which people will not like you. If you pick the db with the lowest memory footprint but not highest speed, people will complain it is too slow. If you pick the fast but more memory consuming other people will complain that there would be a less resource hungry alternative – and so on.

      I wonder why there are so many that allegedly know a better solution but apparently not a single one who codes that back-end. What are the odds? Until those many present some code a project like KDE which has limited resources has to concentrate on one back-end and that one has to be quick and common. Sqlite is not quick. Who cares about some MB of RAM if you get speed for it? Yep, there are some, but my claim is that there are more who appreciate speed.

      And what type of argument is "please kill kickoff"? No reason given, no replacement offered. Neither constructive nor objective but just useless!

      That's exactly what is not needed if one wants to create an atmosphere between users and developers where one can point out existing issues and discuss them in a constructive manner.

      1. When I install any KDE component, MySql and it’s libs gets installed. I simply don’t want to deal with multiple versions of mysql and libs in my environment. I don’t have to deal with this in other DEs.
        “I wonder why there are so many that allegedly know a better solution” There’s no better solution because there was never a problem. Semantic desktop, plasma activities, nepomuk, akonadi… they are all trying to solve problems people don’t have.

        “And what type of argument is “please kill kickoff”?” That’s no argument at all. But the standard reponse “I hardly use kickoff” is not a proper response as well (I know it’s not your response, but that’s what everybody gets when complaining about kickoff).

        I’ve seen Linux usage increase a lot amongst developers, but KDE nowadays is a rare sight. :(

        1. I beg to differ. There are many developers, users who are using KDE4. Many of them are ex-gnome users, because they couldn’t stand disaster like gnome3. KDE’s really the best DE out there.

        2. Why do you have multiple versions of mysql installed? Here all the mysql packages have the same version.

          And why do you have to deal with those versions? If you do not look into the process list you will not even notice that there is mysql running, simply because you do not have to set it up, configure or deal with it in any manner.

          The claim that the semantic desktop is trying to solve problems that people never had seems a bit weak. Of course one can argue that oneself does not find it useful and as with everything there is always a group of people who share that opinion – but to generalise it? There are desktop searches as the one from google etc. – so there seems to be a need for that. Tagging and commenting spreads all over todays’ devices and apps – so there seems to be a need for that as well. And since it does not make sense to tag something in one app but not being able to see or manage those tags in another app (e.g. file manager + photo management) it seems sensible to have a central storage for it including the possibility to write those tags to the files where it makes sense. That’s part of nepomuk + strigi.

          Akonadi is connected to nepomuk but a different type of service, so please do not put them all in one bag. It is just a cache/storage facility. Sharing functionality across apps is common and helps to save development resources. Why would every app have to code how to deal with a ldap addressbook or how to retrieve and store emails? With akonadi and its agents people can concentrate on their app and leave the data management to akonadi.

          In a perfect world there would be enough resources to make everything perfect – but there isn’t in the real world. So trying to share functionality and code seems like a sensible thing to do. Of course there is always room for improvement – but to put akonadi/nepomuk/strigi into one box and simply claim that they are useless does not help.

          And regarding the developers. There were some articles on committing rate on the planet some time ago. They are increasing. So either the developers code more or there are more developers.

          1. regarding MySql

            MySql is probably one of the best database solutions around by no surprise used by majority of websites, where huge workloads requires superior stability, performance and hardware efficiency. The question should not be wherever to use MySql, but why is MySql is runing in embedded mode. Embedded should be used only as fallback

            De facto, KDE would be functionless without MySql (Akonadi, Amarok), so there would not be any additional overhead to have separate easy-to-access server running. Actually it would bring a benefit of factor n-1, as only one instance would be used for all applications. Just set any memory limit, and MySql will totally respect it.

            But the key benefit – any application, that otherwise couldn’t justify using MySql, would have access to it with close to zero footprint. think of DigiKam, plasmoids.

  24. I think these issues should be investigated a lot more, what can we as the user do?, would reporting bugs help?
    Some explanation is needed, why would just some mail and contacts take up 1gb of ram , is every mail cached in ram?, I would say that’s a bug, who is the culprit if it does?, nepomuk? akonadi? strigi?
    Something I think would really help to calm people, would be an explanation of how nepomuk/akonadi works, and what’s that 1gb of ram, I can put more text in 1gb of data that I could read in years, so what’s in there? The frustration of not understanding, makes people mad.
    Akonadi would help making better pim programs, it helps to share data between programs, and do LOTS of awesome stuff, but these rough edges need to be polished.

  25. I tried setting up Akonadi on a brand new computer with 4.7 RC1. I set it up to run on Friday left it to run over the weekend, and when I came back Monday nepomukservices was using over 2 GB of RAM. This is on a brand-new computer with literally none of my files on it besides the emails retrieved through Akonadi.

    1. There are issues with nepomuk – no doubt. Unfortunately there are bugs in RC1 and 2 which make it kind of useless, because it hardly indexes anything and constantly uses the CPU for nothing. Bad luck, but these will be fixed for the final version and are not even only nepomuk’s fault but soprano playing a role.

      However, using available RAM is good. Free RAM is of no use at all. If there is RAM needed by other apps the real quest begins, i.e. what does the kernel do, what does nepomuk do.

      So yes, it’s annoying that there are bugs in the RCs – but that’s just how it is and as long as they are fixed for the final it should be ok. Nepomuk or any other app using available RAM is not a bad thing per se. Leaking memory is bad, grinding the hard disk is bad (a lot worse than high CPU usage regarding slowdown), using the CPU for nothing is bad.

      1. Using RAM is bad, too, if it is not necessary. In my KDE 4.6.5 case, I doubt I will ever benefit from a half GB IMAP stuff in memory even if Kontact is closed. The problem with this contra stuff like disk cache: The kernel can reuse the disc cache memory and expire it => StarCraft stays alive, whereas the memory claimed by akonadi will lead to try of swapping => crash (in my setup, if I would have swap it would write away my poor SSD for nothing)
        But yes, that are bugs, thats no issue per design, only fact is: There are such bugs in the stable release series, which is kind of bad.
        And btw., I know of many bugs in stable Kate, too, but at least, they only occur during using Kate but not delayed even afterwards or during the whole session runtime.

      2. I am not sure whether it was the RAM, hard drive grinding, or CPU but the system was almost completely unresponsive.

  26. Hi Christoph,
    I think it is cool that someone talks about these problems. A lot of people experience various problems like these, and talking about them will help understanding them, and hopefully to fix them at some point. IMO not discussing the problems harms KDE much more than doing so.

    1. I hope so. And beside, I am still using Kontact, its no bad application, actually I use it since it exists (or KMail before). And akonadi does its job function wise here, too. My IMAP account works ok (even if migration failed, which is no big problem for me). Just the rough edges need to be fixed, perhaps KDE 4.7 will do the trick, we will see. Only one thing is not acceptable: denying that any problems exist. They do, but for sure they will be solvable. And I appreciate the work of all pim devs!

  27. Thanks very much for bringing this up. I, too, have been finding KDE completely unusable on 2GB of RAM, and barely usable but still laggy on 4GB of RAM (I added 2 after a day of complete frustration).

    Like Mozilla, we need to put together a team just to hunt down memory issues.

  28. And currently I use latest KDE SC with 512MB or 1GB RAM with 1.6Ghz Atom (N270) and P4 1.7Ghz Mobile CPU.

    The only different is that I dont have Nvidia card. Everything is smooth without problems.
    Oh, and forgot almost to mention, both are laptops. Netbook and a 7 year old Acer Travelmate

    I do have a desktop computer (main use) with 4 year old Nvidia 8600GT and 4GB RAM and 4000+ CPU. But on that computer Plasma lags and gets wrongly drawn graphic bugs.

    So superior computer when compared to old/cheap laptop is on its knees prying mercy and only because nvidia.

    And I dont even like Nepomuk/Strigi. I love their ideas and features. But I really hate the reality what they are with akonadi. Every time I log in, Akonadi pop up saying that it need to convert address book and it was already done. Or it need to do some other checks. I dont even have any app started to use those!
    Nepomuk can be great technology, but the GUI to use its features are non existing. The Dolphin search bar is terrible, it really should be a sidepanel and not a pop-up panel top of the view. There should be very easy way to add semantic data to files when saving or even opening. Not automatic functions but pure human work where person types the info what is needed.

    Since 4.2 the KDE has developed too many features or functions what are not well designed. Simply saying… the were under pressure of users to implent all old features without re-thinking how to do them well.

    I would say two things need to be re-designed and re-implented and they are semantic functionality and activities
    And then there are minor things like mentioned dolphin GUI and so on.

    I only want to create files, add some metadata to them and then later find them easily what ever app I am using from open/save dialog. I want my desktop to list those files. I want my workload to be easier by allowing me to maintain even better order on my computer without just relying folders.

  29. I’m on a 4 year old laptop with 3GB and an NVidia 7600, and KDE is running quite nicely in general. I have the occasional crash, but not often, and then it’s mostly caused by non-core stuff such as bleeding-edge amarok. True, I miss the days when KDE3 used less than 100 MB after first login, but still, KDE4 is more than 200 MB below what Window 7 uses after logon, and still provides much more oomph under the hood. I run memory hogs such as Firefox, X-Plane and VMs and use a ramdisk to compile my Gentoo packages, and very seldomly actually run out of RAM.

    I can understand the frustration and rejection many here express, I sympathise and in parts join the chorus. Example: to load and show my ~, Dolphin needs three to five seconds, whereas Thunar does the same in less than one. The apparent playfulness of some devs which gets a mention here and there doesn’t help much in this respect either (i.e. why, oh just why does Dolphin show all columns except the filename in a toned colouring which makes them illegible from a distance?).

    But to stay on topic: when I open KMail 2 for the first time after boot, it feels like it needs more than a minute just to display my folder structure, accompanied by constant HDD load. It isn’t that much, I only have about 35 folders, of wich only a handful actually change regularly, plus a disconnected IMAP. When that loading doesn’t happen fast enough (which is most of the times), then the folders are shown collapsed. It then takes another long while to display the content of my mailing list folders, again with lots of HDD rattling.
    As fas as I understand it, Akonadi is to serve as a _cached_ interface between the application layer and the actual storage layer. Hence I am quite frustrated that something that is supposed to accelerate access actually makes it take longer. Why use a sophisticated database if you don’t use it to get results quickly?

    My 2 centi-credits. Qapla’

  30. Near the end of trying/testing (for about 4 days now) KDE SC 4.7.0 on 3 different computers (running ArchLinux latest):

    Desktop: Core2Duo on DP45SG (EE) + 4 GB DDR3 + Geforce 9800 GT (blob) + 7200 RPM HDDs on RAID 0
    Desktop: Core2Quad + 4 GB DDR2 + Geforce 7300 (blob) + 7200 RPM HDD
    Laptop (A/C): Core2Quad + 2 GB DDR2 + Radeon 4300HD (both radeon and catalyst) + 5400 RPM HDD

    None of these computers managed to run KDE while still keeping fairly responsive – not anywhere near GNOME2/3/Xfce. Did enough research on these so called bloats & co. in KDE and did everything possible/advised to turn them off for good. Tried turning off eye candies and co. too. What more? Hunted down and eliminated all possible GNOME/GTK leftovers, just in case. CPUs still kept burning while idle. And, then, after many more rounds of tic-tac-toes, we are finally giving up. Going back to GNOME 3.

    I must admit, GNOME lacks some good/complete apps in at least a few area. GNOME 3 (Shell & Co.) is not even complete or perfectly usable (specially on dual-head display and/or for the power user). But, they managed to make a responsive and resource friendly desktop at their very first .0 release, which is, indeed, very important.

    Side by side, I installed GNOME 3 recently on a few more computers owned by novice users (among F&F). Surprisingly, they did not suffer from the problems we power users keep complaining about. It took me some manual tweaks to shape up the shell for desktop use though, which otherwise, suits some imaginary, under research and/or not-yet-released tablets/phones. Overall, GNOME 3 still usable for the average joe, today.

    Another thing is that we have been using GNOME 3 since the RC days (any Archer out there?). Can’t remember any segfaults since then as opposed to a few dozens encountered with KDE, after it’s .7th release. I wonder how it was at KDE 4.0 – never got a chance to run that on real hardware. Trials of 4.1 on VBox didn’t impress us to enough to try it for real.

    Final words. KDE keeps me thinking if are all of these machies are too “low end” for KDE. Again, the entire Linux Desktop developer community keeps me thinking why we are failing to work together towards a nice, stable and complete desktop platform. Our fragmented achievements so far, under different camps, under different religions, has been so good than anyone could imagine! No doubt there. But, many of you out there should agree with me, that we could do better had we have a common goal and sense of responsibility. Freedom, as in software, is not about fragmentation (no matter what you are influenced with) and ego clashes if I got the message correctly. We are in need of a true “Unity” today (again, not the Canonical joke).

Leave a Reply

Your email address will not be published. Required fields are marked *

Scroll to top