Hello, I've been very busy with school so I haven't had time to write much, but I have been keeping tabs on the technology world. During my daily news search I ran across this interesting article on the performance across different desktop environments. KDE did not do so well in this test.
http://www.phoronix.com/scan.php?page=article&item=linux_desktop_vitals&num=1
Any thoughts on this?
Subscribe to:
Post Comments (Atom)
Posts
Those tests are flawed. They represent Kubuntu, not KDE. If most data engines are turned off, compositing turned off, kde services slimmed down to feature partity with Gnome and doing a good packaging job, I mean, with ressources, not just "cmakekdefull" macro + fakeroot (-Os for some packages, -O3 for other, as-needed and fPic tests per packages to see what work the best), then it may be fair. Gnome in Ubuntu get great care and love. The KDE team have less money and workforce.
ReplyDeleteIt show in performances tests like that one.
1) Phoronix is a site run by a single guy who runs his own "test suite" on whatever system he happens to have installed (usually Ubuntu). Sometimes, his "reviews" catch on and people actually talk about them.
ReplyDelete2) KDE 4 is a _lot_ slower than KDE 3 and it gets slower with every minor release. Enabling things like Nepomuk by default and not even asking just shows that the devs are not forced to use older systems and do not give a damn about people who are.
dont get it, I'm using Mandriva and a fresh kde uses about 200 mb and there is some pyton app in the back taking 13 mb so its less than 190 mb, no ware near the 500 mb he was talking about.
ReplyDeleteThese benchmarks have to be considered. I'm suffering because Strigi likes to eat all of the memory (1 GB, it's an old computer), because it wants to index a video. Yes, I know it can be slow, but Strigi doesn't know how to limit its memory consumption, and it's slow. KDE has a lot of room for improvement, and I wish to watch some memory improvements in 4.5.
ReplyDeleteI read through the first 17 pages of comments on that article. Someone dug through the source of the test suite, and apparently it counts caches in the "used memory". I don't know about Gnome, but KDE4 (in every revision I remember using) seems really cache-happy.
ReplyDeleteThere were other thoughts that since the test system was normal Ubuntu with KDE and the others installed after the fact some of the Gnome (or just normal Ubuntu) services might be running in all desktop environments. Pretty much saying that Gnome/GTK+ stuff might needlessly be running in the KDE environment. Not sure how much of that is true.
Pretty much the first part of the first Anonymous post holds true: Phoronix is effectively one dude, and he rarely makes any attempt to hide his bias towards Ubuntu proper and Gnome. That isn't necessarily to say that he would deliberately stack the deck against non-Gnome systems, but he most certainly wouldn't go out of his way to make sure the comparison was a fair as possible. Hell, the first time he gave KDE top-billing in one of those test suite articles, it was a "shoot-out" between Kubuntu and PC-BSD. Many months later, I'm still trying to figure out what that's supposed to accomplish.
I'm actually a little surprised at how well it's actually done.
ReplyDeleteWhen you think about it, KDE has an incredible amount of punch compared to the other desktop environments. From things like KDE animating just about every element of the UI, to having a single built-in search system, to the functionality of plasma; I was quite surprised that it did as well as it did - even if the test was 'flawed', KDE did respectably.
Right now my RAM usage is at 530mb or so, and I have stuff like Firefox (~130mb) on and nepomuk (~50mb) indexing in the background. So it looks like that was not a lean installation by any means.
ReplyDeleteRegarding nepomuk/strigi, in 4.4 you can set a limit on its memory usage - a config option which in my experience so far is well respected. It also automatically suspends indexing whenever CPU loads gets up. Pretty impressive IMO.
Anon up near the top's first point was ok, but then point 2) runs out one of the usual lame wines... what is 'and not even asking?' supposed to mean? Of all the zillion services that are regularly configured to start automatically (by distro's, not up stream desk top software providers, BTW) exactly which ones should they ask permission for, and which ones are Ok to run without your blessing? And of course, they'll be the same for everyone under every circumstance, I'm sure.
ReplyDeleteFrom what I understand, the PTS is actually run by two guys. As for bias, read the comments on this page and you will find bias!
ReplyDeleteAlso, the responses to this test from the KDE camp are typical of what the PTS guys present in their article called "The Five Stages of Benchmark Loss" (http://www.phoronix.com/scan.php?page=article&item=pts_scale_2010&num=1) where people that dislike the outcome of a benchmark will go through five stages, which are:
1 .Shock
2. Denial
3. Discreditation
4. Analysis
5. Acceptance
Problem is, you might not reach past point 3 which is a shame. I am sure that none of you would be posting and saying that this test was flawed if KDE came out on top.
And no, I am not a Gnome user. I have been a faithful KDE user since 2.0.
Comparing a full blown kde install to a full gnome install is like comparing a BMW 7 series to a BMW 3 series, the 7 is heavier and of course needs more energy, but hey, it might even have a button for scratching your butt!
ReplyDeleteNobody expects an ide with this stuff like plasma or even nepomuk consuming as much memory as an ide with win95-style desktop and find -iname "foo" as a search engine. Or are you still using twm?
This is supposed to be stage 5 btw ;)
It really seems that Kubuntu may not be a typical KDE installation when it comes to resource requirements.
ReplyDeleteSee
http://thelinuxexperiment.com/guinea-pigs/tyler-b/big-distributions-little-ram/
where both Mandriva and OpenSUSE (and these are not considered particularly lightweight) consume significantly less than Kubuntu.
@anonymous
ReplyDelete"1 .Shock
2. Denial
3. Discreditation
4. Analysis
5. Acceptance
Problem is, you might not reach past point 3 which is a shame. I am sure that none of you would be posting and saying that this test was flawed if KDE came out on top."
This is total bull. The title of the Phoronix article is misleading and they compared Kubuntu which is full of python scripts, where nepomuk and strigi are running and where Konqueror is preloaded to Ubuntu whith Gnome which is nearly the same as few years ago and is lack of many KDE technologies. Those "out of contest" five points are usually pasted by Icaza or someone and this is idiocy.
Even PTS dev agreed this test is "invalid", because cached memory was measured, so what's your problem?
See Lubos' blog response to the article: http://www.kdedevelopers.org/node/4180
ReplyDelete@Eike Hein
ReplyDeleteYes, nice one. Sadly there's someone who commented and believes Phoronix shows some regressions:
"Their 'strange' benchmarks often spotted regressions in place you could never imagine!"
The best thing I've ever read! Even developers are shocked! Damn...
@Eike Hein, I disagree with this article..
ReplyDeleteThe criticism is that testing filesystems with doing a CPU intensive workload is meaningless, this is incorrect: the filesystem usage use CPU too so in some case this may matter.
OK, this is likely a second order effect where any difference is probably the result of a bug in the filesystem, but this doesn't mean that the test is meaningless..
It would be interesting to compare KDE 3.5.10 with KDE 4.4.1.
ReplyDeleteIn my personal opinion, KDE 3 was faster, slimmer, and less bloated.
It also had less dependencies, causing less headaches (semantic desktop would be the culprit here).
I have a Ubuntu machine, I ran KDE 4.4, I like it. That said, my experience is that the benchmark is indeed correct. KDE 4.4 uses a *LOT-LOT-LOT* more resources than gnome, xfce, lxde and kde3. KDE4 is FAT. But the problem is that it COULD be much better if developers try to optimize and focuse in user (instead of enabling by default things like strigi, neopomuk that most users simply do not use) instead of just adding animations and plasmoid as if those toys were the most amazing thing in the computer land. Guys, those are just toys! Did you not noticed that Windows 98 with the "smart" desktop was a faitule? And it was 12 years ago! Users will use apps, the desktop toys are just to kill time, not the main reason of a DE.
ReplyDelete@Iuri Fiedoruk
ReplyDelete"I have a Ubuntu machine, I ran KDE 4.4, I like it. That said, my experience is that the benchmark is indeed correct. KDE 4.4 uses a *LOT-LOT-LOT* more resources than gnome,"
This is a complete nonsense. The benchmark measured cached memory on Kubuntu and afaik Gnome doesn't cache as much memory as KDE. If you read comments at Phoronix you will notice KDE4 takes about 90MB for some users and probably less then 300MB for any other user! Even PTS contributor agreed the test needs to be re-thinked, so it seems benchmark is not correct. "My KDE" consumes less resources then Ubuntu (so Gnome in Phoronix and yours logic), so my experience is that the benchmark is totally unfair and misleading. Also, KDE is far more advanced, but this is another story. Ubuntu is full of Gnome trash, so don't measure KDE memory usage on this.
"But the problem is that it COULD be much better if developers try to optimize and focuse in user (instead of enabling by default things like strigi, neopomuk that most users simply do not use) instead of just adding animations and plasmoid as if those toys were the most amazing thing in the computer land."
ReplyDeleteThose things are being developed by Gnome devs, because Gnome copies from KDE as usual. However, I agree with this part, but Gnome is starting doing exactly the same now and even worse.
@renoX: Sorry, but testing file system performance with bzip2 compression speed is pretty inane. Hence there's also no appreciable difference between the file systems in the benchmark numbers they're getting. And there's a lot of really dumb filler like that in Phoronix articles.
ReplyDeleteThe underlying problem is that what Phoronix is doing is simply lousy journalism. Anyone can run benchmarks and publish the numbers, that's not journalism. Good journalism is trying to interpret and make sense of the numbers.
For example, raw file system benchmark comparisons are almost always meaningless, because different file systems offer different levels of data integrity and security guarantees. Thus a file system with high data integrity guarantees may appear slow in direct comparison with a file system with low guarantees, but may in fact offer stunning performance in context of those guarantees. Raw benchmark numbers are always uninteresting; benchmark numbers are only ever interesting in context of specific questions and/or requirements.
The problem is that the authors at Phoronix don't seem to have the technological know-how to interpret the numbers they're getting, which makes their articles unfocussed, meaningless, and even actively harmful.
(Regarding desktop environments, the field is very complex as well: Different DEs have different feature sets, so you can't compare e.g. their memory usage 1:1 -- the question is what they use the memory for, and whether they do so appropriately and efficiently. That in turn depends on the user's needs and requirements. Different distribution setups also play a huge role. Auto-starting distro utilities written in Python can severely increase memory footprint due to the CPython overhead, etc.)
First of all, they are making tests on an ALPHA version of an OS, Ubuntu in this case. It is at least "strange".
ReplyDeletethe phoronix numbers say absolutely nothing. kde use much cache if there is free ram and if another application needs ram, it's freed. but if you just want low numbers, here is a screenshot from the german ubuntu forum from kubuntu lucid with 175MiB.
ReplyDeletehttp://media.ubuntuusers.de/forum/attachments/2395451/kde4_low_mem.png
so what does this tell us? the phoronix tests are meaningless!
Again with this stuff "instead of enabling by default things like strigi, neopomuk that most users simply do not use". KDE _Don't_ enable anything by default. Distro's do. And they enable a hell of a lot more than the semantic desktop stuff people like to moan about most theses days.
ReplyDeleteFile indexers are not exactly a new thing, even in most KDE distros pre SC4, beagle was enabled by default by distros, so it was still being done, but it just wasn't as trendy to b*tch in an uninformed way about and wasn't as functional.
And please, point to your reference material that allows you to make the assertion 'most users' don't use a particular technology, I'm really curious to see it.
"And they enable a hell of a lot more than the semantic desktop stuff people like to moan about most theses days."
ReplyDeletePoint to your reference material. KDE uses Strigi not Beagle. Gnome doesn't support about 1/3 of what KDE does, so make comparison fair and don't enable some additional techs. Phoronix measured KDEs cache, so you lost in defending them. I assume it was your intention.
The best part is still that we have a lot of self-proclaimed know-it-alls commenting here and putting Phoronix down. Why doesn't anybody muster up the energy to actually write to them with these views, to defend KDE and to give them a view of how this test could become more fair? And I am surprised to see the mother of all scapegoats: "Kubuntu isn't a decent KDE distro". That might very well be right, but tell me something, what the hell is a KDE distro done right? I have seen internet-hate rain down on Kubuntu, KDE/Fedora and KDE/Suse in different forums, all depending on what direction the wind was blowing from that day.
ReplyDeleteYou want credibility for your views? Take them up with Phoronix. Hell, isn't that what community is all about? Isn't that how we are supposed to improve?
Dabbi, have you tried to open up a dialog with Michael from Phoronix? Not necessarily about this subject, but in general.
ReplyDeleteA while back (a little under two years ago, if I recall), I was having lots of problems with Flash in Konqueror. I mentioned (in the Phoronix forum thread dedicated to advertising concerns) how the Flash ads would often show up as a gray-ed out rectangle with no content shown and no way to interact with them. Michael accused me of stealing from Phoronix.
If you can open up productive dialog with him, congratulations. I've had no such luck.
"KDE uses Strigi not Beagle"
ReplyDeleteLearn to read. I said pre SC4 when discussing beagle as a file indexer used in KDE by distros. Strigi was developed late in the 3.x days and was not used by many distros at all. Download any image from kubuntu, openSuse, fedora, mandriva etc with a pre-4 KDE if you want a reference.
And if you want a reference to other services started, trying looking in your /etc/rc.d/ or distro equivalent.
.... and just in case its not clear, I'm all for the strigi/rdf backend/soprano/nepomuk stack. Trivial comparisons between some gnome-centric distro's package of a gnome DE -vs- their half assed attempt at a KDE DE are a waste of everyone's time IMHO.
ReplyDeleteThese posts have been very interesting. I didn't really spend much time dissecting the article for validity, but I had a feeling that something was up. That said, I do think performance in KDE could benefit from some work.
ReplyDeleteFor the people who are more experienced with these kinds of things, maybe we could get a more proper benchmarking "strategy". Not so we could compare GNOME and KDESC etc. But so that we could compare KDESC4.4 to KDESC4.x and see where we are going and make sure that while we are adding features, we are also being responsible with the system resources because KDESC has a very diverse user base, many of whom may not have the horsepower that some of us enjoy. We can't leave them behind simply because our computers are powerful enough.
"Learn to read. I said pre SC4 when discussing beagle as a file indexer used in KDE by distros. Strigi was developed late in the 3.x days and was not used by many distros at all. Download any image from kubuntu, openSuse, fedora, mandriva etc with a pre-4 KDE if you want a reference."
ReplyDeleteYes, my fault. I thought about KDE4 alpha versions etc.