I am running the 13.10 amd64 version, the package updater is buggy and you often gets "Bla bla bla crashed, click here to report"

Sorry to hear your Dad is having problems but I suppose all auto dependency resolving distros offer the user the only 2 choices of either sticking with the safety of base repos (limiting) or adding wider ones (dangerous). Freedom, with reduction of foolish danger, may be the most compelling reason to use Slackware.

He is not having problems, I am the one having problems xD

I regularly tried out KDE since version 0.9 every year, because WindowMaker got a bit too rural over time. It usually lasted until the first SIGSEGV crash dialog showed up. Also the contents of the .kde directory regularly messed up stuff, especially if they were from an older version. That was not reliable enough for me to depend mission-critical tasks on it.

That was my experience with KDE and it may be result of this:
Another thing of course is the resource utilization. Xfce is snappy on all my systems, from the low-end N270 netbook to the 16 GB Xeon E3. The latter gains some milliseconds, but that's not really noticeable. That's why Xfce4 finally replaced WindowMaker.

KDE is different for two reasons: The higher CPU usage is a result of Qt's signal/slot concept, and that can't be fixed without rewriting everything. Since 4.0, display compositing makes KDE slow on slow GPUs like the GMA 950 or GMA 3150, which I still use. It is not unusable, but it's just not snappy enough and that is important for me.

KDE 4.0 and Windows Vista were both made the same wrong assumption, that hardware will improve faster than the software burns resources. Instead some down-sizing to cheaper, smaller and less powerful systems happened. Additionally high-end CPU development basically stalled.

1 members found this post helpful.

I am wondering if you've read each post of this thread since 3 or 4 people have posted several test results in which CPU usage was a miniscule difference, roughly 4% vs/ 6% at idle, even with full kde library ie: strigi, nepomuk, akonadai, etc. Results would be even closer if those are stripped.

Compositing can be easily turned off if you so desire.

Linux and most desktops are mature enough now that most work very well. There are no gaping differences in performance anymore on even moderately modern hardware. Mostly, all that's left is politics and taste.

There is something wrong, if idle CPU usage goes above 1 % even on an Atom CPU. This will severely reduce battery runtime of any mobile computer. Long story short: 5 % idle CPU usage is completely unacceptable for a desktop environment with actually no applications running!

Then graphics output it's completely unaccelerated and it's up the CPU to draw everything, which is even slower.

But we are talking about less resource usage, that means fewer and simpler GPU drawing calls, which lead to a faster GUI response and less power consumption on a mobile computer.

People have different expectations of GUI performance and stability. Somebody will be just fine with a crawling-slow experience like Windows with two real time anti virus engines installed and maybe somebody will also accept regular application crashes, reload and move on. My requirements are higher. I expect near-instant reaction times ("snappy"). Any UI which can't provide that on every machine I use, it is disqualified.

BTW: That means pure CPU/GPU performance, because every machine in question is SSD-powered. The funny thing is, that with HDD bottleneck removed KDE takes too long to start up, because resource-efficient environments gain a great advantage of not being I/O-bound anymore.

I think you missed the point. That "5%" of application overhead (which is a very small amount) is the applications that make KDE a great desktop experience. Those programs quietly humming away in the background deliver a modern, robust, stable, and fast desktop environment that 'just works' for a lot of people. If you have a specific case where battery life is a greater priority than usability, or your system is old, or ultra-low power, then feel free to choose another DE, or WM. However, I must reiterate that on modern laptop/desktop/server hardware, a 1-3% CPU difference will NOT affect your power consumption or system performance in any noticeable way.

Thank you for your considered response.

I can't speak for others but in my case I rarely use laptops. It's either hardcore desktop or smartphone for me. My test results seem correct for my desktop since I am unconcerned anymore with minimalism. Conky is always running and several widgets (stsin,stdout, weather, time, etc) are always running.Also I'm running Slack 14 32 bit w/ KDE 4.8.5 so in fact I have cpu scaling turned off so I would venture a guess that my system test results are a worst case scenario.

Are you certain of this? On an absolute level Compositing is responsible for orchestrating window repainting, mouse curser etc. but isn't turning off what is listed as "compositing" actually just the eye candy portion? ie - transparency, etc. ?

It seems to me this is simply the tradeoff when dealing with input/output relationships. Analogy - It is possible to restrict gasoline consumption to achieve 100 miles per gallon, but to do so means lower acceleration, top speed, and/or payload capability. We all seek a balance sweet spot that suits us.

In modern PC's, even laptops which employ a number of power managing systems, I don't see this as a big problem if one does a good job of choosing hardware combined with software configuration that suits your workflow. I would imagine most users are concerned more with their productivity, rather than the machine's and get one that fits.

IMHO anyone that accepts a "crawling-slow experience" either can't afford or know how to choose proper hardware nor administer what he has. Crashes? Do those still happen?

I think I get what you actually mean but it seems like you're implying KDE is not resource efficient and I completely disagree with that. I'm quite confident that I can turn off enough services to mimic any DE's usage. KDE is not inefficient anymore than a Mac truck is less efficient than an SUV or a sports car. They just have different capabilities and needs. Don't drive the kids to school in either the semi or the sports car, but don't attempt to haul 30 cubic yards of concrete with a sports car or SUV.

Slackware users commonly drop out of X for certain activities that are resource intensive or provide a simpler environment for deep level work. I see no real obstacle to similarly exchanging DEs on-the-fly if your machine is resource constrained, but when one needs to do the "heavy lifting", it seems silly to not avail oneself of the full potential that KDE offers. KDE simply does more than any other Linux DE, bar none.

Of course it will and you can verify that by running powertop. Other mobile operating systems specifically employ techniques to suspend applications in the background, so the CPU can actually go into deep-sleep (C2 to C6 state).

BTW: Linux still isn't able to compete with Windows and Mac in regard of battery runtime.

I see the issue here.
But everyone else is using notebooks now.

And your smartphone is running KDE, too?

Actually without CPU frequency scaling your numbers looks better than they actually are. CPU usage is calculated relative to the current CPU speed. So if the CPU get's downclocked to 800 MHz and KDE uses 100M cycles of that, it's actually 12,5 %.

That's just a side effect. Xfce4 has a switchable display compositor too, by default it changes nothing about appearance. After turning it on, you can tell for sure if you have a slow GPU.

Power management only works efficiently when CPU and GPU are idle. That means near 0 % load, not 5 %, 10 % or more. Even my high-end desktop nVidia GPU clocks down to 50 MHz, if it is not busy drawing fancy eye-candy. That makes a difference of 60 watts (or $180 per year on the electricity bill for a single PC). So having a energy-efficient desktop actually saves a reasonable amount of money and of course makes the machine less noisy, too.

KDE is rather bloaty, indeed.

I don't. Xfce4 doesn't get in the way and takes less than 100 megs of RAM and almost no CPU cycles for itself. The biggest resource hog I have to deal with on low-end systems is actually the web browser.

1 members found this post helpful.

If you have read the thread you will have noticed that most of the measurements were made by me and that I spoecifically say that on the machines that were used for those measurements background services, like a NFS server and several others, were running. Also, I used htop for the measurements, which naturally produces a little bit of CPU load itself. That the numbers for CPU usage only slightly differ between a plain WM, XFCE and KDE indicates that the "high" CPU usage was caused by those background processes and htop, not by KDE or XFCE.
So, you dislike compositing because it is to slow on your ancient videochip. And you dislike not to have compositing because it does not use your videochip. How could that situation possibly be resolved? Maybe with using a WM that is appropriate for your hardware? If your hardware is to old for Kwin's compositing use KDE with xfwm4, or Openbox, or ... .
What a nonsensical statement. At first, it seems to me that you have no knowledge about properly maintaing a Windows installation, you don't need 2 antivirus engines and programs don't just magically crash every so often. My Windows installation is working fast. For that one reason that I took the time to learn how to properly maintain how to maintain a Windows installation, the same way I have done it for Linux.
But your second statement is even more ridiculous: Go ahead, try Fluxbox on a machine with Pentium I CPU and 32 MB of RAM. If it is not snappy it has been disqualified by your standards.
Really? If nothing else works then it is "it starts too slow"?
Any numbers on that?
It will, they are working on this: http://www.pcworld.com/article/24630...ur_tablet.html
Neither Conky nor htop make show differences in CPU usage dependent on the governor in use. At least not on my machines.
I have never seen a modern Nvidia GPU clocking down to 50MHz, you may restate that with the real numbers.
If a DE with that functionality of KDE is working with that low CPU usage, near to 0%as it seems, and with only 600MB of RAM I wouldn't call that bloated. But may be you show us your expertise and assemble (or write yourself) a DE with the same functions that has better resource usage.

2 members found this post helpful.

I love KDE, now, if you don't, go try something else, why keep bothering us how much you dislike KDE, FFS!

Since much of what we've discussed can be written off to the differences between notebooks and desktops (with a heavy emphasis on battery life) I will respond to this which is what I suspected anyway.

No. It's not. Technically, "Bloat" is strictly defined as "overhead cost that does not contribute to the functioning of an application". IOW, badly written code. This thread has covered how distro devs "jumped the gun" and made KDE4 default when it was basically Alpha level and also didn't bother to pass that little piece of information on to users even as an aside. The finished product is not bloated.

If you don't need a car and a bicycle will do, by all means, save the gas, but don't complain that a 30mpg car carrying 4 in all kinds of weather, in any terrain, with fold-down seats and a trailer hitch is "bloated".


That's curious since 3 tests posted here show more than ~400MB footprint and ~4% CPU at idle with Xfce4. FTR, web browsers are the biggest resource users (not necessarily hogs) on ALL DEs. Are you paring down Xfce?

Again, you don't need to justify your taste preferences by denigrating other "flavors". And also again, I am certain I can match your footprint with KDE even on a notebook by turning off services, so it IS a matter of taste (and right tools for the job) and NOT bloat.

Talking about The Real Resource Hogs...

Look at this little screenshot, from a normal workday of me...

Amazing, right? Yep! This is right: 2.2GB total RAM eaten/allocated/reserved by Firefox and its friend Flash.

PS. You want to make an bet on which one application eat up to 16% from the CPU, too?

PS2. If you are amazed that I work as root, you have to known that operating system used there is a in-house (but Slackware Linux based) distribution, where root is just an ordinary user also/and the only user able to sudo...

Attached Thumbnails

 

Flash is an outrage but still needed, unfortunately

How does that translate to a computer desktop? How would/could you use those "features" to improve the efficiency of the third party applications you use to get whatever it is you do, done?

As to Xfce using 400 megs... That can't be right. It should never use more than 200 megs, if that, including all the panel plug-ins. In another thread someone mentioned excessive RAM usage with Xfce and thinks it is related to a recent update of cairo. He is right. I re-installed the older version and RAM usage dropped considerably, but that isn't the only problem. A recent mesa update (or something else) might also be responsible, but I haven't had a chance to check it out.