KDE 4 Uses 40% Less Memory Than 3 Despite Eye-Candy

An anonymous reader writes "Pro-Linux reports that KDE 4, scheduled to be released in January 2008, consumes almost 40% less memory than KDE 3.5, despite the fact that version 4 of the Free and Open Source desktop system includes a composited window manager and a revamped menu and applet interface. KDE developer Will Stephenson showcased KDE 4's 3D eye-candy on a 256Mb laptop with 1Ghz CPU and run-of-the-mill integrated graphics, pointing out that mini-optimizations haven't even yet been started." Update: 12/14 22:40 GMT by Z : Or, not so much. An anonymous reader writes "The author of the original KDE 3.5 vs KDE 4.0 memory comparison has come out with a more accurate benchmark. In reality, KDE 4.0 uses 110 MB more memory than KDE 3.5.8.

Now if only...

[Anonymous Coward]

...KDE developers had some style. Has anyone looked at the hideous new theme? It looks like a bad Vista rip off. The new panel is freakishly large but is complimented really tiny icons.

Well

[markov_chain]

The laptop was recent, but he limited the memory use and throttled down the CPU to 1GHz. So it still had fancy instructions and a much bigger cache, bus, etc.

Re:Just tried

[tagx]

Lancelot should not even be installed by default. Try removing ~/.kde/share/config/plasma*

Re:Sweet!

[gweihir]

A RC is not non-functioning. It works. As you could have seen from the article.

However it is slower and bigger in the version demonstated, since a lot of debug code is in there.

MS is just looking more and more incompetent all the time.

Re:less memory!

[Anonymous Coward]

The "unused" RAM won't be nice and empty. It'll be used as the system cache to store file data etc. that then can be accessed very quickly. Modern operating systems do not waste RAM by leaving it unused.

Re:Now if only...

[EEPROMS]

It's called a theme. You can change the theme layout to anything you want. I do agree with you though, that is one very ugly theme.

Re:Well

[JonLatane]

1. There's no need for Compiz on KDE4; KWin supports composited window management built-in, and that's what he was using.
2. The computer has Intel integrated graphics, you don't get much lower than that.

Re:To compare with GNOME...

[Elladan]

Keep in mind that even basic modern graphics wastes more memory than that. That background image you have on your 1600x1200 desktop? 5.4 megs. Need a few composite buffers? 5.4 megs each.

Don't have a background? Just the frame buffer to activate that graphics mode itself is 5.4 megs, regardless of what you put on it.

Just to keep things in perspective here. That Commodore 64 you had ran nicely in 64k of ram, but it also only had 320x200 graphics (160x200 in 4-color mode). :-)

Re:misleading article

[dvNull]

I ran OS X server on a 350Mhz PPC with 1.5Gb of RAM. You need at least 256Mb of RAM to run OS X 10.0 - 10.2 usably, 1G or more is ideal. Still the CPU was a bottleneck.
I don't even want to think how Tiger would have run on it. The latest version I ran on it was Panther.

Backwards compatibility?

[filbranden]

Try backwards compatabiltiy and then on top of that inspection and maintaining of it's own integrity.

Strange, considering everything I read about Vista, and my current experiences (problems installing Adobe Reader, impossible to run PDFCreator, some hardware that didn't work well), Vista broke much of backwards compatibility. So as XP broke it too, by not running DOS programs anymore. Therefore, the idea that Vista is bloated because of backwards compatibility sounds strange to me.

On the other hand, I recall reading something about network traffic problems on Vista when copying files, and IIRC it was related to it doing some fiddling on the network stack to make it more difficult to copy media files, that is, DRM related.

I actually tend to believe that more of Vista's bloat is due to DRM than it's due to backwards compatibility, of which it actually has very few.

Re:less memory!

[pherthyl]

Might as well just make another login that boots into twn and a terminal for when you need to do work. I assume you're not working on huge datasets all the time, so then you can run KDE/Gnome most of the time and still have the option of going minimal when you need it.

Re:setup

[pherthyl]

I did in fact use the setup I described... and you can check that imacs were sold with 32 megs on wikipedia. Please check your facts before calling me a lier.

Sorry but you are completely full of shit. OS X does not run for any reasonable definition of "run" on 32mb of RAM.

Have a look at the minimum requirements for OS X 10.1 which you say was the most efficient OS X.

http://en.wikipedia.org/wiki/Mac_OS_X_v10.1#System_Requirements [wikipedia.org]

Notably: RAM required 128 megabytes

And you're saying you did OpenGL development on a quarter of the minimum requirements. Riiight.

Troll. Nice one though. The moderators believed you at least.

Results are completely false

[Anonymous Coward]

Has no one pointed out that the numbers are actually completely, utterly wrong? See Lubos and Thiagos (two high-ranking KDE and Qt devs) comments here:

http://www.kdedevelopers.org/node/3138

See the original authors retraction, here:

http://www.jarzebski.pl/read/kde-3-5-vs-4-0-round-two.so

In similar conditions KDE 3 consumed 97 MB on memory, whereas KDE 4 about 170 MB.

So really, it should be "KDE4 uses 75% more memory", which is actually incredibly lame, but doesn't make for as good a title. I'm absolutely amazed that usually cynical slashdot readers have accepted this so uncritically.

Re:Hmm

[SoapDish]

Well, KDE is listed on the Free Software Directory (directory.fsf.org). Also, just recently, RS was quoted commending KOffice devs, and challenging Gnome devs over their stances on ODF.

Re:Actually...

[bjourne]

When was the last time you used windows? What you have wrote just isn't true. The memory footprint for apps such as Word, Excel and Powerpoint are much lower than comparable Linux apps like OpenOffice, AbiWord and KWrite. Almost all Linux desktop programs takes longer to cold start than their windows (XP or 2000, I've no experience of Vista) equivalents thanks to the huge amount of dynamic libraries Linux uses. EOG is slower than the image viewer in Windows, GEdit is much slower than notepad.exe, write.exe and so on. Internet Explorer and even Firefox starts faster on Windows than Linux.

Re:Here we go again

[Zoolander]

I don't know what Vista uses all its memory for, but KDE using less memory means Linux can use the left over memory as file system cache, as it has done for quite a few years.

Re:To compare with GNOME...

[Fweeky]

Well, I can't decipher your 2nd sentence, vista "turned" what?

I believe the word he's missing is "off". "GDI primitives like LineTo and Rectangle are now rendered in software rather than video hardware [regdeveloper.co.uk]".

Re:Actually...

[FrostedChaos]

Come on. You're going to have to do better than that.

I was running Debian on a 300 MHz Pentium II back in 2003, when the rest of the world was using Windows XP. Performance wasn't an issue. Windows XP wouldn't have even installed on that hardware, much less run. (I did have Windows 2000 creaking along for a while, though, to run some Windows-only apps.)

P.S. Dynamic libraries actually reduce memory consumption because multiple apps can share the same memory.

Re:Results are completely false

[Anonymous Coward]

Aaargh - it get's worse. In the new analysis, he doesn't even include X-server pixmap usage, which Qt4 abuses more than Qt3: in Qt4, all widgets are double-buffered by default, and since the majority of apps are basically windows that are almost 95% covered in widgets, this adds up, fast - a kwrite window, maximised on a 1600x1200, 24-bit screen will gobble up a whopping 6MB almost, just in double-buffering. When you take into account the fact that composite then redundantly double-buffers the entire window *again* (12MB per window, now!), it just gets even worse! So KDE4 is likely using more than twice as much RAM as KDE3, yet the headline reads "KDE4 uses 40% less memory than KDE3" and is tagged "amazing" - what a clusterfuck!

And since people have short-memories, when they do discover that KDE4 takes up hugely more memory than KDE3, they'll remember "KDE developers said it used less, not much more - liars!" rather than "Someone not affiliated with KDE published incorrect benchmarks and we didn't take time to verify them". As if the KDE guys need more abuse hurled at them :/

Re:Actually...

[cp.tar]

When was the last time you used windows? What you have wrote just isn't true. The memory footprint for apps such as Word, Excel and Powerpoint are much lower than comparable Linux apps like OpenOffice, AbiWord and KWrite. Almost all Linux desktop programs takes longer to cold start than their windows (XP or 2000, I've no experience of Vista) equivalents thanks to the huge amount of dynamic libraries Linux uses.

First, you have just switched the issue from the OS to the applications.
That's almost-justified, as users generally care more about their apps than the OS.

Anyway, I won't challenge the fact that MS Office is made well, at least in the features vs. footprint and speed respect.
The UI is a whole new part of discussion, and quite irrelevant here.

Anyway, footprint and startup times are not necessarily equivalent.

EOG is slower than the image viewer in Windows, GEdit is much slower than notepad.exe, write.exe and so on. Internet Explorer and even Firefox starts faster on Windows than Linux.

Don't know about EOG, but Enlightenment's image viewer is about the fastest I've ever seen. I haven't measured its startup time, but I have never seen anything display or resize pictures faster.

Notepad cannot be compared to any other editor, as it is the most useless piece of crap in the editor world.
GEdit has tabs, syntax highlighting, and a whole bunch of other features that Notepad doesn't have.

And yet again: startup times and memory footprint are not the same.

Anyway, the issue here was the OS and its interface; KDE vs. Aero, if you like.
KDE added new features, and so did Aero; KDE has a lower memory footprint than the previous version, while Aero patently doesn't.
On Linux, a compositing UI is available on a much lower-spec'd machine than on Windows.

I have absolutely no idea how their startup times compare, but once up and running, the difference is evident.
I have two 600 MHz machines, one with Linux, the other with WinXP.
Linux is slow, especially if running Gnome, like most people do, but WinXP is a slideshow.
And if you start E17 under Linux, the difference is amazing.

Re:To compare with GNOME...

[EsbenMoseHansen]

I like the top bar and separate window bar at the bottom.

That is very easy to setup yourself in KDE, if you please, Just right-click the panel, add an extra panel and move the stuff you need around. Personally, I think one taskbar wastes too much screen space, so mine is hidden by default. I have yet to find anything to use a 2nd for :)

Re:Wow.

[sqldr]

Funny as it is, the 640k thing is a myth. Asked about the subject, Mr Gates replied "I've said some pretty stupid things in my time, but not that". Sorry to ruin that for you :-(

Most eye Candy doesn't use much memory.

[jellomizer]

It uses CPU Cycles, not Memory for most cases. With faster CPU's expected you can use less memory for more eye candy

Lets take the bouncing Icon. There are two normal ways to program this. Get the icon render each frame for each bounce and save it in memory. And just load the memory and play it. That way it plays smooth and quick every time, because it is in memory all pre-rendered. Now with a faster CPU which spend most of its time idle it can render the icon on the fly between each frame and still keep it smooth so all it needs to do is store the main image the next image to be displayed and perhaps what is currently on the screen. So with a 16x16x8 icon that is around 2k of ram using the CPU method it will only take 6k of ram. vs around 40k of ram for the bouncing icon. But if the CPU couldn't do the work in the time needed to get it done using the memory is the only good option. Memory vs. CPU has always been a balance.

Re:Bad measurements

[Andrei D]

For a more accurate memory usage report, use

pmap -d `pidof $application`

Re:I'm Curious

[virtual_mps]

How does the number of dynamic libraries affect it? Linux running on a desktop is made up of thousands of smaller projects and libraries. Microsoft is able to consolidate these into fewer, larger, libraries. Does that have any advantage? In other words, could Linux benefit from combining lots of the smaller dynamic libraries into more monolithic libraries?

I just looked at random windows & linux sample systems. I found about 700 shared libraries on the linux system and about 2000 dll's on the windows system. I think you're starting from a flawed premise.

Re:Hmm

[chill]

2000 called and they want their FUD back.

Trolltech released QT v2.2 under the GPL back in September 2000 [linuxdevices.com], after which RMS stopped complaining and granted forgiveness [linuxdevices.com] as they did what they wanted.

Re:setup

[podperson]

I think you're being a bit harsh.

He said 10.0 not 10.1. Even so, I think folks are entitled to be a little hazy on machine specs from yesteryear. OS X 10.0 "ran" quite well on 64MB of RAM. (And 333 MHz is actually a pretty liberal estimate. I had 10.2 work quite nicely on an aging G3 Powerbook.)

http://en.wikipedia.org/wiki/Mac_OS_X_v10.0 [wikipedia.org]

Re:Unbloating?

[mr_mischief]

I grew up in Hannibal, Missouri. We studied Mark Twain for 13 years in public school there, for obvious reasons. I'm not familiar with Mark Twain as the source of this quote, although he may have repeated it. I've always heard it attributed to Blaise Pascal, but it seems he may have been paraphrasing someone earlier [stanford.edu]. Pascal lived well before Sam Clemens.

It possibly dates back to St. Augustine or even Cicero, but the most common wording of the idea in English is a straightforward translation from Pascal's.

Re:Wow.

[coolGuyZak]

Since our colleagues insist on being pricks, I found this on the Security Now [grc.com] site:

Leo: Well, ironically, they've made it more susceptible to malware. These tilt bits - talk a little bit about the tilt bits, Peter.
PETER: Right. So what tilt bits are is - the name's taken from pinball machines. We had tilt sensors to monitor physical interference with the device.

Leo: Yeah, if you pick up the machine and get the ball in the hole, it's tilted, and it fails.
PETER: Right. And so Microsoft have done or required that hardware manufacturers do pretty much exactly the same thing. The nasty thing with this is that, well, to put it bluntly, it makes your hardware in your system a lot less reliable. The typical PC is thrown together out of all sorts of random bits and pieces with different tolerances; and half the parts are made by the cheapest possible manufacturer, so a lot of them are cheap and nasty. So they're designed to have a certain amount of tolerance for voltage fluctuations and strange bus signals and bugs in device drivers that set hardware bits wrong and so on and so forth. The problem is that, if you do get these strange voltage fluctuations or strange noise on the system bus or whatever, that could also be a sign of attack. And so Microsoft have said that hardware has to monitor for any of these peculiarities. And if they're found, then it sets these tilt bits in a register somewhere. Vista polls these tilt bits; and if any of them are set, it reacts in some vaguely specified but somewhat drastic manner.