[gentoo-user] bloated vs. lightweight WM's
Lloyd H. Meinholz
have not found that the lightweight WM's that I have worked with have
improved my life that much. Out of curiosity, I decided to measure
resource usage on my workstation with my workload in as meaningful way
as I could (without spending too much time on it). If my measurements
could be easily improved, please let me know. I've compared gnome 2.6,
kde 3.2.2, xfce4 4.0.5, openbox 3.2 and iceWM 1.2.13.
In my measurements, I don't consider build times since that happens
infrequently (though gnome and KDE are definitely bloated in this
regard) or disk usage (again gnome, kde probably much bigger) or startx
times (since I usually have my desktop up a minimum of one week between
reboots). I'm only looking at memory and process numbers with top, which
isn't 100% accurate, but close enough.
I have a P4 2.3 GHz, with 1GB memory. I use this as a development
workstation and have lots of server type stuff running (apache, tomcat,
openldap, mysql, postfix). After I start X, I usually have a 3 big
applications (evolution, firefox, idea (my Java IDE)) and lots of
xterms.
conosle (no X) - 179 MB, 127 procs
gnome 2.6 - 522 MB, 232 procs
kde 3.2.2 - 513 MB, 215 procs
xfce4 4.0.5 - 515 MB, 209 procs
openbox 3.2 - 438 MB, 197 procs
iceWM 1.2.13 - 441 MB, 202 procs
All of these WM's (except gnome, I've added a couple of applets so the
memory is probably higher than out of the box) use the default
configuration and are running just my desktop apps.
It seems xfce4 isn't all that lightweight considering it's resource
usage. openbox and iceWM seem to be somewhat smaller (and more
primitive).
It would be interesting to know how much resources XFree takes to be
able to compare only the WM's.
This is a real world (my world anyway) usage comparison, but I'm not
sure how technically accurate it is (top accuracy, valid use case,
etc.). I am disappointed by the lightweight WM's performance especially
considering how much people seem to be gnome/kde/bloatware bashing.
>From a design and developers point of view, I agree that bloated
software is bad and irritating, but from a practical point of view, in
this specific case, I don't see a dramatic improvement from using the
lightweight WM's.
Lloyd
--
gento...@gentoo.org mailing list
James Larkby-Lahet
am not familiar with all the Wms you listed, it has been my experience
that blackbox will run fine on boxes where gnome+kde will drag things
to a crawl.
Thanks,
James
--
gento...@gentoo.org mailing list
dav...@servidor.unam.mx
>
> conosle (no X) - 179 MB, 127 procs
> gnome 2.6 - 522 MB, 232 procs
> kde 3.2.2 - 513 MB, 215 procs
> xfce4 4.0.5 - 515 MB, 209 procs
> openbox 3.2 - 438 MB, 197 procs
> iceWM 1.2.13 - 441 MB, 202 procs
>
I would have measured X alone too, maybe with a single "xterm" running
-------------------------------------------------
www.correo.unam.mx
UNAMonos Comunicándonos
--
gento...@gentoo.org mailing list
Ciaran McCreesh
<lloyd.m...@bmpcoe.org> wrote:
| I've been reading a lot of stuff about bloated and lightweight WM's
| and have not found that the lightweight WM's that I have worked with
| have improved my life that much. Out of curiosity, I decided to
| measure resource usage on my workstation with my workload in as
| meaningful way as I could (without spending too much time on it). If
| my measurements could be easily improved, please let me know. I've
| compared gnome 2.6, kde 3.2.2, xfce4 4.0.5, openbox 3.2 and iceWM
| 1.2.13.
| I have a P4 2.3 GHz, with 1GB memory.
Do the same benchmarks on a box no faster than, say, 200MHz, with a
fairly basic graphics card and only 512MBytes RAM. Now, you may not
run one of those boxes for every day usage, but do you really want to
sacrifice so much of your super-expensive faster CPU on doing trivial
window management?
--
Ciaran McCreesh : Gentoo Developer (Sparc, MIPS, Vim, Fluxbox)
Mail : ciaranm at gentoo.org
Web : http://dev.gentoo.org/~ciaranm
Lloyd H. Meinholz
are what my typical work environment is, and since I have the exact same
apps open each time, I thought that was fair enough.
Lloyd
--
gento...@gentoo.org mailing list
Lloyd H. Meinholz
> fairly basic graphics card and only 512MBytes RAM. Now, you may not
> run one of those boxes for every day usage, but do you really want to
> sacrifice so much of your super-expensive faster CPU on doing trivial
> window management?
CPU usage seems to be minimal in all WM's, so I don't see that as an
issue. Besides the quality of moving around windows in gnome seems to be
more attractive and smoother than all the other WM's I just tested and I
am willing to expend an extra .01% cpu usage for that. Memory is the
real issue and my suspicion is that my base system and XFree itself is
using so much memory that the WM memory usage is trivial, thus
irrelevant. So the whole lightweight argument from a performance
perspective seems to me to be incorrect, at least in this my case.
I would however agree that the lightweight WM argument is probably a
good one from a design perspective, but that is different from
performance.
My home machine is a P3 733 MHz with 256 MB of ram. I have been a long
time gnome user and my user experience has been becoming unpleasant on
this machine for a long time. Over the past year I've been reading a lot
about lightweight WM's, so I tried openbox, then xfce4. Neither of them
gave me much better performance on this machine. Maybe the machine is
too slow for anything to help, but openbox and xfce4 certainly didn't
help.
Looking at my memory numbers (256 MB on my home box and a minimum 441 MB
for iceWM) I probably was using swap no matter which WM I was using and
I am going to be slow no matter what...
I don't know that I have the ambition to configure my home machine to
test all this anytime soon (I've already spent more time on this than I
meant to), but maybe I will bring it in to the office and try it...
James Larkby-Lahet
what is the room for improvement due to wm choice. As David
suggested,,, you could simply start an xterm, instead of a window
manager, to approximate this.
--
gento...@gentoo.org mailing list
dav...@servidor.unam.mx
> How would I measure X alone? I want the applications running since they
> are what my typical work environment is, and since I have the exact
> same
> apps open each time, I thought that was fair enough.
>
> Lloyd
>
I think
X :1
should work. It will open a server in display :1
Also keep in mind that after you run, say, gnome, all of its libraries remain
in memory, so for a safe bet you should reboot between each test.
and try to do it running the least apps possible and then with your apps (as it
is very likely that your apps use or not shared libraries that were already
loaded by the wm)
just my 2 cents.
and to be safer, make sure you are not running a dm either (xdm/kdm/gdm/etc)
Lloyd H. Meinholz
> Any chance you could get some numbers for fluxbox or blackbox? While I
> am not familiar with all the Wms you listed, it has been my experience
> that blackbox will run fine on boxes where gnome+kde will drag things
> to a crawl.
fluxbox 0.9.9 - 454 MB, 202 procs
blackbox 0.65.0-r3 - 451 MB, 202 procs
Seem to be a little higher than openbox, but not much.
I actually didn't have the experience with blackbox that you had
(openbox in my case). Maybe my requirements were so high and memory so
low that it didn't matter what I used... (see other post).
Ciaran McCreesh
<lloyd.m...@bmpcoe.org> wrote:
| > Do the same benchmarks on a box no faster than, say, 200MHz, with a
| > fairly basic graphics card and only 512MBytes RAM. Now, you may not
| > run one of those boxes for every day usage, but do you really want
| > to sacrifice so much of your super-expensive faster CPU on doing
| > trivial window management?
|
| CPU usage seems to be minimal in all WM's, so I don't see that as an
| issue.
You are very very very wrong here.
| Looking at my memory numbers (256 MB on my home box and a minimum 441
| MB for iceWM) I probably was using swap no matter which WM I was using
| and I am going to be slow no matter what...
You must be doing something wrong. I've got a diskless Sun JavaStation
JK with 64MBytes of RAM that runs xorg-x11 and fluxbox quite happily.
Lloyd H. Meinholz
> I think
> X :1
> should work. It will open a server in display :1
Maybe I'm a bonehead but this didn't work. I get a black screen and X,
but left and rick clicking the mouse don't bring up any menu to start an
xterm. If I comment out everything in rc.conf, I get a really primitive
WM that is twm or fwm or something? Would this be enough?
> Also keep in mind that after you run, say, gnome, all of its libraries remain
> in memory, so for a safe bet you should reboot between each test.
I should have mentioned, I am rebooting between each test, out of deep,
deep paranoia..
> and to be safer, make sure you are not running a dm either (xdm/kdm/gdm/etc)
I'm not running a DM.
dav...@servidor.unam.mx
> On Mon, 2004-05-24 at 13:53, dav...@servidor.unam.mx wrote:
> > I think
> > X :1
> > should work. It will open a server in display :1
>
> Maybe I'm a bonehead but this didn't work. I get a black screen and X,
> but left and rick clicking the mouse don't bring up any menu to start
> an
> xterm. If I comment out everything in rc.conf, I get a really primitive
> WM that is twm or fwm or something? Would this be enough?
>
Yes. thats right. you are running only X and nothing else.
you can go to another term and check stats there.
you can also do this instead
X :1 & DISPLAY=:1 xterm &
Simon Roby
> I think
> X :1
> should work. It will open a server in display :1
Actually it's preferable to run `startx -- :1`.
Speaking of which, it's a little known fact that X11 can run in multiple
instances quite well. Hell, you can even start X11 from another X11 in a
terminal emulator. It will use the next available vc slot, and you can switch
back in forth with the usual ctrl-alt-f* combo. Try it, it's fun, and can be
used as a crude way of doing that "fast user switching" feature that's so
popular in recent proprietary desktop OSes.
--
gento...@gentoo.org mailing list
Ciaran McCreesh
wrote:
| Speaking of which, it's a little known fact that X11 can run in
| multiple instances quite well. Hell, you can even start X11 from
| another X11 in a terminal emulator. It will use the next available vc
| slot, and you can switch back in forth with the usual ctrl-alt-f*
| combo. Try it, it's fun, and can be used as a crude way of doing that
| "fast user switching" feature that's so popular in recent proprietary
| desktop OSes.
And don't forget Xnest. It lets you run an X server inside a window on
another X server. Note that most window managers get a bit picky with
focus issues when you do that...
Chris Woods
Hash: SHA1
Simon Roby wrote:
| Speaking of which, it's a little known fact that X11 can run in multiple
| instances quite well. Hell, you can even start X11 from another X11 in a
| terminal emulator. It will use the next available vc slot, and you can switch
| back in forth with the usual ctrl-alt-f* combo. Try it, it's fun, and can be
| used as a crude way of doing that "fast user switching" feature that's so
| popular in recent proprietary desktop OSes.
This is actually a menu choice in gnome these days. There are two choices:
"New Login" (starts X on another VC in the manner you describe, while starting
up xscreensaver on the original one) and "New Login in a Nested Window", which
starts xnest, so one can have a second X session in a window inside an
already-running X session.
- --
~ Chris Woods || the....@comcast.net
~ ICQ 21740987 || AIM gnarrlybob
~ YIM cjwoods || Jabber Gnar...@jabber.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAskKvcv1XM0B+gpsRAuIaAJ4wO868r3inF0JbWYjBGmwxpK9mkgCePLzd
HwpIvRcykCOZN1sLCiWi/ko=
=KjaI
-----END PGP SIGNATURE-----
--
gento...@gentoo.org mailing list
Lloyd H. Meinholz
> On Mon, 24 May 2004 13:48:52 -0400 "Lloyd H. Meinholz"
> <lloyd.m...@bmpcoe.org> wrote:
> | > Do the same benchmarks on a box no faster than, say, 200MHz, with a
> | > fairly basic graphics card and only 512MBytes RAM. Now, you may not
> | > run one of those boxes for every day usage, but do you really want
> | > to sacrifice so much of your super-expensive faster CPU on doing
> | > trivial window management?
> |
> | CPU usage seems to be minimal in all WM's, so I don't see that as an
> | issue.
>
> You are very very very wrong here.
I'm looking at top, I see nothing that is X related eating up CPU time.
What makes you say that gnome (for example) is using up CPU time? I can
find nothing to support your statement. I could imagine gnome using more
CPU time since the window decorations are more detailed, but from
looking at top, it seems pretty miniscule.
> | Looking at my memory numbers (256 MB on my home box and a minimum 441
> | MB for iceWM) I probably was using swap no matter which WM I was using
> | and I am going to be slow no matter what...
>
> You must be doing something wrong. I've got a diskless Sun JavaStation
> JK with 64MBytes of RAM that runs xorg-x11 and fluxbox quite happily.
Again, what does this have to do with CPU usage? Memory seems to be a
much bigger issue than CPU usage.
What else are you running on your JavaStation? My development
environment without X uses 179 MB. I use 3 pretty heavy X applications
(evolution, firefox and idea). This takes a lot of resources.
Run apache, tomcat, mysql, openldap, postgres, courier-imap, evolution,
firefox and idea in your 64 MB's. These are my requirements for my
development workstation. I am trying to compare my options based on my
requirements. My requirements may not be the same as yours, so while a
64 MB JavaStation my fulfill your requirements, they most certainly do
not fulfill mine, even if I were running fluxbox.
You're comparing apples to oranges. I'm trying to compare my machine
implementing my requirements using different window managers. My
experience with my requirements show that the choice of WM doesn't make
a whole lot of difference.
What am I doing wrong? I think you misunderstand my requirements and
what I am trying evaluate. If I am wrong, please tell me what to change
in my testing to make it correct?
Eamon Caddigan
> On May 24, 2004 1353, dav...@servidor.unam.mx wrote:
>> I think
>> X :1
>> should work. It will open a server in display :1
>
> Actually it's preferable to run `startx -- :1`.
Following David's instructions starts X alone, without touching
~/.xinitrc or /etc/X11/xinit/xinitrc. I'm pretty sure that was
intentional.
> Speaking of which, it's a little known fact that X11 can run in multiple
> instances quite well. Hell, you can even start X11 from another X11 in a
> terminal emulator. It will use the next available vc slot, and you can switch
> back in forth with the usual ctrl-alt-f* combo. Try it, it's fun, and can be
> used as a crude way of doing that "fast user switching" feature that's so
> popular in recent proprietary desktop OSes.
My family has been using multiple X sessions to simulate "fast user
switching" for quite a while, but it leaves much to be desired. I'm
actually planning to start a "X fast user switching" project in the next
couple weeks[1]. Anyone interested in stealing my thunder and going
ahead with this is welcome to get in touch with me; I believe I've
figured out about half of the issues that need to be addressed to make
this work smoothly. I'm thinking that the freedesktop.org people /might/
be interested in this once some actual code is written.
-Eamon
[1] My 2004 Year's resolution was that I would only begin new projects
when I've finished old ones. It's frustrating, but I have too many
half-finished projects.
--
gento...@gentoo.org mailing list
Chris Woods
Hash: SHA1
Ciaran McCreesh wrote:
|
| You must be doing something wrong. I've got a diskless Sun JavaStation
| JK with 64MBytes of RAM that runs xorg-x11 and fluxbox quite happily.
Color me impressed. So X + fluxbox is running completely in memory - in 64MB?
And this isn't running on another machine and simply being displayed via X to
the JavaStation?
- --
~ Chris Woods || the....@comcast.net
~ ICQ 21740987 || AIM gnarrlybob
~ YIM cjwoods || Jabber Gnar...@jabber.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAskNWcv1XM0B+gpsRAio3AKCNBUYKSqlC9O4btpugJ/12BIjNJgCfYuk6
jwchTvmmglwyep+5gN0jeIc=
=i+S/
Lloyd H. Meinholz
> Any chance you could get some numbers for fluxbox or blackbox? While I
> am not familiar with all the Wms you listed, it has been my experience
> that blackbox will run fine on boxes where gnome+kde will drag things
> to a crawl.
fluxbox 0.9.9 - 454 MB, 202 procs
blackbox 0.65.0-r3 - 451 MB, 202 procs
Seem to be a little higher than openbox, but not much.
I actually didn't have the experience with blackbox that you had
(openbox in my case). Maybe my requirements were so high and memory so
low that it didn't matter what I used... (see other post).
Lloyd
--
gento...@gentoo.org mailing list
Eamon Caddigan
>
> Simon Roby wrote:
>| Speaking of which, it's a little known fact that X11 can run in multiple
>| instances quite well. Hell, you can even start X11 from another X11 in a
>| terminal emulator. It will use the next available vc slot, and you can switch
>| back in forth with the usual ctrl-alt-f* combo. Try it, it's fun, and can be
>| used as a crude way of doing that "fast user switching" feature that's so
>| popular in recent proprietary desktop OSes.
>
> This is actually a menu choice in gnome these days. There are two choices:
> "New Login" (starts X on another VC in the manner you describe, while starting
> up xscreensaver on the original one) and "New Login in a Nested Window", which
> starts xnest, so one can have a second X session in a window inside an
> already-running X session.
Awesome... I've heard that KDE can do the same. Out of curiosity, what
happens to GDM when you start a new X session? Does it have its own
display, doling out new displays for each subsequent login?
As I mentioned elsewhere, I'm interested in doing this in a
desktop-environment agnostic way. I think it would be a real boon for
Linux on the desktop.
-Eamon
--
gento...@gentoo.org mailing list
Qureshi, Affan
But it doesnt make sense that memory usage does not decrease significantly even when all the (helper) apps KDE and Gnome run are absent.
Maybe its more of a subjective issue. For some people saving the (rather small) amount of resources we saw may prove significant while it wont matter for the rest.
Just-watching-this-thread-from-the-side-lines...
Affan
Lloyd
--
gento...@gentoo.org mailing list
--
gento...@gentoo.org mailing list
Chris Woods
Hash: SHA1
Eamon Caddigan wrote:
|
| Awesome... I've heard that KDE can do the same. Out of curiosity, what
| happens to GDM when you start a new X session? Does it have its own
| display, doling out new displays for each subsequent login?
It would appear as though a completely new X session starts up in either case,
along with a "special" instance of gdm that has a "Quit" option along with the
usual "Session" and "System" menus. In the case of the session starting up on
the alternate VC, it switches over and I see my nvidia splash screen - so X is
starting fresh there - and then the gdm Login screen. In the case of starting
up inside xnest, I don't see the nvidia splash screen - just the gdm Login
screen with the extra "Quit" option.
- --
~ Chris Woods || the....@comcast.net
~ ICQ 21740987 || AIM gnarrlybob
~ YIM cjwoods || Jabber Gnar...@jabber.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAskdqcv1XM0B+gpsRAiAqAKCFnn3U7BxH6ntyFEzxmVZwNe/fQwCffRZ4
c9s7O9C7aA6CxoY84bqo94g=
=5G4G
Ciaran McCreesh
wrote:
| | You must be doing something wrong. I've got a diskless Sun
| | JavaStation JK with 64MBytes of RAM that runs xorg-x11 and fluxbox
| | quite happily.
|
| Color me impressed. So X + fluxbox is running completely in memory -
| in 64MB? And this isn't running on another machine and simply being
| displayed via X to the JavaStation?
Yup. It's doable. I can even run a few instances of urxvt on it. Of
course, firefox and friends are out of the question locally. X11
forwarding comes in handy there.
For anyone interested in the JavaStation JK, it's a totally silent Sun
box that you can get fairly cheaply on eBay. They make rather nice thin
clients. If you stop by #gentoo-sparc on freenode I can tell you how to
get them up and running. Photos at:
http://dev.gentoo.org/~ciaranm/boxen/javadrop/
Simon Roby
> I've been reading a lot of stuff about bloated and lightweight WM's and
> have not found that the lightweight WM's that I have worked with have
> improved my life that much. Out of curiosity, I decided to measure
> resource usage on my workstation with my workload in as meaningful way
> as I could (without spending too much time on it). If my measurements
> could be easily improved, please let me know. I've compared gnome 2.6,
> kde 3.2.2, xfce4 4.0.5, openbox 3.2 and iceWM 1.2.13.
For the 8234092th time, KDE and Gnome are *not* WINDOW MANAGERS, they are
DESKTOP ENVIRONMENTS. When you load their main processes (startkde or
gnome-session, respectively), it doesn't load up a window manager, but rather
a _session_ manager, which loads a bunch of stuff, the window manager (kwin
or metacity, respectively) being only one of those apps. The rest are
support/utility daemons and applets that provide extra functionality to the
desktop or apps.
> In my measurements, I don't consider build times since that happens
> infrequently (though gnome and KDE are definitely bloated in this
> regard) or disk usage (again gnome, kde probably much bigger) or startx
> times (since I usually have my desktop up a minimum of one week between
> reboots). I'm only looking at memory and process numbers with top, which
> isn't 100% accurate, but close enough.
Gnome and KDE are DEs, so they are a _collection_of_applications_. Why would
that be called "bloat"?
> I have a P4 2.3 GHz, with 1GB memory. I use this as a development
> workstation and have lots of server type stuff running (apache, tomcat,
> openldap, mysql, postfix). After I start X, I usually have a 3 big
> applications (evolution, firefox, idea (my Java IDE)) and lots of
> xterms.
Running evolution and firefox under KDE is pretty much missing the point of
it. *Of course* the memory usage is going to be high, you'll have two toolkit
libraries running (not to mention gnome support libs and daemons), and the
precached stuff and services KDE loads at startup are going to end up unused.
> conosle (no X) - 179 MB, 127 procs
> gnome 2.6 - 522 MB, 232 procs
> kde 3.2.2 - 513 MB, 215 procs
> xfce4 4.0.5 - 515 MB, 209 procs
> openbox 3..2 - 438 MB, 197 procs
> iceWM 1.2.13 - 441 MB, 202 procs
>
> All of these WM's (except gnome, I've added a couple of applets so the
> memory is probably higher than out of the box) use the default
> configuration and are running just my desktop apps.
Again, you compared WMs with DEs. Apples and oranges.
> It seems xfce4 isn't all that lightweight considering it's resource
> usage. openbox and iceWM seem to be somewhat smaller (and more
> primitive).
>
> It would be interesting to know how much resources XFree takes to be
> able to compare only the WM's.
>
> This is a real world (my world anyway) usage comparison, but I'm not
> sure how technically accurate it is (top accuracy, valid use case,
> etc.). I am disappointed by the lightweight WM's performance especially
> considering how much people seem to be gnome/kde/bloatware bashing.
But there are also many other people (like me) who are perfectly happy with
those. You just don't notice us, because we don't constantly bitch about
stuff ;)
> From a design and developers point of view, I agree that bloated
> software is bad and irritating, but from a practical point of view, in
> this specific case, I don't see a dramatic improvement from using the
> lightweight WM's.
Just what does "bloat" mean anyway? Merriam-Webster Online doesn't even have a
definition for it in a computers/software context.
--
gento...@gentoo.org mailing list
Todd de Gruyl
> Just what does "bloat" mean anyway? Merriam-Webster Online doesn't even have a
> definition for it in a computers/software context.
not to get too deep into the actual issue here, you can probably get a
decent idea here:
<http://www.catb.org/~esr/jargon/html/B/bloatware.html>
--
Todd de Gruyl
http://www.tdegruyl.com
--
gento...@gentoo.org mailing list
Brenden Walker
> From: Todd de Gruyl [mailto:to...@tdegruyl.com]
> Sent: Monday, May 24, 2004 3:36 PM
> To: gento...@lists.gentoo.org
> Subject: Re: [gentoo-user] bloated vs. lightweight WM's
>
>
>
> > Just what does "bloat" mean anyway? Merriam-Webster Online doesn't
> > even have a
> > definition for it in a computers/software context.
>
> not to get too deep into the actual issue here, you can
> probably get a
> decent idea here:
> <http://www.catb.org/> ~esr/jargon/html/B/bloatware.html>
I think this is getting a bit twisted around due to incorrect terminology
and the like. What we're really talking about here (I think) is Desktop
Environment's vs. simply running a window manager.
If what you want and/or need is a complete desktop enviroment, then do so.
Personally I have no need of all the extra's.
Lloyd H. Meinholz
> For the 8234092th time, KDE and Gnome are *not* WINDOW MANAGERS, they are
> DESKTOP ENVIRONMENTS. When you load their main processes (startkde or
> gnome-session, respectively), it doesn't load up a window manager, but rather
> a _session_ manager, which loads a bunch of stuff, the window manager (kwin
> or metacity, respectively) being only one of those apps. The rest are
> support/utility daemons and applets that provide extra functionality to the
> desktop or apps.
Then what is something like openbox without gnome? Does that mean I am
running a window manager without a desktop environment? Or does it mean
openbox is functioning as a desktop environment in this context but as a
window manage when used in conjunction with gnome? Sorry I got the
terminology wrong...
> Gnome and KDE are DEs, so they are a _collection_of_applications_. Why would
> that be called "bloat"?
That wasn't my term. It is a term many others have been using. I am
actually trying to prove the point that (at least in the way my desktop
environment is set up) the window manager is really more a matter of
taste than of performance or resource usage. I have read many posts to
this list stating that lightweight WM's/DE's whatever perform better
than bloated (not my word choice) kde or gnome.
> Running evolution and firefox under KDE is pretty much missing the point of
> it. *Of course* the memory usage is going to be high, you'll have two toolkit
> libraries running (not to mention gnome support libs and daemons), and the
> precached stuff and services KDE loads at startup are going to end up unused.
Well it is my point and I don't think I'm missing it. I'm describing the
way I use my development workstation and trying to evaluate the
performance of and resource usage of each DESKTOP ENVIRONMENT.
>
> > conosle (no X) - 179 MB, 127 procs
> > gnome 2.6 - 522 MB, 232 procs
> > kde 3.2.2 - 513 MB, 215 procs
> > xfce4 4.0.5 - 515 MB, 209 procs
> > openbox 3..2 - 438 MB, 197 procs
> > iceWM 1.2.13 - 441 MB, 202 procs
> >
> > All of these WM's (except gnome, I've added a couple of applets so the
> > memory is probably higher than out of the box) use the default
> > configuration and are running just my desktop apps.
>
> Again, you compared WMs with DEs. Apples and oranges.
If I am using blackbox without gnome, is blackbox a WM or DE or both?
Lloyd H. Meinholz
> Quoting "Lloyd H. Meinholz" <lloyd.m...@bmpcoe.org>:
> > conosle (no X) - 179 MB, 127 procs
> > gnome 2.6 - 522 MB, 232 procs
> > kde 3.2.2 - 513 MB, 215 procs
> > xfce4 4.0.5 - 515 MB, 209 procs
> > openbox 3.2 - 438 MB, 197 procs
> > iceWM 1.2.13 - 441 MB, 202 procs
> >
>
> I would have measured X alone too, maybe with a single "xterm" running
Similar to your suggestion, I tried it with just X running, and with X +
evolution, firefox and idea. All tests are with server apps running
(apache, tomcat, openldap, mysql, postfix). Here's my new results:
Without evolution, firefox and idea running:
conosle (no X) - 179 MB, 127 procs
X only - 198 MB, 129 procs
With evolution, firefox and idea running:
DE/WM TOTAL MEM/PROC USAGE DE/WM MEM/PROC USAGE
================== ==================== ====================
X - 335 MB, 158 procs
gnome 2.6 - 522 MB, 232 procs 187 MB, 74 procs
kde 3.2.2 - 513 MB, 215 procs 178 MB, 57 procs
xfce4 4.0.5 - 515 MB, 209 procs 180 MB, 51 procs
openbox 3.2 - 438 MB, 197 procs 103 MB, 39 procs
fluxbox 0.9.9 - 454 MB, 202 procs 119 MB, 44 procs
blackbox 0.65.0-r3 - 451 MB, 202 procs 116 MB, 44 procs
iceWM 1.2.13 - 441 MB, 202 procs 106 MB, 44 procs
Again, the base for my tests:
* My workstation is a P4 2.3 GHz, with 1GB memory.
* All tests were performed with the following servers running: apache,
tomcat, openldap, mysql, postfix.
* 3 big X applications running: evolution, firefox, idea.
The reason I did the test the way I did was to use a real world scenario
I was familiar with (my own). I have been hearing for a long time about
using a WM performing better than big DE's and that just wasn't my
experience (at least in the environment I am working in) and I wanted to
find out what the difference actually was.
Lloyd H. Meinholz
> I think this is getting a bit twisted around due to incorrect terminology
> and the like. What we're really talking about here (I think) is Desktop
> Environment's vs. simply running a window manager.
>
> If what you want and/or need is a complete desktop enviroment, then do so.
> Personally I have no need of all the extra's.
You are probably right and as Simon pointed out, I've misused the terms
DE and WM and this is probably a source of some confusion. I'm sorry for
this because I was simply trying to do something good by comparing and
documenting performance of different WM/DE combinations in my
environment.
Many of the arguments I've heard in favor of lightweight WM choices have
mentioned performance and resource usage as an advantage. This wasn't
something I observed in my environment and I was just trying to figure
out what was the reality.
It is obviously true that if a WM only approach works for you, then
there is no reason to install an entire DE, but that choice has nothing
to do with performance or runtime resource usage (except compile time
and disk usage of course :) ).
Sean
> and the like. What we're really talking about here (I think) is Desktop
> Environment's vs. simply running a window manager.
>
> If what you want and/or need is a complete desktop enviroment, then do so.
> Personally I have no need of all the extra's.
I agree completely. I've gone through both KDE and Gnome phases, but
after noticing that I really didn't use a lot of the "desktop
environment" features, I dropped back down to a straight window manager,
and have been quite content.
--
GPG Key => http://tinyurl.com/2yud3
Simon Roby
> On Mon, 2004-05-24 at 15:26, Simon Roby wrote:
> > For the 8234092th time, KDE and Gnome are *not* WINDOW MANAGERS, they are
> > DESKTOP ENVIRONMENTS. When you load their main processes (startkde or
> > gnome-session, respectively), it doesn't load up a window manager, but
> > rather a _session_ manager, which loads a bunch of stuff, the window
> > manager (kwin or metacity, respectively) being only one of those apps.
> > The rest are support/utility daemons and applets that provide extra
> > functionality to the desktop or apps.
> Then what is something like openbox without gnome?
A window manager.
> Does that mean I am
> running a window manager without a desktop environment?
Yes.
> Or does it mean
> openbox is functioning as a desktop environment in this context but as a
> window manage when used in conjunction with gnome? Sorry I got the
> terminology wrong...
Regardless of whether you use it with Gnome or not, openbox remains a window
manager. It manages windows. See my original post for a description of a
desktop environment.
> > Gnome and KDE are DEs, so they are a _collection_of_applications_. Why
> > would that be called "bloat"?
>
> That wasn't my term. It is a term many others have been using. I am
> actually trying to prove the point that (at least in the way my desktop
> environment is set up) the window manager is really more a matter of
> taste than of performance or resource usage. I have read many posts to
> this list stating that lightweight WM's/DE's whatever perform better
> than bloated (not my word choice) kde or gnome.
Sorry, I wasn't really targetting you when I was bitching about the use of the
word "bloat", just the people you were referring to.
> > Running evolution and firefox under KDE is pretty much missing the point
> > of it. *Of course* the memory usage is going to be high, you'll have two
> > toolkit libraries running (not to mention gnome support libs and
> > daemons), and the precached stuff and services KDE loads at startup are
> > going to end up unused.
>
> Well it is my point and I don't think I'm missing it. I'm describing the
> way I use my development workstation and trying to evaluate the
> performance of and resource usage of each DESKTOP ENVIRONMENT.
>
> > > conosle (no X) - 179 MB, 127 procs
> > > gnome 2.6 - 522 MB, 232 procs
> > > kde 3.2.2 - 513 MB, 215 procs
> > > xfce4 4.0.5 - 515 MB, 209 procs
> > > openbox 3..2 - 438 MB, 197 procs
> > > iceWM 1.2.13 - 441 MB, 202 procs
> > >
> > > All of these WM's (except gnome, I've added a couple of applets so the
> > > memory is probably higher than out of the box) use the default
> > > configuration and are running just my desktop apps.
> >
> > Again, you compared WMs with DEs. Apples and oranges.
>
> If I am using blackbox without gnome, is blackbox a WM or DE or both?
See above, but s/openbox/blackbox .
Anyway I'm off this thread, I don't like discussing about so-called "bloat"
issues.
--
gento...@gentoo.org mailing list
Eric Paynter
I don't know what you're measuring, but I've had xfce4 running on a system
with less than 29 MB in use - total for the entire system. You're not
including cache/buffers, are you?
-Eric
p.s. sorry for breaking threading - I missed the original post :\
--
arctic bears - affordable email and name services @yourdomain.com
http://www.arcticbears.com
--
gento...@gentoo.org mailing list
Ajai Khattri
> conosle (no X) - 179 MB, 127 procs
> gnome 2.6 - 522 MB, 232 procs
> kde 3.2.2 - 513 MB, 215 procs
> xfce4 4.0.5 - 515 MB, 209 procs
> openbox 3.2 - 438 MB, 197 procs
> iceWM 1.2.13 - 441 MB, 202 procs
Is this the total usage for all GUI-related apps?
I have 228Mb in my laptop and I have no problem running xfce4 without
swapping...
--
Aj.
Sys. Admin / Developer
--
gento...@gentoo.org mailing list
Eric Paynter
> On Mon, 24 May 2004, Lloyd H. Meinholz wrote:
>
>> conosle (no X) - 179 MB, 127 procs
>> gnome 2.6 - 522 MB, 232 procs
>> kde 3.2.2 - 513 MB, 215 procs
>> xfce4 4.0.5 - 515 MB, 209 procs
>> openbox 3.2 - 438 MB, 197 procs
>> iceWM 1.2.13 - 441 MB, 202 procs
>
> Is this the total usage for all GUI-related apps?
>
> I have 228Mb in my laptop and I have no problem running xfce4 without
> swapping...
It seems high for anything. I have xfce4, mozilla (with over 20 tabs
open), blender, a couple dozen aterms, apache and mysql running on my dev
desktop right now and it is sitting around 215MB used. As I said earlier,
I've had xfce4 with the panel and task bar but no apps running at <29MB
*total system usage*. xfce4 is great if you are low on memory - it will
not require swapping and it is very light on CPU. It has enough
-Eric
Lloyd H. Meinholz
how accurate it is for any one elses use case. I just meant it as an
example of my particular situation.
Lloyd
--
gento...@gentoo.org mailing list
Lloyd H. Meinholz
and gui apps). This was a rough estimate arrived at by subtracting what X
running all apps used from what xfce4 running all apps used. I am running
two big java applications, tomcat and idea, which both take lots of memory
and processes. That may explain the big numbers.
Lloyd
--
gento...@gentoo.org mailing list
Peter Gordon
--
Peter Gordon, Student
Bachelor of Computer Science & Engineering
Cal. State Univ. Fullerton, Class of 2008
E-Mail: pg15...@student.fullerton.edu
GnuPG Public Key: 0xE721C290
Chris Woods
Hash: SHA1
Peter Gordon wrote:
| Huh? Is it possible to run XFce-4 without running X?
I think he was just doing the math to estimate how much memory xfce4 itself
was taking -- subtracted the total resource usage of X and other stuff to come
to a more "accurate" number.
I don't think he was actually running xfce4 without X :)
- --
~ Chris Woods || the....@comcast.net
~ ICQ 21740987 || AIM gnarrlybob
~ YIM cjwoods || Jabber Gnar...@jabber.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAsqSbcv1XM0B+gpsRArE4AJoCvr8MJwFY/zYQ2HtTNBQHkxDXbwCeIMrk
Wb9e/GT0YXg0oWSs/Qxqt/k=
=yFQT
Peter Gordon
Chris Woods wrote:
> Peter Gordon wrote:
> | Huh? Is it possible to run XFce-4 without running X?
>
> I think he was just doing the math to estimate how much memory xfce4 itself
> was taking -- subtracted the total resource usage of X and other stuff
> to come
> to a more "accurate" number.
>
> I don't think he was actually running xfce4 without X :)
>
> --
>
> ~ Chris Woods || the....@comcast.net
> ~ ICQ 21740987 || AIM gnarrlybob
> ~ YIM cjwoods || Jabber Gnar...@jabber.org
>
--
Lloyd H. Meinholz
terrible mistake.
What I mean by "minus X" is my estimate at the resource xfce4 is using
(total memory with xfce4 running minus total memory with only X running).
Lloyd
--
gento...@gentoo.org mailing list
Chris
> Sorry I was confusing. I'm beginning to think this whole topic has been a
> terrible mistake.
It wasn't a mistake in my book, I find this very interesting. :)
--
Chris P. Carter
Nothing happens in contradiction to nature, only to what we know of nature.
Registered GNU/Linux user #349582
Gentoo Linux : KDE 3.2.1 : Kernel 2.4.26-pre5-gentoo
--
gento...@gentoo.org mailing list
Tero G
would matter on a machine with 1GB of ram. With that memory you can do
allmost anything without memory being an issue. But if you have only
=<512MB of memory a lighter wm could make a big difference in performance.
You are right about the CPU usage - it doesn't make a big difference
in a performance of a WM or DE unless your WM uses alot of
transparency. Also some apps can be unusually slow with text scrolling.
Many people use CFLAGS like -O3 or -funroll-loops etc. for their desktop
software which simply does not make any sence because desktop apps don't
benefit from faster execution time. That kind of aggressive flags only make
their desktop slower because of larger code size which fill up the caches
and cause more disk and memory i/o. In other words the real bottle necks
become more tighter.
More than anything else hard disk performance and the amount of memory are
the key factors in desktop performance IMHO.
I find it weird that many people with >PIII >~500GHz and >256Mb in their
machines complain how slow their desktops are.
I have a Pentium MMX 233@262Mhz with 128Mb of ram and I'm quite happy with
the performance... OOo is a bit slow because its so big - but once it is
loaded the performance is good enough for that app too. Java is usable now
thanks to NPTL.
I could upgrade my machine any time but I refuse to do so. I like
tweaking and find it more satisfactory to tweak this old box with the
power of Gentoo Linux. It is fascinating that, while playing an mp3 song,
the CPU usage is now 0%! On Win98 it was about 40%...
My current wm is IceWM and I mostly use GTK1 apps - I don't even have GTK2
installed on my system. GTK2 is not that good any way. I also don't
use any apps with gnome/kde dependencies. Some of the apps I use are
Firefox, Gimp, Gentoo and Sylpheed. Btw AA fonts suck on linux at the
moment so I avoid them at all costs.
It's been forgotten in this discussion that with gnome/kde applications
you also get their dependencies as a "bonus". So with these apps, no
matter how simple your WM, the resource usage can be allmost as if you
were using the full DE.
Tero G
On Mon, 24 May 2004, Lloyd H. Meinholz wrote:
> I've been reading a lot of stuff about bloated and lightweight WM's and
> have not found that the lightweight WM's that I have worked with have
> improved my life that much. Out of curiosity, I decided to measure
> resource usage on my workstation with my workload in as meaningful way
> as I could (without spending too much time on it). If my measurements
> could be easily improved, please let me know. I've compared gnome 2.6,
> kde 3.2.2, xfce4 4.0.5, openbox 3.2 and iceWM 1.2.13.
>
> In my measurements, I don't consider build times since that happens
> infrequently (though gnome and KDE are definitely bloated in this
> regard) or disk usage (again gnome, kde probably much bigger) or startx
> times (since I usually have my desktop up a minimum of one week between
> reboots). I'm only looking at memory and process numbers with top, which
> isn't 100% accurate, but close enough.
>
> I have a P4 2.3 GHz, with 1GB memory. I use this as a development
> workstation and have lots of server type stuff running (apache, tomcat,
> openldap, mysql, postfix). After I start X, I usually have a 3 big
> applications (evolution, firefox, idea (my Java IDE)) and lots of
> xterms.
>
> conosle (no X) - 179 MB, 127 procs
> gnome 2.6 - 522 MB, 232 procs
> kde 3.2.2 - 513 MB, 215 procs
> xfce4 4.0.5 - 515 MB, 209 procs
> openbox 3.2 - 438 MB, 197 procs
> iceWM 1.2.13 - 441 MB, 202 procs
>
> All of these WM's (except gnome, I've added a couple of applets so the
> memory is probably higher than out of the box) use the default
> configuration and are running just my desktop apps.
>
> It seems xfce4 isn't all that lightweight considering it's resource
> usage. openbox and iceWM seem to be somewhat smaller (and more
> primitive).
>
> It would be interesting to know how much resources XFree takes to be
> able to compare only the WM's.
>
> This is a real world (my world anyway) usage comparison, but I'm not
> sure how technically accurate it is (top accuracy, valid use case,
> etc.). I am disappointed by the lightweight WM's performance especially
> considering how much people seem to be gnome/kde/bloatware bashing.
>
> >From a design and developers point of view, I agree that bloated
> software is bad and irritating, but from a practical point of view, in
> this specific case, I don't see a dramatic improvement from using the
> lightweight WM's.
>
> Lloyd
Sami Samhuri
> I have a P4 2.3 GHz, with 1GB memory. I use this as a development
> workstation and have lots of server type stuff running (apache, tomcat,
> openldap, mysql, postfix). After I start X, I usually have a 3 big
> applications (evolution, firefox, idea (my Java IDE)) and lots of
> xterms.
>
> conosle (no X) - 179 MB, 127 procs
I suppose I have a similar system:
Athlon XP 2800+
512MB memory
I too run apache2 (mod_ssl, mod_php), mysql, postfix, cups,
courier-imap, ftpd, and then some.
When I run free after booting I have about 50M used! After starting X
with xfce4 or ratpoison that jumps up by a hundred MB or so, at least. X
is a hog, no doot aboot it. Your initial amount of memory being used
seems excessive though. Is my system just running much less than yours,
is there something _huge_ running on yours? It seems... just too much.
--
Sami Samhuri
Lloyd H. Meinholz
> I suppose I have a similar system:
>
> Athlon XP 2800+
> 512MB memory
>
> I too run apache2 (mod_ssl, mod_php), mysql, postfix, cups,
> courier-imap, ftpd, and then some.
>
> When I run free after booting I have about 50M used! After starting X
> with xfce4 or ratpoison that jumps up by a hundred MB or so, at least. X
> is a hog, no doot aboot it. Your initial amount of memory being used
> seems excessive though. Is my system just running much less than yours,
> is there something _huge_ running on yours? It seems... just too much.
It seems to be tomcat. When I don't start tomcat, I only use around
60MB. Tomcat only uses 63MB from top, but is multi-threaded (starts out
with around 19 threads). I'm not sure how 60 + 63 = 179, but those are
the numbers top gives me (I've done this several times). I've heard that
top isn't necessarily 100% accurate and multi-threaded apps are more
difficult... I can't wait until I have the time to reinstall my system
with 2.6 and nptl.
Lloyd H. Meinholz
> > It seems to be tomcat. When I don't start tomcat, I only use around
> > 60MB. Tomcat only uses 63MB from top, but is multi-threaded (starts
> out
> > with around 19 threads). I'm not sure how 60 + 63 = 179, but those are
> > the numbers top gives me (I've done this several times). I've heard
> that
> > top isn't necessarily 100% accurate and multi-threaded apps are more
> > difficult... I can't wait until I have the time to reinstall my system
> > with 2.6 and nptl.
>
> Well all I can say is I bet you are sure glad you have 1G of RAM with
> the amount your work requires. :)
yup, that's why I was interested in the WM only approach in the first
place and was disappointed when it didn't help me much. When I first
started investigating WM only approach, I only had 512MB and the problem
was more urgent.
Simon Roby
> I think it is rather silly to think that the memory usage of a WM or DE
> would matter on a machine with 1GB of ram. With that memory you can do
> allmost anything without memory being an issue. But if you have only
> =<512MB of memory a lighter wm could make a big difference in performance.
Come on now, 512MB, or even 256MB is still much more than most people actually
need. A different WM would only make a difference under 128MB (IMHO).
> Many people use CFLAGS like -O3 or -funroll-loops etc. for their desktop
> software which simply does not make any sence because desktop apps don't
> benefit from faster execution time. That kind of aggressive flags only make
> their desktop slower because of larger code size which fill up the caches
> and cause more disk and memory i/o. In other words the real bottle necks
> become more tighter.
O3 on x86 is overrated. Its heavy optimisations enlarges code significantly,
which may be more efficient in the short term, but in the end a bigger
program _will_ hurt overall performance for sure, especially if the whole
system is compiled that way.
> More than anything else hard disk performance and the amount of memory are
> the key factors in desktop performance IMHO.
Definitely.
> I find it weird that many people with >PIII >~500GHz and >256Mb in their
> machines complain how slow their desktops are.
Indeed. I run KDE on a Celeron 550MHz with 192MB or RAM and it runs fine and
dandy. And on my Athlon XP 2500+ with 512MB of RAM, the extra relative
performance gains are not really that significant, despite being almost 4
years more technologically advanced.
> I could upgrade my machine any time but I refuse to do so. I like
> tweaking and find it more satisfactory to tweak this old box with the
> power of Gentoo Linux. It is fascinating that, while playing an mp3 song,
> the CPU usage is now 0%! On Win98 it was about 40%...
That must've been a pretty shitty player you were using under Windows. Winamp
doesn't burn that much.
> My current wm is IceWM and I mostly use GTK1 apps - I don't even have GTK2
> installed on my system. GTK2 is not that good any way. I also don't
> use any apps with gnome/kde dependencies. Some of the apps I use are
> Firefox, Gimp, Gentoo and Sylpheed. Btw AA fonts suck on linux at the
> moment so I avoid them at all costs.
AA fonts under Linux are highly configurable. The way I tweaked them, my fonts
render almost like in Mac OS X on my Linux boxen. And properly rendered AA
fonts are much easier to the eyes. Nowadays I get disgusted whenever I see
non-AA fonts (usually when I reboot to WIndows), they are so much less
readable than in my current setup.
--
gento...@gentoo.org mailing list
Chris Woods
Hash: SHA1
Simon Roby wrote:
| On May 24, 2004 2240, Tero G wrote:
|>I could upgrade my machine any time but I refuse to do so. I like
|>tweaking and find it more satisfactory to tweak this old box with the
|>power of Gentoo Linux. It is fascinating that, while playing an mp3 song,
|>the CPU usage is now 0%! On Win98 it was about 40%...
|
|
| That must've been a pretty shitty player you were using under Windows. Winamp
| doesn't burn that much.
Actually, on windows, and especially on win9x, the number of entries in the
playlist directly affects the performance of winamp. If I loaded a playlist of
more than a couple hundred titles, winamp would *crawl*, and my system would
lag for a few seconds between songs. It was significantly better on win2k (and
by proxy winxp), but the problem was still there.
Just a little anecdotal input on one of the issues mentioned :)
- --
~ Chris Woods || the....@comcast.net
~ ICQ 21740987 || AIM gnarrlybob
~ YIM cjwoods || Jabber Gnar...@jabber.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAs8jJcv1XM0B+gpsRAqSDAJ44NGbyeCu5utBW1Kzc16i8xtIXyACfS+vU
HSbXgRk5L86rXfFuH7jmCZM=
=2GIT
Sami Samhuri
> I think it is rather silly to think that the memory usage of a WM or DE
> would matter on a machine with 1GB of ram. With that memory you can do
> allmost anything without memory being an issue. But if you have only
> =<512MB of memory a lighter wm could make a big difference in performance.
clearly using more memory than any of us (on our desktops). In
addition, The fact that X+<your fav DE> can easily take up a few hundred
MB of that memory, well that's 1/3 of your gig right there.
--
Sami Samhuri
Chris Woods
Hash: SHA1
Sami Samhuri wrote:
|
| I don't think it's ever silly to worry about memory usage. The OP is
| clearly using more memory than any of us (on our desktops). In
| addition, The fact that X+<your fav DE> can easily take up a few hundred
| MB of that memory, well that's 1/3 of your gig right there.
I've outgrown worrying about how much memory my desktop apps use, and am more
concerned with obtaining the features I would like to have in my environment.
If I see a feature that I would like to use, and I don't have enough memory to
use it, well... I go get more memory. The benefit gained by using gnome in all
its glory (same philosophy applies to KDE) far outweighs the cost in system
resources -- for *me*.
I understand that this is not a common mentality, and that's ok. Diff'rent
strokes :)
A lesson hard-learned over many years: hardware is a LOT less expensive than
people time. However, if somebody is at least as productive in an environment
that uses less hardware resources, great. It's not a serious concern for me,
though. It's not hard to justify dropping another $200 to get a reasonable
amount of memory in a desktop machine.
When I'm writing software for someone else's use, I will naturally design it
so that it runs in the most resource-efficient manner possible, but it's not
the only consideration. In fact, it is very often not even near the top of the
priority list for my clients, at least not on the level of memory and CPU and
disk space. They're typically more concerned with ease of use, learning curve
(or lack thereof), and productivity.
I've got what I consider to be a pretty beefy desktop machine, and it cost me
$800 a year or so ago. This machine paid for itself within 2 months - easily.
I use gnome with lots of what people would consider "eyecandy", but it's what
I'm most comfortable using, and makes my work (and play) experience much more
tolerable. It's hard to quantify the gain there, but it's worth it to me.
I don't presume to dictate just what sort of environment is "acceptable" for
anyone else but myself.
- --
~ Chris Woods || the....@comcast.net
~ ICQ 21740987 || AIM gnarrlybob
~ YIM cjwoods || Jabber Gnar...@jabber.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAs/7Hcv1XM0B+gpsRAsLlAJ91rtZ4ELqKll40ly0VIaoLZoRG5ACeLamC
rajRYPzV2xC3otZNRbGszsk=
=AUcK
Ajai Khattri
> It is for all server and GUI apps. This is just my use case. I'm not sure
> how accurate it is for any one elses use case. I just meant it as an
> example of my particular situation.
I think the main difference between "just running a window manager" and
running a full environment is the number and size of the processes
involved. I have recently started using ratpoison - when I look at what's
running I see X, xfs and ratpoison AND NOTHING ELSE. When you run
something like GNOME, you are running X + xfs + window manager + various
taskbars and applets + backend stuff like oafd, orbit, etc etc just for
the basic environment (and that's before you've launched any apps).
This is why some of us prefer the minimal approach.
Tero G
<snip>
> Come on now, 512MB, or even 256MB is still much more than most people actually
> need. A different WM would only make a difference under 128MB (IMHO).
I'd think 256MB is *barely* enough for Gnome or KDE. But if you also want
to run applications too... well I don't know actually..
<snip>
> O3 on x86 is overrated. Its heavy optimisations enlarges code significantly,
> which may be more efficient in the short term, but in the end a bigger
> program _will_ hurt overall performance for sure, especially if the whole
> system is compiled that way.
I couldn't emphasize this enough...
<snip>
> > I could upgrade my machine any time but I refuse to do so. I like
> > tweaking and find it more satisfactory to tweak this old box with the
> > power of Gentoo Linux. It is fascinating that, while playing an mp3 song,
> > the CPU usage is now 0%! On Win98 it was about 40%...
>
> That must've been a pretty shitty player you were using under Windows. Winamp
> doesn't burn that much.
Well the player was Winamp.. And I only got to 0% on Linux after I
upgraded to 2.6 kernel (love-sources) and changed the internal kernel
frequency to 100Hz. Before that it was about 9%. And with 2.4 kernel about
3-4%..
I recommend for everybody who is running 2.6 kernel on an older box to get
this patch and change the kernel frequency to 100Hz.
<snip>
> AA fonts under Linux are highly configurable. The way I tweaked them, my fonts
> render almost like in Mac OS X on my Linux boxen. And properly rendered AA
> fonts are much easier to the eyes. Nowadays I get disgusted whenever I see
> non-AA fonts (usually when I reboot to WIndows), they are so much less
> readable than in my current setup.
Yeah I know they're very tweakable but for some reason I'm happy with
non-AA truetype fonts "out of the box". With AA fonts I allways keep
thinking whether my fonts are good or not.
--
gento...@gentoo.org mailing list
Tero G
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Simon Roby wrote:
>
> | On May 24, 2004 2240, Tero G wrote:
> |>I could upgrade my machine any time but I refuse to do so. I like
> |>tweaking and find it more satisfactory to tweak this old box with the
> |>power of Gentoo Linux. It is fascinating that, while playing an mp3 song,
> |>the CPU usage is now 0%! On Win98 it was about 40%...
> |
> |
> | That must've been a pretty shitty player you were using under Windows. Winamp
> | doesn't burn that much.
>
> Actually, on windows, and especially on win9x, the number of entries in the
> playlist directly affects the performance of winamp. If I loaded a playlist of
> more than a couple hundred titles, winamp would *crawl*, and my system would
> lag for a few seconds between songs. It was significantly better on win2k (and
> by proxy winxp), but the problem was still there.
>
> Just a little anecdotal input on one of the issues mentioned :)
That number (40%) was without any playlist.
--
gento...@gentoo.org mailing list
Tero G
With desktop software it's the same like with decorating - people like
different things.
I wouldn't fill my house with useless stuff no matter how much money I
had. If I had the best computer money can buy, I probaply still wouldn't
use gnome, kde or other "gadgets". I simply like simplicity.
It is great that in Linux we _still_ have the possibility for choice.
However I'm a little afraid of the future in that regard.
On Tue, 25 May 2004, Chris Woods wrote:
--
gento...@gentoo.org mailing list
Tero G
Hmm... I'm not sure if the OP measured the memory usage correctly. AFAIK
the more you have memory the more they *seem* to be using according to top or
similar programs.
I'm under the impression that it is more complex to measure the true
memory usage than that. But I'm not competent enough to explain it...
All I know is that I've used xfce4 with 64mb of memory and the memory
meter showed that only half of it was used after a fresh reboot... I had
to open several Mozilla windows to get the system to swap.
--
gento...@gentoo.org mailing list
Simon Roby
"X+<your fav DE>" won't take up "a few hundred MB of that memory". That's a
gross exageration. If I close all windows except KMail and Konsole on my KDE
desktop right now, I get a (-/+ buffers/cache) system memory usage of 118MB.
--
gento...@gentoo.org mailing list
Simon Roby
> I'd think 256MB is *barely* enough for Gnome or KDE. But if you also want
> to run applications too... well I don't know actually..
You're just making this up. See my previous post.
--
gento...@gentoo.org mailing list
Andy Herrman
Right now I have X running using the kahakai window manager. I have
thunderbird, firefox, gaim, and a few other apps running, along with a
few wm docapps. I'm at about 180MB total being used. Without
firefox/thunderbird the number would probably be significantly less.
Personally, I chose to use kahakai for performance reasons. However, I
don't mean the performance of the machine, I mean overall performance
when using it. I was able to set it up to do exactly what I wanted, and
I am much more efficient when using it than other window managers.
However, this is completely due to my personal style of using my
machine. I couldn't get KDE or Gnome to do a few things I wanted, so I
don't use it. If it does what you want, go for it. I doubt there's
much of a difference as far as resources go. If you like lots of helper
apps running all the time it won't matter which window manager you use,
you'll still end up using up more resources than someone who doesn't use
those apps.
Just my $.02
-Andy
Simon Roby wrote:
--
gento...@gentoo.org mailing list
Chris Woods
Hash: SHA1
Simon Roby wrote:
| On May 26, 2004 0543, Tero G wrote:
|
|>I'd think 256MB is *barely* enough for Gnome or KDE. But if you also want
|>to run applications too... well I don't know actually..
|
|
| You're just making this up. See my previous post.
I think there is a lot more at play here than any of us is seeing. My previous
machine was a 700MHz Athlon with 384MB, and gnome was sluggish. KDE was next
to unuseable. Granted, on that machine, I wasn't running gentoo - I was
running a more-or-less homegrown distribution (started with SuSE barebones,
then downloaded sources and built everything by hand) - but I'm pretty sure I
got optimizations down pretty nicely there too. Prior to my getting into
gentoo a few months ago, I only ever really bootstrapped my own linux boxes
and built everything by hand. That process got to be too time-consuming, hence
my discovery of gentoo.
One thing I've learned is that with linux, if the memory is available, it will
be used to near capacity, I believe due to the filesystem caching mechanism in
linux. I've got 1GB of RAM here, and "top" shows 950MB of that in use. I'm
running several apps, but still -- 950MB in use? A better indicator of how
much memory is really being used by the system is to look at how much swap is
engaged. If you show lots of swap used, then perhaps it is time to either get
more RAM or slim down your app usage.
- --
~ Chris Woods || the....@comcast.net
~ ICQ 21740987 || AIM gnarrlybob
~ YIM cjwoods || Jabber Gnar...@jabber.org
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAtLNCcv1XM0B+gpsRAje8AJ96ulr4+8USJJD+hgyPtYOn+wx2cQCfXCnI
UtGDu5iOhO7FhmYI/vU656M=
=t4/c
al...@halogen-dg.com
Hi
On Tue, 25 May 2004, Simon Roby wrote:
> On May 24, 2004 2240, Tero G wrote:
> > I think it is rather silly to think that the memory usage of a WM or DE
> > would matter on a machine with 1GB of ram. With that memory you can do
> > allmost anything without memory being an issue. But if you have only
> > =<512MB of memory a lighter wm could make a big difference in performance.
>
> Come on now, 512MB, or even 256MB is still much more than most people actually
> need. A different WM would only make a difference under 128MB (IMHO).
>
I have 512RAM but pretty slow old server (Celeron 466). I *hate* when
something is slow, while both my CPU and HDD are very slow.
KDE and Gnome were working very badly here. So when I switched to IceWM
I found a WM of my life :) It loads in 2 seconds (compare to 1 minute
of KDE startup). It eats RAM less then single 'xterm' does. It is very
fast in switching between windows / desktops (it was noticeable under
KDE to switch between windows, especially in Editor<->Browser chain, when
developing web software).
Now, even on very fast computer at home, I am still running IceWM.
> Indeed. I run KDE on a Celeron 550MHz with 192MB or RAM and it runs fine
> and dandy.
Depends on tasks you run there. For me, on Celeron 466 it was very
noticiable when I switched desktops and windows in KDE. May be its
my vide card, but perfomance was very bad with KDE.
Also, IceWM does have very good and easy configuration, menu, taskbar,
themes, docking and many things which you did not find in KDE or Gnome.
The best features I like in IceWM are:
1. Very easy to configure menu - you can edit it in editor, and change
appears immediately. It is *much* more comfortable to edit it this
way, comparing to KDE/Gnome menu editors.
2. Taskbar gets configured in seconds...
3. Windows have Alt F1 - Alt F10 commands, and its very easy to use
those shortcuts. UNder KDE I have to redefine some of them manually.
4. There are more things you can do with windows in IceWM then in KDE.
I like "hide" feature for example.
5. IceWM is extremely fast. It used to work on my 32RAM computer, and
even there load time and usability were very good.
So, in fact, when choosing the right WM you must ask you which tasks
are most common for you.
In my own setup I am using:
- a number of terminals (terminal start time is also very important
to me. using kopete or gnome-terminal on my computer is too slow)
- IM messaging
- docking (for xmms, email, IM, meeting)
- mail program
- Browser
Almost any WM will work for me, but I found KDE very slow, Gnome slow
and buggy, and IceWM the best :)
--
Alex V. Koval
http://www.halogen-dg.com/
http://www.zwarehouse.org/
--
gento...@gentoo.org mailing list
Tero G
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Simon Roby wrote:
> | On May 26, 2004 0543, Tero G wrote:
> |
> |>I'd think 256MB is *barely* enough for Gnome or KDE. But if you also want
> |>to run applications too... well I don't know actually..
> |
> |
> | You're just making this up. See my previous post.
As I said I don't actually _know_. In fact I have more experience on
Windows XP than Gnome or KDE because some of my friends use it. But I
_think_ that WinXP is pretty much comparable to Gnome and especially KDE
when it comes to resource usage. So thats how I have come to the
conclusion that 256MB would not be enough to comfortable use Gnome or KDE.
But I could be wrong.
--
gento...@gentoo.org mailing list
Simon Roby
> I think there is a lot more at play here than any of us is seeing. My
> previous machine was a 700MHz Athlon with 384MB, and gnome was sluggish.
> KDE was next to unuseable.
I take it that was back before KDE 3.2 ? KDE has made gigantic leaps and
bounds in terms of efficiency in the past two major updates. Even KDE 3.1.x,
being a lot better than before, was a pain on my old laptop, but now with the
3.2 series it just flows. Believe me, it's really awesome.
> One thing I've learned is that with linux, if the memory is available, it
> will be used to near capacity, I believe due to the filesystem caching
> mechanism in linux. I've got 1GB of RAM here, and "top" shows 950MB of that
> in use. I'm running several apps, but still -- 950MB in use? A better
> indicator of how much memory is really being used by the system is to look
> at how much swap is engaged. If you show lots of swap used, then perhaps it
> is time to either get more RAM or slim down your app usage.
That's because top includes disk buffers and cache in the "memory use" value.
Unless you recently booted, or you recently shut down a memory-heavy
application, it's perfectly normal for memory to be exausted, since the
kernel uses as much buffers and cache as it can. Whenever more memory is
needed, the kernel starts flushing disk buffers and clearing unused cache.
This is the normal behavior of any recent operating system, only you usually
don't notice it because those other operating systems tend to have better
memory reporting facilities (and it's a well known fact that Linux truly
sucks in terms of memory reporting).
To know how much total physical memory your applications are _really_ using,
run `free -m`, and look at the "-/+ buffers/cache" line.
--
gento...@gentoo.org mailing list
pur...@sympatico.ca
> KDE has made gigantic leaps and bounds in terms of efficiency
> in the past two major updates. Even KDE 3.1.x,
> being a lot better than before, was a pain on my old laptop,
that's certainly my own experience, tho' i use XFCE as desktop manager.
i use several KDE apps & there's no problem with speed.
> top includes disk buffers and cache in the "memory use" value.
> Unless you recently booted or shut down a memory-heavy application,
> it's perfectly normal for memory to be exausted,
> since the kernel uses as much buffers and cache as it can.
you are correct: this seems to be a common mistake by newer users.
i did a short set of tests on my own machine,
which has an Athlon XP2500+ processor with 512 MB memory.
each test follows a reboot; i used 'free' & 'top' to get the figures;
no other applications were running except Xscreensaver for XFCE/KDE
& a Kterm or Xterm to get at the figures;
the memory figure is from Free's 2nd line, ie w/o buffers/cache;
i boot into a raw terminal, then start/stop X by hand (no Xdm etc).
i can't test Gnome, as i don't have the whole thing installed.
immediately after boot : memory 21 MB ; tasks 28 .
after starting XFCE 4.0.5 : memory 39 MB ; tasks 40 ; start-up 9 s .
after starting KDE 3.2.2 : memory 51 MB ; tasks 49 ; start-up 20 s .
w XFCE after starting 2 terminals, Gkrellm1, Kdeinit & Mutt : 51 MB ;
ditto + Galeon : 64 MB .
the actual net saving in memory with XFCE is 12 MB , ie 18 vs 30 ;
also, there is not much difference in number of tasks or start-up times.
however, there mb other performance considerations
when you start to handle multiple apps & frequently switch desktops.
XFCE has always been very efficient there, while KDE used to be rougher:
i don't know how far KDE has improved in that area.
does anyone want to give comparable figures for other machines/WMs/DEs ?
--
========================,,============================================
SUPPORT ___________//___, Philip Webb : pur...@chass.utoronto.ca
ELECTRIC /] [] [] [] [] []| Centre for Urban & Community Studies
TRANSIT `-O----------O---' University of Toronto
--
gento...@gentoo.org mailing list