PAE

David Bolt wrote:
> On Thursday 20 Aug 2009 19:55, Van Chocstraw played with alphabet spaghetti
> and left this on the plate:
>
>> I have 2.225 gig of ram. OpenSUSE 11.1 loads the PAE kernel by default.
>> Then I added Virtualbox and it added the KMP PAE kernel.
>> I don't want PAE running and want it to only use 2.0 gig of ram.
>
> Why would you have the extra 256MiB of RAM and not want to use it?
>
>> I loaded the default kernel but the system won't boot completely.
>> Complains of part 5 and part 6 of the harddrive and drops to the $.
>
> Sounds like it's not got the correct modules in the initrd. Try using the
> rescue system to redo the boot loader. As a part of this process, a new
> initrd is created. An alternative is to create the initrd manually.
>
>> How do I turn off PAE?
>
> Unless you're using real-time programs, you won't notice the slowdown
> because of the PAE kernel. So, unless there's another compelling reason,
> there's no real point to removing the PAE kernel.
>
> However, if that's what you're wanting to do, the easiest way is to use YaST
> to install the default kernel with the same version number.
>
>
> Regards,
> David Bolt
>
I did use yast to install the same version default kernel. It's in the
grub menu but has the above problem.

Speed is all I had on my mind in changing. I just thought Opensuse would
run faster without it, if not, I guess I'll forget it.
Comparatively, on the same system, W2Kpro and WindowsME run considerably
faster.

On Thu, 2009-08-20 at 16:07 -0400, Van Chocstraw wrote:
> David Bolt wrote:
> > On Thursday 20 Aug 2009 19:55, Van Chocstraw played with alphabet spaghetti
> > and left this on the plate:
> >
> >> I have 2.225 gig of ram. OpenSUSE 11.1 loads the PAE kernel by default.
> >> Then I added Virtualbox and it added the KMP PAE kernel.
> >> I don't want PAE running and want it to only use 2.0 gig of ram.

PAE really is only interesting past 4G anyhow on 32bit machines.

> >
> > Why would you have the extra 256MiB of RAM and not want to use it?
> >
> >> I loaded the default kernel but the system won't boot completely.
> >> Complains of part 5 and part 6 of the harddrive and drops to the $.
> >
> > Sounds like it's not got the correct modules in the initrd. Try using the
> > rescue system to redo the boot loader. As a part of this process, a new
> > initrd is created. An alternative is to create the initrd manually.
> >
> >> How do I turn off PAE?
> >
> > Unless you're using real-time programs, you won't notice the slowdown
> > because of the PAE kernel. So, unless there's another compelling reason,
> > there's no real point to removing the PAE kernel.

Memory performance penalty is less than 5%. Usually unoticeable.

> >
> > However, if that's what you're wanting to do, the easiest way is to use YaST
> > to install the default kernel with the same version number.
> >
> >
> > Regards,
> > David Bolt
> >
> I did use yast to install the same version default kernel. It's in the
> grub menu but has the above problem.
>
> Speed is all I had on my mind in changing. I just thought Opensuse would
> run faster without it, if not, I guess I'll forget it.
> Comparatively, on the same system, W2Kpro and WindowsME run considerably
> faster.

Well.. in the general case, openSUSE will be faster.

This message is not archived

houghi wrote:

> Chris Cox wrote:
>>> Speed is all I had on my mind in changing. I just thought Opensuse
>>> would run faster without it, if not, I guess I'll forget it.
>>> Comparatively, on the same system, W2Kpro and WindowsME run
>>> considerably faster.
>>
>> Well.. in the general case, openSUSE will be faster.
>
> And then muc`h will depend on the application one is running. And if
> speed is the objective, openSUSE is a strange choice. I would go for
> something like DSL. At least something that does not run KDE or GNOME as
> default.
>

The choice of default DE has usually nothing at all to do with speed of
applications. Except if you are in a extreme memory shortage.
Which nowadays usually does not apply. If you run KDE or Gnome, you use
around 100KByte more memory than in a very frugale WM.

--
Another name for a Windows tutorial is crash course

houghi wrote:
> Chris Cox wrote:
>> Well.. in the general case, openSUSE will be faster.
>
> And then much will depend on the application one is running.

When I switched from using Mozilla on Windows for reading news, to using
Mozilla on SuSE 9.1 on the same PC, it was very noticeable how much
faster it was to compact the message folders. At least twice as fast.


> And if
> speed is the objective, openSUSE is a strange choice. I would go for
> something like DSL. At least something that does not run KDE or GNOME as
> default.

Sure, but it's valid to compare Windows with KDE (or another desktop
environment with similar functionality to Windows.)



--

Ulick Magee

Free software and free formats for free information for free people.
Open Office for Windows/OSX/Linux: http://www.openoffice.org
OpenSUSE Linux: http://en.opensuse.org

On Thursday 20 Aug 2009 22:42, Peter Köhlmann played with alphabet spaghetti
and left this on the plate:

> houghi wrote:

>> And then muc`h will depend on the application one is running. And if
>> speed is the objective, openSUSE is a strange choice. I would go for
>> something like DSL. At least something that does not run KDE or GNOME as
>> default.
>>
>
> The choice of default DE has usually nothing at all to do with speed of
> applications.

It can, depending on just what eye-candy is being displayed.

> Except if you are in a extreme memory shortage.

Define extreme memory shortage.

> Which nowadays usually does not apply. If you run KDE or Gnome, you use
> around 100KByte more memory than in a very frugale WM.

You've tested that out? I gave it a simple test in three different desktop
environments on 11.1. The test grabbing a copy of the free memory at three
different points:

1, the console after the system has completed booting and is waiting at the
KDM login screen;
2, after starting up the desktop environment, waiting for it to fully load,
opening an xterm;
3, after leaving the desktop and returning to the KDM login screen.

The three desktop environments I tried[0] were WindowMaker, KDE 3.5 and KDE
4.1. Here's the results:

Windowmaker
total used free shared buffers cached
Mem: 383772 172676 211096 0 556 151356
-/+ buffers/cache: 20764 363008
Swap: 573432 0 573432
Total: 957204 172676 784528
total used free shared buffers cached
Mem: 383772 203820 179952 0 556 161756
-/+ buffers/cache: 41508 342264
Swap: 573432 0 573432
Total: 957204 203820 753384
total used free shared buffers cached
Mem: 383772 183756 200016 0 556 161892
-/+ buffers/cache: 21308 362464
Swap: 573432 0 573432
Total: 957204 183756 773448

KDE3.5
total used free shared buffers cached
Mem: 383772 172676 211096 0 556 151344
-/+ buffers/cache: 20776 362996
Swap: 573432 0 573432
Total: 957204 172676 784528
total used free shared buffers cached
Mem: 383772 339960 43812 0 556 257364
-/+ buffers/cache: 82040 301732
Swap: 573432 0 573432
Total: 957204 339960 617244
total used free shared buffers cached
Mem: 383772 297360 86412 0 556 273024
-/+ buffers/cache: 23780 359992
Swap: 573432 0 573432
Total: 957204 297360 659844

KDE4.1
total used free shared buffers cached
Mem: 383772 172924 210848 0 556 152504
-/+ buffers/cache: 19864 363908
Swap: 573432 0 573432
Total: 957204 172924 784280
total used free shared buffers cached
Mem: 383772 334884 48888 0 556 216988
-/+ buffers/cache: 117340 266432
Swap: 573432 0 573432
Total: 957204 334884 622320
total used free shared buffers cached
Mem: 383772 253744 130028 0 556 230324
-/+ buffers/cache: 22864 360908
Swap: 573432 0 573432
Total: 957204 253744 703460

Surprisingly, KDE3.5 uses up a little more memory than KDE 4.1, but both use
up 130MiB more than WindowMaker. In the end, just staring up the desktops
consumed a little over 30MiB, KDE 4.1 used up a bit more than 158MiB, and
KDE 3.5 used up a little more than 163MiB. That's just a little bit more
than your suggestion that they would only use 100KiB more than a frugal WM.


[0] I might go back and add the IceWM figures at a later stage, and maybe
the XFCE figures as well. The XFCE figures will have to wait though, I don't
have it installed on that particular VM.

Regards,
David Bolt

--
Team Acorn: www.distributed.net OGR-NG @ ~100Mnodes RC5-72 @ ~1Mkeys/s
openSUSE 10.3 32b | openSUSE 11.0 32b | |
openSUSE 10.3 64b | openSUSE 11.0 64b | openSUSE 11.1 64b | openSUSE 11.2m5
RISC OS 3.6 | RISC OS 3.11 | openSUSE 11.1 PPC | TOS 4.02

Van Chocstraw wrote:

> David Bolt wrote:
>> On Thursday 20 Aug 2009 19:55, Van Chocstraw played with alphabet
>> spaghetti and left this on the plate:
>>
>>> I have 2.225 gig of ram. OpenSUSE 11.1 loads the PAE kernel by default.
>>> Then I added Virtualbox and it added the KMP PAE kernel.
>>> I don't want PAE running and want it to only use 2.0 gig of ram.
>>
>> Why would you have the extra 256MiB of RAM and not want to use it?
>>
>>> I loaded the default kernel but the system won't boot completely.
>>> Complains of part 5 and part 6 of the harddrive and drops to the $.
>>
>> Sounds like it's not got the correct modules in the initrd. Try using the
>> rescue system to redo the boot loader. As a part of this process, a new
>> initrd is created. An alternative is to create the initrd manually.
>>
>>> How do I turn off PAE?
>>
>> Unless you're using real-time programs, you won't notice the slowdown
>> because of the PAE kernel. So, unless there's another compelling reason,
>> there's no real point to removing the PAE kernel.
>>
>> However, if that's what you're wanting to do, the easiest way is to use
>> YaST to install the default kernel with the same version number.
>>
>>
>> Regards,
>> David Bolt
>>
> I did use yast to install the same version default kernel. It's in the
> grub menu but has the above problem.
>
> Speed is all I had on my mind in changing. I just thought Opensuse would
> run faster without it, if not, I guess I'll forget it.
> Comparatively, on the same system, W2Kpro and WindowsME run considerably
> faster.

I know it doesn't help solve the original problem but IMHO it's unrealistic
to expect an operating system built in 2009; with 2009 hardware in mind, to
perform faster than an OS built in 2000 with 2000 hardware in mind. Not to
belabour the point but put win 3.1 on a pentium 4 and it'll fly like
nobody's business.

I apologize for not actually contributing to finding a solution to the
original problem. As far as this goes the only thing that I can think of
is trying to add apci=off to the boot config line. This may be way off
base, but at least it would rule out power management as being part of the
problem.

David Bolt wrote:

> Surprisingly, KDE3.5 uses up a little more memory than KDE 4.1, but both
> use up 130MiB more than WindowMaker. In the end, just staring up the
> desktops consumed a little over 30MiB, KDE 4.1 used up a bit more than
> 158MiB, and KDE 3.5 used up a little more than 163MiB. That's just a
> little bit more than your suggestion that they would only use 100KiB
> more than a frugal WM.

Thanks for sharing the figures. Did you check how long it takes for the
desktop environment to start up also? With RAM being so cheap these days
the main reason for me to use one of the light desktops or window
managers is that a login is much faster even with current hardware. This
is not much of an issue on a system where on logs in once a day or even
more rarely. But then I quite often just want to check something like a
printer queue and log in only for a minute or so.

Günther

This message is not archived

This message is not archived

This message is not archived

On Friday 21 Aug 2009 17:48, houghi played with alphabet spaghetti and left
this on the plate:

> David Bolt wrote:
>> Surprisingly, KDE3.5 uses up a little more memory than KDE 4.1, but both
>> use
>
> If by "Surprisingly" you mean you had expected a lot more, then I can
> agree. A lot of work in KDE 4.x went into reducing memory usage and
> increasing speed.

I know. I was expecting KDE4 to use quite a bit less memory, and to start at
least as quickly, if not to be quicker at loading.


Regards,
David Bolt

--
Team Acorn: www.distributed.net OGR-NG @ ~100Mnodes RC5-72 @ ~1Mkeys/s
openSUSE 10.3 32b | openSUSE 11.0 32b | |
openSUSE 10.3 64b | openSUSE 11.0 64b | openSUSE 11.1 64b | openSUSE 11.2m5
RISC OS 3.6 | RISC OS 3.11 | openSUSE 11.1 PPC | TOS 4.02

On Friday 21 Aug 2009 17:48, houghi played with alphabet spaghetti and left
this on the plate:

> David Bolt wrote:
>> Surprisingly, KDE3.5 uses up a little more memory than KDE 4.1, but both
>> use
>
> If by "Surprisingly" you mean you had expected a lot more, then I can
> agree. A lot of work in KDE 4.x went into reducing memory usage and
> increasing speed.

I know. I was quite expecting KDE4 to use a lot less memory, and at the very
least to start as quickly, if not to be quicker at loading up.

What was really surprising was the amount of memory XFCE used, and the speed
at which it was fully loaded. I'm going to have to look a bit more closely
at it since it's been quite a while since I sat down and actually used it
extensively. I'm curious as to whether that initial loading time carries
over onto the the rest of its use.


Regards,
David Bolt

--
Team Acorn: www.distributed.net OGR-NG @ ~100Mnodes RC5-72 @ ~1Mkeys/s
openSUSE 10.3 32b | openSUSE 11.0 32b | |
openSUSE 10.3 64b | openSUSE 11.0 64b | openSUSE 11.1 64b | openSUSE 11.2m5
RISC OS 3.6 | RISC OS 3.11 | openSUSE 11.1 PPC | TOS 4.02