Running Ubuntu 9.10 With Older PC Hardware

[V!NCENT]

They provide some bug reporting tool - apport and maybe some debugging options must be turned on to make it working.

Maybe it's a JIT bedugger?

[kraftman]

Maybe it's a JIT bedugger?

I don't know exactly.

https://wiki.ubuntu.com/Apport

It seems it will be disabled in stable release:

Apport is not enabled by default in stable releases, even if it is installed. There are two ways to enable it.

[kuolas]

Yesterday I tried Ubuntu 9.10 Beta on a Gateway E4300.

The specs:
Pentium III 500MHz
Intel 440BX
RAM 512MiB
HDD 80GiB WD
Promise Ultra66.
Savage4 32MB
Philips SonicEdge

I must say... It runs.

Of course, Dapper was speedier than Karmic, but having the latest "vital" software pack worth it (mostly Firefox and OpenOffice.org).

I used Dapper Xubuntu on a Pentium Classic 133MHz and 64MiB of RAM... slow but runs Firefox, and I mean ONLY Firefox another app and the system was unusable.

My currently workhorse machine is "NapaValley" HP 530, an Intel Napa based platform.

The specs:
Core Duo T2400 1.83GHz
Intel 945GME
RAM 1GiB
HDD 120GiB WD Scorpio SATA150

It works flalessly since the day one... the day that I've installed Ubuntu 8.04 Beta. Since then this OS is still beating, and I pretend to be so until the next Ubuntu LTS... I have an encrypted LVM, so the upgrade in this scenario is critical. I hope that the next LTS will come with a decent Intel video driver.

I bought Intel because of the "strong linux driver support", It's a lie. I hate Keithp because he constantly breaks the X windowing system for no good reason (developing GEM was and is still a bad idea, UXA is another nightmare), and don't follow a good development model, nobody learned anything from the development of the Linux kernel? That's the best example on how to develop software. That and the WiFi... their drivers are updated between eons.

So, in the end... vintage hardware is excellent. You shouldn't measure performance numbers but "usability" feeling of a "vintage" machine. For me, I leave the old hardware (Pentium and 486DX2) for the command line. In that way the latest software is available... not the latest GUI.

My machine is today a "netbook" class, abeit with a speedier processor. I don't know why people complaint about the 945GME, It's a very good chipset. It should be better since a long standing hardware means mature and stable drivers... of course in the intel arena it isn't.

[V!NCENT]

@kuolas: Currently Linux is in FLOSS graphics driver hell (and it was probably worse back in the day, but my (even then old) Pentium II computer had a nVidia graphics card and with their proprietary driver (back then I didn't even knew KDE wasn't the 'Linux user interface') I wasn't noticing it, but it was probably ten times worse...

Hopefully (and probably) with the AMD documentation ATI Gallium3D FLOSS drivers will end this nightmare.

That said computers today are a complete mess. Windows XP sucks and newer Windows versions are too heavy for my liking still, Apple is too expensive and nothing really satisfies me completely about it and Linux is stuck with graphics suckage... What about nVidia binaries? Don't even get me started on their silicon failure rate (which is probably also linked to why they are not going FLOSS)

*sigh* <- that's all I can currently say about it

My advice to you is to 'sit it out' and buy a cheap ass ATI card when ou know the Gallium3D drivers are solid

[tuxdriver]

WTF! That's an almost generalized huge performance drop!
What about the old refrain stating that Linux is so suited for older rig?

Last year I tried a Xubuntu 7.10 on a P3 500MHz laptop... ok quite old, but it was almost unusable (but I wrote my PhD thesis on it with win2000!!!!)

I think that unless you get Linux From Scratch or distros that openly aim at small/weak/old hardware... Linux has in time evolved aiming at bleeding edge rig. I'm not angry about it at all... but at least defy the old beliefs about Linux being able to resuscitate old HW.

IMHO, I think that two areas are mainly responsible for that performance drop: modern graphic drivers (blobs) are getting more and more complicated and designed over newest hardware.. and the kernel itself (I have modern schedulers in mind).
So.. is Linux getting more and more feature rich at the cost of getting heavier and heavier?
I couldn't explain otherwise the performance increase with modern HW and a slow-down with the older.

Anyone has different opinions or more clues to explain the results of the test?

Try this experimental build of Karmic Koala on your P500 laptop:

http://puppylinux.com/test/karmic-puppy-20091024/

and feel the huge difference in speed

It's an Ubuntu / Puppy Linux hybrid created from Karmic Koala packages.

This is a snapshot of "Karmic Puppy", a puppy built from Ubuntu packages, specifically Karmic Koala 9.10RC. It was built using the 'Woof' build system, on 24th October 2009.

The live-CD 'upup-432.iso' is an example build. It has the 2.6.30.5 SMP 
kernel that's used in the latest version of Puppy.

Since this is an experimental build, I'm not really interested in bug reports related to the build itself, only in anything related to the build system.

[tormod]

Have you looked at kernel config? Debugging is turned on.

Some kernel debug flags are on and have always been. Some modules will print out a bit more messsages, and the kernel binary is slightly bigger than it has to be, but the performance impact should be minimal.

[Peter_Cordes]

I am particularly intrigued by the last test: MAFFT Multiple alignment program. What could this be due to? Is there anything in gcc or libc that could be making this slight but consistent improvement over time?

Yes, newer versions of gcc can generate better code from the same source. This is the whole point of what a lot of the changes to gcc! Optimizing compilers are a hard problem. Lots of developers spend lots of time working on gcc to make it generate better code. They've obviously had some measure of success, as MAFFT runs faster when compiled with newer gcc.

[Apopas]

Yes, newer versions of gcc can generate better code from the same source. This is the whole point of what a lot of the changes to gcc! Optimizing compilers are a hard problem. Lots of developers spend lots of time working on gcc to make it generate better code. They've obviously had some measure of success, as MAFFT runs faster when compiled with newer gcc.

Indded. I made also some multimedia tests under Gentoo and found that they performed better with tha latest gcc.
For example that's what I got with lame:

gcc-4.3.4 + O3 = 1m6.207s
gcc-4.4.2 + O2 = 1m5.916s
gcc-4.4.2 + O3 = 1m5.716s

I repeated the tests five times.
Well not a tremendous boost, but even O2 is a bit faster than an older gcc with O3.

[Panix]

karmic is abysmal

C'mon, tell the truth.

Ubuntu 9.10 is absolutely crap on older hardware. I cannot use it on my T41. I tried the LiveCD and it crashes each time. It's practically the same as it was at the Alpha stage. Imho, the developers didn't care about older hardware and didn't address the issues. What other reason could there be?

Ubuntu has a rep, too, for horrible support for users with ATI cards, the older ones or the new.

What a POS! I'll be installing sidux or Debian, thank you!

[tormod]

Truth is that Ubuntu 9.10 has pretty much the same Xorg stack as in Debian sid, so please keep us posted about your findings.