He has an older unit. He already stated that his unit is not using a lot of swap which is why I felt it was safe to make my recommendation.

Lufbery,

Your experience seems very anecdotal. I'm not sure how you are going to tackle a speed issue without trying to measure speed. However, I can understand a degraded performance (which isn't necessarily a speed thing).

So as for comparison, I can tell you just a month ago, I installed slackware 13.37 on a 500 MHz celeron laptop specifically for DVD and video playback, with only 128 MB of ram. I'm not sure if I even remembered to install a swap partition, on its 10GB hard disk. Anyway, the initial boot was slow, but it boots right into XFCE. After that, it felt very snappy. I did go ahead and use Xine (I've always seen Xine being slower than MPlayer). It plays back DVDs perfectly, without slow down, even if the graphics chip (a cheap onboard brand) offers no acceleration to speak of. I also point out, that this 12 year old machine can play DVDs better than Vista machines (their video skips) my dad was using. It also plays youtube videos fine through the web browser using MPlayer through a plugin. (not using flash of course--I wouldn't even use that on my quad core _with_ acceleration) And browsing did not feel sluggish.

I've been using Slackware for a very long time now. I've not noticed a dramatic resource hog with much anything. The only thing that comes to the point of being slow is KDE4. And please people, don't preach the virtues of the triplets. I literally don't care--even after you let them index, KDE does feel sluggish as a WM, and even if you do use completely supported graphics drivers. Based on reading your history, I would think your main beef is with KDE4.

3 members found this post helpful.

Many of the software packages naturally get bigger and better as the standard hardware gets more capable. That seems to be the nature of things. I used to run Slackware 10 with full GNOME desktop on my Thinkpad T20 (~2000 vintage) and it seemed pretty snappy. Today I run fluxbox or LXDE and it's about the same, both in terms of performance and features.

Heck, I remember installing full slackware on a 500MB HD on the old 486 back in the day!

I think you have two choices, running the latest version with lighter software and fewer services, or running an older version. Nothing wrong with the old versions for old hardware; there are still security updates for Slackware 12.0, which is over four years old.

This is an interesting topic. I too have had the perception that something is not quite right with free/libre software and older hardware. I presumed the problem to be the kernel, udev, hal, dbus, or something similar.

I have a PI (400 MHz K6-III+, 256 MB RAM) and PII (350 MHz Deschutes, 448 MB RAM) that I tinker with occasionally. The PI machine still has WFWG 3.11 and NT4 partitions installed and those systems are snappy. Not so with the free/libre partitions.

I never dug deep to learn more. Nothing serious but seems to me that as the years have passed free/libre software gets slower on these machines. I have a 13.1 partition on a spare drive that I swap between the two machines and 13.1 crawls compared to 12.2, which also is installed.

I run a bare minimum of services on these machines.

I wonder whether these machines would be faster with the 2.4 series kernel. The hardware in these machines are static. Other than a a CD player in each machine there is no need for dynamic detection of devices. The old hotplug system worked okay. with that.

I suppose running 11.0 --- the last release with the 2.4 kernel, is doable and I could rebuild many packages to update that overall environment. Security updates is not a serious issue because at some point maintaining old hardware becomes a hobby rather than a necessity. Sounds like an interesting experiment....

One thing I have not imagined: free/libre web browsers never ran smooth on these two machines. Both machines have 10/100 NICs.

I'd be interested in learning how a person could test in a meaningful and controlled manner this perception of running slower.

1 members found this post helpful.

Comparing NT4 to any current Linux distribution seems irrelevant to me. Heck, Winodws XP is 10 years old and probably wouldn't run on those boxes.

I have not noticed any slowdown with newer Slackware versions. However, I do not use ancient hardware, so I don't know. In fact, I use only x86_64 on all my machines.

It would be nice if you had some real benchmarks to better quantify the performance loss, and if it wasn't due to some other issues (hardware).

Current may be slower after the recent jump to a 3.2 kernel, as this introduced a bug (started in 2.6.39) which causes much slower flash drive access and probably slows harddisk reads too.

I've found the 3.2.x kernel to be much faster than other kernels. They fixed a major issue relating to cache usage.

Here's some anecdotal data from me:

The machine is 4.5 years old ThinkPad R61i with Intel Core2 Duo 1.5 GHz CPU and 4GB DDR2 RAM, so it can still be considered 'modern'. Started with KDE 3.5 when I bought it (32-bit Slackware back then), now using Eric's KDE 4.8 packages (Slackware64-current with latest version of custom-built kernel). The infamous KDE4 triplet is disabled, because I simply do not have a need for it and I dislike having some services running and using resources needlessly.

During those years of usage I haven't noticed any significant slowdown. On the contrary, I believe that latest KDE 4.8 is faster than KDE 3.5 ever was at drawing windows. I have compositing enabled, and all the animations and redraws feel quite smooth (Intel 965GM graphics chip). Also Opera (my main browser) got much faster.

That's good to know.

I've built Linux From Scratch on this laptop using Slackware as a host. I'm about to get started on LFS 7.1 (when they release it), and it uses the 3.2.5 kernel. We'll see how it goes once I get networking, X, (perhaps) KDE, Seamonkey (and/or Opera) and LibreOffice installed.

By the way, GLibc took several hours to compile on the laptop!

Meanwhile, I fully admit that my experience is absolutely anecdotal. The laptop is still fast enough for my needs, but I have perceived a little bit of a slowdown over time.

FWIW, my desktop computer, where I do most of my work, is a dual-core Athalon with a good Nvidia graphics card and 4 GB of RAM. It is very fast, and each new version of KDE seems to get better on it.

Regards,

My thought is that later release software expects to be able to use more RAM. Currently I have just Firefox 10.0, Dolphin and a bash shell open in KDE 4.5.5 and 'free' reports
So half of the available 1GB of RAM is in use. Older machines do not have the RAM available, so use disk swap, which is slow.

It could be like allend said, that newer apps require more RAM, but I haven't found this to be the case. I have gkrellm available all the time and my system uses the same memory usage as it always has plus or minus some megabytes.

The only thing that made a small bit of difference in RAM usage was enabling huge pages all the time and zcache, but it is still only a small bit higher than usual. I only have 2 GB or RAM and it rarely uses more than about 500 MB of RAM, same for cache.

Although firefox does tend to use a lot of RAM, I think recent versions are better than older ones. They've fixed a lot of memory leaks.

This is what leads me to believe that the slowdowns that you see may be caused by ancient hardware getting older and older. I would run 'badblocks' or a smartctl long test on the HDD.

The problem is not well defined; you use a big bunch of software, and therefore I'd expect you to find the newer releases are slower than old ones, if the hardware is the same. The "Slackware release with all the bells and whistles turned on" probably does get slower (should--I don't believe the software can be optimized well enough to actually make it faster while adding features all the time), but "Slackware" as a bare system is a different thing. If you had only those parts of the system updated that really are required to be updated and then ran the same benchmarking software (same versions) on all of them, you might get slightly different results--not necessarily the opposite, but I'd guess "better" compared to this situation. A minimal install, with no other software than what is absolutely required, and a few pieces of test software that run on each of the releases, and there you go.

If you can't run KDE4 with today's software today, I'm pretty sure you couldn't have done that six years ago. That's the point.

I never said I couldn't run KDE4. I run it and use it every day on my ancient system. It works well. It seems, however, a tiny bit slower on the same machine in Slackware 13.37 than the version of KDE that runs under Slackware 13.1.

I do wonder if a base Linux system is slower than it used to be. Do you think that Polkit, Consolekit, and various changes to X, udev, HAL (are we using HAL any longer?), are slowing things down a little bit?

Regards,

In my opinion, it is the graphics chip.
According to a search engine, your Thinkpad comes with an ATI Rage 128 graphics chip (ATI RAGE Mobility M3) which is currently supported via the r128 Xorg driver.
The performance of this r128 driver has become worse with recent releases.

The first problem ist that this driver is not actively maintained.

The second one ist that r128 still relies on XAA for 2D hardware acceleration and through an Xorg commit in mid-2008 big parts of this XAA support have been disabled.

I made similar observations like you:
On my r128 machine openSUSE 11.1, Debian Lenny and Slackware 11.0 worked quite well, although 3D support was always a little buggy.
Now in openSUSE 12.1 this ATI Rage 128 card is unusable. If I try to change the size of a window, Xorg freezes. In contrast to this, the older Debian Squeeze still works, but Squeeze is slower on this machine than Lenny. But it is usable.

But the main problem is the future.
So here is the third problem:
With the switch from Mesa 7.11 to Mesa 8.0 Mesa support for r128 has been abandoned (as well as for all old Mesa DRI1 drivers). So with Mesa 8.0 onwards 3D hardware acceleration is disabled for r128.
I suppose that the whole desktop experience could become a little bit "slower".

And here is the fourth problem:
With a recent commit for xserver-xorg 1.13 XAA has been removed completely, so that r128 chips will also lose 2D hardware acceleration in the future, because r128 relies on XAA to get this job done (see above). 2D hardware acceleration for r128 could be "reinstated" with the switch from XAA to EXA, but (as far as I know) no one is working on this. But there exists something like an unofficial (?) r128 EXA port from GSOC 2006, but I do not know how usable this is.

So the r128 graphics chip will probably lose 2D and 3D hardware acceleration support in the nearer future. So, yes, you are right, your machine has become slower and will become slower as a result of massive Xorg and Mesa changes and the fact that there is no maintainer for the r128 driver.

At this future point of time, one possibility could be to use r128 (or vesa or fbdev) with "shadowfb" turned on for better performance.

2 members found this post helpful.