...at least on my laptop, anyway. It's great on my desktop.

Really the main thing that is slow is program loading times. It seems like it might be disk I/O related, but the bottom line is that certain things are significantly slower on this machine in FC5 than when it was running FC4. The most annoying trouble areas would be...

* Firefox taking 22 seconds to load and become responsive
* The GNOME Desktop "hangs" for about 12-15 seconds at the splash screen after logging in (the Fedora splash screen goes up but no progress icons appear for a long delay).
* Adobe Acrobat takes a long time to load (haven't timed it). This is the same for either the standalone viewere or the Firefox plugin. This might not be FC5 specific, however; I didn't test this on FC4.

These two delays make for painful wait between the login screen and being able to actually use the Internet. Moreover, they're not really "hangs" that I'm experienceing because the HDD light is reading the whole time, so it's doing something... just doing something as slow as molasses. I notice similar delays in other areas (i.e. even Terminal takes about 6-7 seconds to hit the screen). This is a Turion64 3700+ laptop with a gig of RAM and a fast HDD and should not be experiencing such delays.

Just for the heck of it, here's what 'hdparm -i /dev/hda' says:

If anybody has been experiencing similar delays or can help me out, I'd really appreciate it!

Thanks

Have you left your machine up for a day after installing, to allow prelink to run? If not, you can run it manually (as root):

/etc/cron.daily/prelink

Hi All,

I'd just like to chime in that from a newbie prespective it does appear slow. The computer is old, but before installing I made sure I met the system requirements. At start up is there a way to make Fedora boot automatically into the non-interactive mode to speed up start up times? Right now it comes up and prompts me to select interactive. When the GUI prompt appears (I am using KDE) I have seen the logon process takes seconds and other times up to 15 - 20 seconds. I have 200MB of RAM in the system. Do I need more? Finally, if someone can help me configure my wireless card that would be appreciated. I have a D-Link WDA-2320 and Fedora is having problems auto detecting it. Shut down times are on par if not faster than shutting down a XP machine. Thanks.

I too am having problems with FC5 detecting my wireless card, which is a realtek clone based on rt2500???I think!!!
everything else about FC5 is fine, including a 1st class display and installation - though I don't understand what the VOLs are!
if I could get the wireless card working I would be a happy bunny!

There's another performance issue, as described in this bug. You can add:

export MOZ_DISABLE_PANGO=1

to your ~/.basrc or ~/.bash_profile and log out / back in to work around it.

Dykesy61: You should start a new thread for a different issue. However, you can download the software for rt2500 cards from here. The installation is just:

macemontea: Thank you for your responses. I have tried both running prelink and disabling Pango as you said, but as the following benchmarks suggest, it made little difference.

Program loading time was recorded with a stopwatch from the instant the application was clicked or Enter was hit (i.e. for logins). All times are for the first time a program was loaded on a fresh boot (i.e. it was not cached). Both benchmarks were done on the same machine and both distros had the proper ATI propietary drivers installed. Fedora Core 5 numbers are listed below and the numbers for Ubuntu Dapper (latest beta with all updates) are listed in parenthesis, where available...

1. GNOME Login time (from "Enter" until Nautilus appears): 30 seconds (Ubuntu: 16 seconds)
2. GNOME Login time (from "Enter" until the clock applet loads: 38 seconds (Ubuntu: 24 seconds)
3. OOo Writer: 38 seconds (Ubuntu: 30 seconds)
4. amaroK: 16 seconds (Ubuntu: 15 seconds)
5. gedit: 8 seconds (Ubuntu: 3 seconds)
6. Adobe Reader 7.0.5: 19 seconds
7. FC5: Firefox (Pango disabled): 19 seconds to window, 21 seconds until usable
Ubuntu: Firefox (no modifications--does it have Pango by default?): 7 seconds until usable
8. Totem: 7 seconds (Ubuntu: 2 seconds)

As you can tell, many of the FC5 numbers lag far behind Ubuntu on my machine. My intention is obviously not to start a flame war between which distro is "better"; I just wanted to set a reference point for this machine. In reality, I'm not terribly pleased with Ubuntu's performance on this machine either, but it is far more usable than FC5. Both distros use GNOME 2.14.1, so this should be apples-and-apples.

Is anybody else experiencing this kind of dissapointing performance on their machines? Any help is greatly appreciated!

Unfortunately, there are many variables between distributions, and there are tradeoffs. Simply having the same level of Gnome (of the thousand packages that make up a typical distribution installation) is not enough.

Once all the packages are at the same level with the same patches, there are tuning options that differ. Which I/O scheduler is in use? What readahead settings are configured? Are hard drive parameters configured the same? Are filesystem parameters configured the same?

Many of these choices are intentionally set to more conservative values, because they work better across a wider array of usage patterns. For example, while the load time may be slower in FC5, is it more consistent when the system is under stress?

You can certainly spend considerable time tuning the various parameters. If you go down that path, just remember that your system has a finite set of resources, and improving one will come at some cost to another.

macemoneta,

What you say is all definitely true, and I expect there must be differences in overall optimization between distros. However, some of these differences are on the order of magnitude of 3 to 4 times faster on Ubuntu. If this were the kind of performance differential that everybody was seeing between FC5 and other distros, then I expect we would be seeing many many more reports of people being unhappy with FC5's performance. To the contrary, I've actually been seeing more praise for FC5's performance than anything else--and I've experienced that on my desktop box.

I think a more appropriate comparison here would be between FC4 and FC5 on my machine (I wish I could do that but I've run out of extra partitions). I can confidently say that FC4 was much more usable on this laptop than FC5 is, but I'm always reluctant to make such claims without backing them up with numbers.

I have to think that with those kinds of performance differences that there is something wrong--either a hardware compatibility issue or configuration or otherwise. Either that, or everybody else is experiencing similar performance drops after upgrading to FC5 and just aren't saying anything (but I doubt that ).

The most obvious hardware compatibility issue that would impact load times is in the detection and setting of hard drive parameters. Is DMA enabled? I/O unmasking? 32-bit transfers? Use the hdparm command to check. For example:

By default, interrupt unmasking and 32-bit I/O transfers were not enabled; that's why I placed this in /etc/rc.local :

I posted the output for "hdparm -i /dev/hda" for FC5 in the original post. For FC5, "hdparm -v /dev/hda" says:

Again, these are not the default settings but rather the settings after I added the line in rc.local . Incidently, neither interrupt unmasking nor 32-bit I/O transfers are enabled in Ubuntu by default, either, yet I still get better performance in Dapper. Here's what "hdparm -v /dev/hda" says in Ubuntu...

FC5 does seem slow compared to FC3 - I find that apps take 10-20 secs to load. Even a terminal window takes 10 secs! On FC3 most things were availible in 1-2 secs. With the specs for my machine below, I'm wondering where the bottleneck is.

My PC has:
A P4 2.8GHz Chip
Gigabyte GA-8IPE1000 Pro 2 Motherboard - 800MHz FSB
400MHz DDR dual channel Memory Slots with 512Mb Ram
150Mb/s SATA with 200Gb Samsung SP2004C Serial ATA Hard Drive connected on /dev/sda no VolGroup
IBM 20Gb IDE on /dev/hda with VolGroup

Cable Internet Connection - usually runs at 170Kb/s

Kernel: (2.6.16-1.2122_FC5smp)

hdparm -tT /dev/hda
/dev/hda:
Timing cached reads: 3324 MB in 2.00 seconds = 1664.63 MB/sec
Timing buffered disk reads: 90 MB in 3.01 seconds = 29.95 MB/sec

hdparm -tT /dev/sda

/dev/sda:
Timing cached reads: 3272 MB in 2.00 seconds = 1636.84 MB/sec
Timing buffered disk reads: 172 MB in 3.01 seconds = 57.21 MB/sec

Now - what I have done so far:

I installed FC3 on the 20Gig IBM drive(now small!) 2 years ago & everything was fine.
The other day I installed the 200Gig SATA Drive & FC5, immediately noticed the difference speed of loading apps. Not good considering the above specs & the new faster drive.
SELinux is disabled in case that was slowing things up - I don't really need it considering I am the only user.

ALL the packages on this machine have been updated - definite overkill with an 800Mb download!!

Is there more that I can do with the SATA drive? It seems that most of the hdparm options don't apply to SATA drives. I know that hd makers overstate the specs a bit, but the mother bd is capable of 150Mb/s & it's seems mine is only acheiving 57Mb/s.

A VolGroup is active on /dev/hda even though it is not always mounted, - are there hassles running 2 drives - 1 with a VolGroup & 1 without?

May be there is a problem with the kernel. Both the current one (installed yesterday) & the original one on the DVD (2.6.15-1.2054_FC5)are slow.


I agree with Sancho - perhaps FC5 shouldn't be this slow out of the box.

FC5 isn't this slow; there is something wrong on your machine. I'm running FC5 on a 5+ year old P4M 1.7GHz laptop, and operations are 3-4x faster than you report (even with a slow laptop hard drive), and I run with SELinux.

Have you stopped unneeded services? Does 'top' show excessive CPU, memory or swap usage? Are there any unusual messages in 'dmesg' or /var/log/messages?

Thank you for your prompt reply.
Today I noticed everything was going much faster, although I hadn't done anything. Then I saw a reply of yours which mentioned cron/prelink - That's probably it!!

As well as that I looked at your unneeded services page & stopped about a dozen services - so that reduced RAM usage by 10 %. The RAM usage wasn't high before, but I think that your unneeded services page is very useful & we are all lucky to have the vast experience of people like yourself to help everyone out.

Thanks Heaps

David Saunders (worzel68)