Ubuntu 7.04 to 8.10 Benchmarks: Is Ubuntu Getting Slower?

[kuolas]

You should check this:

https://bugs.launchpad.net/ubuntu-mobile/+bug/228559

It's the kernel taking too much time for itself?

Try with PowerTOP and the tell me. I've always get 25% of CPU when doing nothing.

[marakaid]

These are games in Wine right? Some of them have bad performance because Direct3d is not optimised in wine. Native linux games are generally same speed or faster on linux I thought?

My findings are with ETQW 1.5 native.

[Sacha]

Linux I barely see Fps higher than 55 playing with Low/Normal settings, dropping to their 30s or even 20s most of the time (sometimes in heavy shootings indoor their 10s!), even with small shootings outdoor; the difference is very noticeable, I can assure you.

These are games in Wine right? Some of them have bad performance because Direct3d is not optimised in wine. Native linux games are generally same speed or faster on linux I thought?

[marakaid]

Have you some data? Is it like 80 -> 40 or 80 -> 70 ?
That's another thing. Intrepid is using the xorg 1.5 server. The driver should be nearby beta status. Perhaps there are some more fixes in the official driver?

Performance is horrible in Linux (Fedora, Suse, Mandriva, Ubuntu, Debian, ARch):
while in Windows I play nicely at highest settings with 60 fps average (frames locked and vsync on, so maybe I could see even more Fps) while using Linux I barely see Fps higher than 55 playing with Low/Normal settings, dropping to their 30s or even 20s most of the time (sometimes in heavy shootings indoor their 10s!), even with small shootings outdoor; the difference is very noticeable, I can assure you.

[Beleriand]

@Beleriand

You are testing different Nvidia drivers, best use my script:

Well, thats my intention. Imho the tests should show the results of the distros as they come along in the default installation. GPU-drivers make an exception to me because they can't be part of the standard installation as they are under closed license. But I installed the drivers available in the official repositories, as this should be the comfort way a typical users would take. I didn't make any other manual improvements and no additional installations than build-essential and libgtk2.0-dev as they are required by the test suite.

By the way these are my

• compilation results.

To be fair you should acknowledge, that code optimization is the attention getter of compilation and not the compilation itself. That may cause the time needed in these tests grows over the distro versions. Maybe the code is faster then, which unfortunately can't be proofed with the suite. Would be a good idea too, wouldn't it?

[Kano]

@Beleriand

You are testing different Nvidia drivers, best use my script:

http://kanotix.com/files/install-nvidia-debian.sh

To install always the latest (177.80 currently) driver on every distro. I optimized the script that it should work without problems with 32+64 bit Ubuntu 8.10 (in case of 8.10 it has to enable multiverse too).

[Sacha]

This is all just a trick to get us to use Phoronix's benchmarking tool and prove them wrong.

[Beleriand]

I also have 2 more tests (see my post on page 8 for more).

computational
gtkperf TotalTime

As you might criticize, theres some difference in screen resolutions.
This should not affect the computational result but maybe the gtkperf a bit. I will proof later that its not a big issue.

Furthermore I re-emphasize that I believe that something went wrong with the Phoronix test on audio-encoding. I wait in suspense for results on ohter platforms.

[glasen]

I'm currently running the audio-enconding tests of the test suite on a Thinkpad R31 (P-3 1GHz, 512MB RAM) on Ubuntu 8.10 and the tests that are finished (LAME, OGG, FLAC) are only ~10% lower than the numbers Phoronix got for their T60.

The T60 has a Core2Duo with 1.87Ghz, it should be twice as fast as a P-3 with 1GHz. Remember the scores for the audio-tests are in seconds!

       R31         T60
LAME  133.07s    120.83s
OGG   84.40s     69.81s
FLAC  62.13s     56.19s

Does now someone believe that the numbers Phoronix got for its T60 and Ubuntu 8.10 can't be right?

P.S :

I will upload the complete results when the tests are finished and post the link here in the forum.

Edit :

The complete results can be found here :

http://global.phoronix-test-suite.co...14-14176-26881

[tommcd]

I would be interested to see a benchmark comparison between Ubuntu and Debian. There is no question that Debian runs lighter and faster than Ubuntu. This would be the litmus test for just how bloated Ubuntu really is.