Benchmarks: Mandriva, PCLinuxOS, Ubuntu, openSUSE

[RealNC]

You didn't post a single argument, but I did. Also, flames came with you here. I explained later why it sucks for me. I don't consider PCLinuxOS won graphic tests thanks to BFS and that's why I replied.

I don't believe it either, since BFS does not improve performance. I had to react though because you keep flaming at a project that is a Win with capital "W".

So, in the same spirit: CFS sucks. I urge everyone not to use that crap. It's crap on my machine, therefore it's crap everywhere.

[kraftman]

I don't believe it either, since BFS does not improve performance. I had to react though because you keep flaming at a project that is a Win with capital "W".

BFS does improve performance in some things - afaik 5% improvement in video encoding, but it's due to different design of both schedulers, so 5% gain is probably some loose in different thing. It also shows advantages in some Phoronix benchmarks. However, It's not a "Win" for me, because it makes things only worse on my PC right now. However, it was a "Win" in some part for CFS.*

So, in the same spirit: CFS sucks. I urge everyone not to use that crap. It's crap on my machine, therefore it's crap everywhere.

I explained why I don't love BFS too much yet - because of occasional problems with input which sometimes stops working, sometimes X Server is being killed and you return to KDM/GDM and some people reported, it doesn't boot with some devices plugged in. If such things don't happen to you it's ok, but your problems with CFS are nothing compared to mentioned problems with BFS. I also don't urge anybody to not use it. What I know and you probably don't such problems with BFS are worked on and should be fixed some day, but I'm talking about its current version. You were talking in the same spirit about CFS many times.

Now,come on,...kraftman you really don't like to BFS take any credit in this

Thanks to Con and his BFS there were some issues with CFS fixed* There are some things why I don't like it, but there are also some positive ones. I don't agree BFS is better for desktop, but it doesn't matter. My point was PCLinuxOS victory in many tests isn't only BFS merit, but maybe I should say this in a different way CFS is more bullet proof at this point, but BFS is better for many people (but the same can be said about CFS).

[Zajec]

That's intresting. Why screen size affects only UMS performance and not KMS?

In case of KMS we use memory management and we can allocate memory for screen dynamically. It's not possible in UMS where we do not use memory management.

It may happen you allocate memory for bigger size than you actually need (it's for UMS) and it impacts performance making it worse.

[mitcoes]

Phoronix: Benchmarks: Mandriva, PCLinuxOS, Ubuntu, openSUSE

Last week we delivered benchmarks of Fedora 13 Alpha and Ubuntu 10.04 (along with testing the Fedora 11 and 12 too), but today we have a new set of comparative benchmarks that are covering the latest development versions of Ubuntu 10.04, Mandriva 2010.1, PCLinuxOS 2010, and openSUSE 11.3. Here they are.

http://www.phoronix.com/vr.php?view=14669

For compare , it would be better to make an average result.
Being the best in 10 of 12 test, can be not the best, if you win by a 5% 10 tests and lose by 40% 2 versus the same OS in thsi case, the other is better.

Phoronix bench MUST have a average number result, with the difficulty of giving a % of score for each bench. And a "standard" 100 base that can be changed with years (as it was IBM AT), where every benchmark would be normalized to 100.

This normalization where you can print the brute value, and the normalized one would be easier to understand, and easier to make a choice when you read a benchmark.

Next Ubuntu LTS with a new i3, 5, 7 or 9 Apple model can be the standard to compare with other OSs and computers. With all benchs normalized to 100 - Apple model x - and ubuntu 10 LTS for next 4 or 5 years, and change then to a new base.

Stats must be easily understandable or are not a good tool.

The conclusion itself of the article is not for a good statistical understander is more for a sport score fan - of course a team that loses 10 matches by 5% score, an wins 2 by double, being better than other, won't win the competition, but machines are not sports and benchmarks are for knowing what is better in average, even average must be cooked, and give the results for several tests for knowing where you have your bests and your worsts.

[DarkFoss]

kms is disabled by default using mandriva ..you have to add radeon.modeset=1 to the kernel line in grub...on my 64 bit agp system I also had to add radeon to /etc/module.preload and disable speedboot.

if you use the alternate 2.6.33-tmb-desktop-5mdv provided by Thomas Backlund he's enabled kms by default as well as backported evergreen support (and more) from the .34 kernel and I just needed the radeon module listed in module.preload.

[Dr_ST]

I'm unhappy to see those missing numbers from some distributions in a comparison test.

I've made my tests for Linux Identity Kit (soon to appear) and most of the time there are little compilations issue that can be overcome easily (on stable version of the distributions, Fedora, Ubuntu, Mint, Mandriva and openSUSE, not PCLinuxOS).

If you see have the logs of the installation, i'd happily continue to contribute to phoronix-test-suite with patches for fixing the problems (on priority for Mandriva since i know it best, but it will probably apply for other distributions).