Fedora 18 vs. Ubuntu 12.10, Ubuntu 13.04 Benchmarks

[phoronix]

Phoronix: Fedora 18 vs. Ubuntu 12.10, Ubuntu 13.04 Benchmarks

As the next chapter after the Fedora 17 vs. Fedora 18 benchmarks for the Red Hat sponsored Linux distribution, here are benchmarks comparing Fedora 18 to Ubuntu 12.10 and Ubuntu 13.04 on two separate PCs...

http://www.phoronix.com/vr.php?view=MTI4NzQ

[cardboard]

Why the discrepancy?

[Sergio]

Fedora has always been buggy

Absolutely EVERY time i've given Fedora a chance there is ALWAYS an issue... last time (Fedora 17) had issue with both Intel and ATI GPU's; had the latest versions of the Xorg drivers but X wouldn't start (my chips are not latest generation so I suppose they must be supported), and Catalyst gave segmentation fault when starting X.
Ubuntu, on the other hand, just works.
Even if buggy Fedora showed faster, I would take Ubuntu (or Gentoo, or FreeBSD) any time over it.

[FourDMusic]

Not representative

I've def had my problems with Fedora. But these benchmarks are totally unrepresentative in comparing the two distros.

In the article, there were 6 benchmarks shown. 3 of which ubuntu was outperformed fedora.

The total number of benchmarks run by Michael was 29. There are only THREE of which ubuntu definitely outperformed fedora (I'm not counting the cases in which the difference in performance is negligible). That's right. Most of the rest of the benchmarks, the two distributions were about equal. Two of the other benchmarks, fedora outperformed ubuntu.

I don't know if Michael just grabbed a random 6, but I should hope that he pays more attention in the future so readers do not form incorrect impressions or conclusions.

Here's a link to the other benchmarks from the article: http://openbenchmarking.org/result/1...OM75&obr_imw=y

[mmrezaie]

The Reason

Can anyone explain the reason in some of the benchmarks (like apache) fedora is so slow? I wonder if there is a bug or ubuntu is using a special patch or something! In NAS benchmark seems ubuntu team are maintaining their repository very good.

[Tiger_Coder]

Can anyone explain the reason in some of the benchmarks (like apache) fedora is so slow? I wonder if there is a bug or ubuntu is using a special patch or something! In NAS benchmark seems ubuntu team are maintaining their repository very good.

Only a guess, but security system of Fedora like SELinux and/or firewall is the cause

[Rexilion]

The total number of benchmarks run by Michael was 29. There are only THREE of which ubuntu definitely outperformed fedora (I'm not counting the cases in which the difference in performance is negligible).

You prefer an article which shows 26 times the same conclusion (it's equal!) with a graph?

Reporting the major differences makes it interesting. All the more, as there should be no (or neglible) change between these distributions since they are released on about the same time.

[phil]

poor form

Dude, very poor form on only showing the benchmark graphs where Ubuntu was clearly "faster" than fedora. Yet, in another post, you compare the gaming performance of different compositors on ubuntu. From the OB.org results it seems that for some games and work loads Fedora is clearly faster than Ubuntu. Please edit the comparison post to include the subjectively relevant in game performance figures. Yes, you included a link to the openbenchmarking.org stats, but it seems to be that you have cherry picked certain results.

PLUS, if you run Fedora as a production web server then you should possibly reconsider that due to it's short development and production life cycles; unless you have a sweet as lifetime management system. The more relevant comparative bench marks, IMO, are those which are relevant to end users - desktop/laptop performance.

It would also be nice if you included the Fedora version that more closely matches the equivalent Ubuntu versions you benchmarked against to provide a more balanced comparison - either Fedora 17 or the rawhide F19 (I'm pretty sure that 13.04 is in a similar state to 19 - pre-release).

Ta,

Phil

[Rexilion]

Dude, very poor form on only showing the benchmark graphs where Ubuntu was clearly "faster" than fedora.

These were the only differing benchmarks AFAIK. No point in showing benchmarks were everything is the same.

Yet, in another post, you compare the gaming performance of different compositors on ubuntu. From the OB.org results it seems that for some games and work loads Fedora is clearly faster than Ubuntu.

Different compositors and different distributions are indeed a seperate comparison. So yes, the outcomes can be different. What is the point?

Please edit the comparison post to include the subjectively relevant in game performance figures. Yes, you included a link to the openbenchmarking.org stats, but it seems to be that you have cherry picked certain results.

FPS figures are subjective? Please explain.

PLUS, if you run Fedora as a production web server then you should possibly reconsider that due to it's short development and production life cycles; unless you have a sweet as lifetime management system. The more relevant comparative bench marks, IMO, are those which are relevant to end users - desktop/laptop performance.

This argument does not in any way invalidate said results. You argue about the different distributions used in this test. So what? Furthermore, Ubuntu and Fedora both have approx. 6 month development cycles.

It would also be nice if you included the Fedora version that more closely matches the equivalent Ubuntu versions you benchmarked against to provide a more balanced comparison - either Fedora 17 or the rawhide F19 (I'm pretty sure that 13.04 is in a similar state to 19 - pre-release).

I'm not sure, packages are frozen before the final version is rolled out.

[phil]

These were the only differing benchmarks AFAIK. No point in showing benchmarks were everything is the same.

Agreed that showing benchmarks that are the same across all platforms is not really intuitive as it just shows that nothing was done in that particular area of development, but there was a great variety of results in the game benchmarks. Between the F and both U's, this really highlit the steps forward in only 6 months of development. This in part answers another of your questions below. My argument is in part in relation to where you would use a cutting edge distro and IMO as a storage device, i would strongly advise against using the latest of any distribution in this role. In reality these distro's are more likely to be used for gaming than RHEL or an LTS release, so, why not show the benchmarks for that target??

Different compositors and different distributions are indeed a seperate comparison. So yes, the outcomes can be different. What is the point?

Well, the point of benchmarking is to eliminate as many variables as possible and provide comparative results. As gnome-shell is available on the U's it would be greatly beneficial to provide comparative results with gnome-shell to give a more accurate cross distribution comparison. By showing the in game results on this F vs U benchmark it makes it easier for people to correlate those results with the cross compositor benchmarks of http://www.phoronix.com/scan.php?pag...desktop1&num=1

FPS figures are subjective? Please explain.

the subject in the case of benchmarking is, as i put just above, benchmarking that which something is mostly used for. For current/modern distro's game performance is becoming an important metric, so why leave it out?

This argument does not in any way invalidate said results. You argue about the different distributions used in this test. So what? Furthermore, Ubuntu and Fedora both have approx. 6 month development cycles.

During the life cycle of a product many tweaks, patches and updates are applied. For server/enterprise class tests (web perf / db perf / io capacity) it is a bit futile to use a development distribution. For those tests it would be more appropriate to benchmark the U LTS and Current RHEL versions as they share a similar market space that is targeted toward http/db/fs.

I'm not sure, packages are frozen before the final version is rolled out.

so, why if the package versions are still in flux is it in a benchmark that is designed to assess consistency? Plus there is the added fact that while in beta generally full debugging is enabled which impacts performance.


I'm not out looking for a fight, just seeking equality, consistency and unbiased journalism. It's pretty much a known fact that people only read what's in front of them so to have relevant data hidden away in another post is a bit irresponsible for providing openly comparative data. To this it'd be great to see some SuSe results in the mix also in addition to Debian, but I understand that there are only so many hours in a day and limited resources.