Page 1 of 1

Phoronix Benchmark

PostPosted: May 28th, '13, 01:09
by mithion
Hey a benchmark done by Michael Larabel at Phoronix just got published today pitting multiple Linux distros including Mageia 3. A lot of the benchmarks showed Mageia to be measurably slower than other comparable distros. Just thought I'd post the link and see what people thought about this and just to outline as something we should look into. Take a look here http://www.phoronix.com/scan.php?page=article&item=bsd_linux_8way&num=1

Re: Phoronix Benchmark

PostPosted: May 28th, '13, 20:01
by doktor5000
Does it also feel slower than other distros?

Also for the games, Mageia is one of the faster ones, sometimes even leading. Remember the rest are mostly synthetic benchmarks.
Maybe the one about LAME is not that good :(

Apart from that, the summary mentioning the different frequency-scaling governor is interesting, as that one is quite easy to fix,
and for fairness could have been egalised before running the test. But TBH the test probably had a different intent, so i'm fine with that and thankful for the results :)

Re: Phoronix Benchmark

PostPosted: May 28th, '13, 20:38
by mithion
I completely agree, Mageia doesn't feel any slower than other distributions. I just found it odd to see measurable differences in the synthetic benchmarks especially compared to Fedora which shares a lot in common with Mageia. The one thing I find odd is that mention about the cpu governor. I have a fresh mageia 3 install, and the tool cpupower reports the governor to be ondemand. I don't know where Michael got his performance governor from unless he changed it by hand.

Re: Phoronix Benchmark

PostPosted: May 28th, '13, 20:48
by doktor5000
Please reread his comment on that.

Re: Phoronix Benchmark

PostPosted: May 28th, '13, 21:12
by mithion
The way I understand it, Mageia defaults to performance governor when using Sandybridge EE processor? Am I getting this right? If yes, is that by design or a bug?

Re: Phoronix Benchmark

PostPosted: May 28th, '13, 21:27
by doktor5000
He doesn't mention explicitly which governor is used, and probably it doesn't matter if Ivybridge or Sandybridge, it's probably the default governor which is problematic.
Seems to have been posted already in viewtopic.php?p=34497#p34497 but not reported yet as a bug.

Re: Phoronix Benchmark

PostPosted: May 31st, '13, 19:47
by mithion
I may start a bug report on this. It seems the ondemand governor isn't working well anymore. I've switched to conservative governor as per the link you posted and I seem to be getting better performance and slightly smoother experience. Most likely people with moderately beefy systems won't notice a difference, but on my hardware limited system, I can definitely tell things are slightly faster even using the conservative governor. I saw a report on Phoronix where kernel devs were saying that even the Performance governor (where cpu is just kept at max clock all the time) may provide better power saving than ondemand.

Re: Phoronix Benchmark

PostPosted: Jun 3rd, '13, 00:44
by ghmitch
I have filed a specific bug report on the ondemand governor situation. I meant to do this previously and lost track of it, but it is now filed: https://bugs.mageia.org/show_bug.cgi?id=10405