If that’s true, thus a noob like me should switch to another distro like debian or fedora?
by the way, I’ve found Leap very interesting and easy to use.
Well, they tested a beta, it might be interesting if you try it now, then
try those other distros listed in that benchmarks and compare it for yourself.
Actually I did try some of those some months ago, but I’ve found Leap wonderful,
Leap is very stable, fast and awesome, and that’s why I got puzzled when I saw those benchmarks
I recall that this was discussed, back in the day, when Phoronix published that. One thing that you should keep in mind is that, presumably for reasons of the amount of work to do, Phoronix does their testing in ‘default’ conditions, and, as the defaults are different for Leap than for the other distros tested, that’s one of the things that gets tested.
I don’t know how much can be laid directly at the door of BTRFS, but that will be a big lump. Don’t like it? Change from BTRFS and live without the benefits it can bring!
The change between Leap beta and Leap final was substantial.
There are problems with some Nvidia cards, but that’s probably not a legitimate complaint about Leap. It’s either a problem due to an inadequate driver from Nvidia, or due to Plasma 5 issues.
My only “issue” with Leap, is that startup time is longer than it should be. But I don’t reboot all that often, so it’s not a big issue for me.
They are also doing all tests on a “default” install (of the Leap Beta at that).
This presumably means plasma 5 circa August 2015 with compositing turned on?
One thing I found with the move to Plasma 5 is that compositing it has a much bigger overhead than before. I don’t think RAM is much different so it must be to do with taking up more GPU cycles.
I ran glmark2 and used to get scores in the mid 5000’s on my KDE4 desktop and this didn’t differ much whether composition was off or on.
Now in Plasma 5, if I leave compositing on I only get 1500 or so. Turn it off and run again, magically back to pre-plasma5 levels.
Needless to say I went directly into KDE’s settings and set it to stop compositing when a full screen app is in focus.
To answer the ‘Do you think that it’s right?’ question, yes (and also no). I do think that the results published are absolutely the results that were measured, so in that sense, they are right.
There is a more difficult question ‘…and are they relevant…’ to something or another, and that gets messier.
Firstly, there are some more up-to-date results that you should look at (link1 and link2), and those are significantly different (they measure different things, so they get different results). I was going to try to summarise all the results, but it is too much like work and probably doesn’t tell you anything that you really want to know. As high/low lights:
Intel’s ClearLinux is fast, but they are (almost) insanely adventurous with compiler flags to get there
In places, there are bigger differences between Leap and Tumbleweed than I’d have expected, but the compiler versions are quite different (GCC 4.8.5 vs GCC 5.1)
Just to illustrate how careful you have to be with these things, the second and third tests here show PC-BSD as nearly the fastest thing going, and quite slow, both in compile bench testing (depending on which exact test that you look at)
Or, you could look here, where, in the NAS Parallel Benchmark, Tumbleweed is one of the fastest and Leap is one of the slowest (on a test with a low-ish total spread). Or, the next test (MAAFT), in which Leap is second fastest, and Tumbleweed is about half way down
Next up is the ‘Himeno’ behchmark, which I’d have expected to be down to math (maybe it does lots of i/o), and you’d have expected that would be one of the areas where the aggressive flags of Clear would have given it the edge, but it is solidly mid-table (Tumbleweed is second and Leap is last here, but again the spread isn’t very big).
And, just for one last cut at the same theme, the next test is an Apache compilation test (not that you’ll do that often), and Leap wins that and Tumbleweed is just below mid-table (and that Clear Linux that was crushingly fast, earlier on, is now plumb last)
So, you really have to be very, very, careful with benchmark results. It is easy to just read one or two and not understand what is actually being measured and come to quite unjustified conclusions.
Yes, the best benchmarks are “real world” performance.
I’m playing Bioshock Infinite now with settings maxed out, including the FOV, and it’s running smooth as butter on my Leap 42.1 install.
I could not say that just a month ago when I was running Kubuntu 14.04.