According to these charts it was
The other Distros minus clear linux did worse across the board yet every single one was ranked higher?
Have you read what iâve written? 
You can see all 210 individual tests.
-
Note that first / last places is calculated from 210 number of all tests. 53 / 210 first places shows nothing about how
performed in other tests. -
If you want something to be high in geometric mean - extremes (first / last places) doesnât matter that much, unless if itâs more than 50% 1st places or higher than middle result through all data.
All of them did more consistent - that is what geometric mean shows (their results was less volatile than across extremes)
P.S. That being said - iâm not sure that geometric mean is best technique to show average / smooth results for distro tests, because a lot of people miss itâs meaning and might get mislead by it.
1 Like
Please keep in mind that phoronix uses different clock speeds in that test:
CPU is Intel Core i9-13900K but clock speed is 4 GHz for Centos, Debian and Ubuntu and clock speed is 5.50 GHz for Clear Linux, EndeavourOS and Fedora.
gcc versions used are also different, compiler flags are also different.
That makes any comparison of the results impossible.
1 Like
Yes, that particular test doesnât even try to be fair for settings or software used, they just measure âout of the boxâ user experienceâŚWeird idea, but whatever
but if you go through the tests page by page, EOS was above fedora in almost every single test.
I think im just expecting something more along the lines of Harmonic mean than Geometric
The harmonic mean results seems more sensible but even still, going through the results 1 by one shows EOS consistently leading Fedora 37
If anyone is curious, here are the Fedora vs EOS results going through the tests 1 by 1. Several of the tests they both failed to run so that actual test total is 206 out of 210
These numbers represent wins vs the other
EOS - 129
Fedora - 77
Overall anything Gaming related EOS will win pretty much across the board, workstation tasks its a toss up but Fedora is a workstation distro so makes sense
EDIT:
Ubuntu vs EOSâŚbig oof for ubuntu this round
EOS - 153
ubuntu - 57
CentOS vs EOS, it was close for a while but blender and some other tests took the crown for EOS
EOS - 129
Cent - 76
EOS vs Debian, another close one till Blender and a few other benchmarks Debian was ahead or close to even with EOS
EOS - 115
Debian - 95
EOS vs Clear, i think i might have slightly fudged Clear linux as being higher than EOS by too much. They traded blows so much it got hard to keep track. Theyre so close though who really cares?
EOS - 104
Clear - 106
vs any given Distro on this test vs EOS you will find EOS coming out on top most of the time minus vs clear linux where they trade blows across most tests.
1 Like
That one is the most impressive to me, i would never thought that Debian come even close to EndeavourOS - thatâs very impressive on Debian part.
Well, donât forget that test is not fair in the first place, as @mbod have pointed out, also winning test relatively doesnât tell you by which margin in absolute numbers (some test in absolute numbers for Clear Linux are miles away from anything else)âŚ
Actually when i come to think about it, maybe they calculate Geometric mean by using absolute numbers of each test?
Yeah when definitively Clear won a benchmark, it usually did so by miles ahead of everyone else. I dont think anyone else had as big a wins as Clear but Clear also had some pretty horrible losses in some cases. I can also go through each benchmark and tally who won what by what margin
I dont have time at the moment but later i will go through the benchmarks 1 by 1 and figure out the average deviation from distro to distro across the data and note any significant outliers along with noting what may be causing significant swings in any given result,etc.
Im not sure the clock speed was actually locked to these speeds as the processor info may be simply read out differently. There would be much larger deviation in performance if the CPU was locked to 4ghz as the performance vs 5.5ghz is MUCH lower. CentOS is listing that 4ghz but its results are pretty consistent with everyone else who is boosting so i feel that bit of info is wrong.
I think this is just CPU info being displayed differently not an actual disparity on clocks, for instance depending on how i check CPU frequency on my 5600x you would think it never reaches boost clocks (but it does)
EDIT: If i had extra drives to do some testing i would probably do this suite of tests along with logging frequency/temperature across the range of tests and ensuring theyre all configured as close as possible. Unfortunately i dont have the storage for doing this kind of testing lol