Ubuntu 16.04 vs. vs. Clear Linux vs. openSUSE vs. Scientific Linux 7

[phoronix]

Phoronix: Ubuntu 16.04 vs. vs. Clear Linux vs. openSUSE vs. Scientific Linux 7

Here are some extra Linux distribution benchmarks for your viewing pleasure this weekend...

http://www.phoronix.com/scan.php?page=news_item&px=Ubuntu-XX-CLR-SUSE-Scientific

[HenryM]

Hmm. any guess as to why Ubuntu 16.04 sucks at C-Ray and Dolfyn? it's doing fine everywhere else.

[F1esDgSdUTYpm0iy]

Hmm. any guess as to why Ubuntu 16.04 sucks at C-Ray and Dolfyn? it's doing fine everywhere else.

More interesting to note is the dramatic regression there, for Dolfyn at least, between 14.04.4 and 16.04. Might be the novelty of 16.04. Or, maybe, the compiler version used to compile the packages of those 2 Ubuntu versions? As in, not distribution regression but compiler regression maybe?

[arjan_intel]

More interesting to note is the dramatic regression there, for Dolfyn at least, between 14.04.4 and 16.04. Might be the novelty of 16.04. Or, maybe, the compiler version used to compile the packages of those 2 Ubuntu versions? As in, not distribution regression but compiler regression maybe?

Hmm Ubuntu 16.04 and Clear Linux use the same compiler pretty much (gcc 5.3.0 + 5.3.1-maintenance branch)... (the others in the test seem to be using gcc 4.8.x)

[F1esDgSdUTYpm0iy]

Hmm Ubuntu 16.04 and Clear Linux use the same compiler pretty much (gcc 5.3.0 + 5.3.1-maintenance branch)... (the others in the test seem to be using gcc 4.8.x)

Well, Clear Linux does a lot of things differently; more very specific compiler flags, deep optimizations and what have you. So, that could explain the difference between 16.04 and Clear Linux. Not to mention those optimizations being Intel specific.

[arjan_intel]

Well, Clear Linux does a lot of things differently; more very specific compiler flags, deep optimizations and what have you. So, that could explain the difference between 16.04 and Clear Linux. Not to mention those optimizations being Intel specific.

that explains the small gap between Clear and the others... not the huge gap between everyone and Ubuntu 16.04
(and yes I understand people consider SSE4.2 Intel specific... AMD only has it for a few years by now)

[c2h5oh]

Michael Those results are very, very hard to read:

• there are no units for the measurements
• higher-is-better tests are mixed with lower-is-better tests, without an indication which is which
• because of the two above your only option is to rely on colors (green = best, red = worst)
• there are multiple tests with the same label (multiple compile, pgbench and redis rows)

[Michael]

Michael Those results are very, very hard to read:

• there are no units for the measurements
• higher-is-better tests are mixed with lower-is-better tests, without an indication which is which
• because of the two above your only option is to rely on colors (green = best, red = worst)
• there are multiple tests with the same label (multiple compile, pgbench and redis rows)

That table is just meant for a quick overview. Click the Openbenchmarking.org link to see all the data. The intention is the user clicks over and checks out the OB data. If you wish to make improvements to the table, patches welcome to the graphing code @ http://www.github.com/phoronix-test-...ix-test-suite/

[F1esDgSdUTYpm0iy]

that explains the small gap between Clear and the others... not the huge gap between everyone and Ubuntu 16.04
(and yes I understand people consider SSE4.2 Intel specific... AMD only has it for a few years by now)

Actually, I meant it as an explanation of why there is a difference between Clear and 16.04 if they're on the same compiler version. Regarding the Intel specificity of Clear Linux -- https://clearlinux.org/

Read it and weep. Not only is it implied, it is actually intended to be Intel specific.

[c2h5oh]

That table is just meant for a quick overview. Click the Openbenchmarking.org link to see all the data.

It doesn't provide a quick overview, because you have to "solve" it to understand it. It is faster to click the link and check it there.

I'll take a stab at improving the graphing code, but I haven't written any PHP in years, so the results might not look like regular PHP.