Clear Linux vs. Ubuntu 16.04 On The Xeon E3-1280 v5 Skylake Workstation

[phoronix]

Phoronix: Clear Linux vs. Ubuntu 16.04 On The Xeon E3-1280 v5 Skylake Workstation

With Clear Linux continuing to outperform other Linux distributions on Intel hardware, I was curious to see how the Intel OTC Linux distribution was performing when trying it with one of the new Xeon CPUs at our disposal for testing...

http://www.phoronix.com/scan.php?page=news_item&px=Clear-Linux-vs-1604-E3-1280v5

[Sloth]

What were the mount options for the postgres test?

[tigerroast]

The Xeon processor assortment looks like a candy bar for a fucking Terminator.

[Michael]

What were the mount options for the postgres test?

the filesystem mount options are shown on the system table page.

[kgardas]

the filesystem mount options are shown on the system table page.

Michael, unfortunately they are not. You just list ext4 there -- or am I completely blind.

In addition doing pgbench on such different drives like Triton150 and Samsung 850 -- btw, was is Pro or EVO? -- is simply not correct.

[intelmatt]

Michael- Why is the ubuntu machine clocked at 4.0 GHz but the Clear machine is at 3.7 GHz? Would that have any effect on the results?

I'm referring to the comparison chart here: http://openbenchmarking.org/result/1...GA-CLEARUBUN61

[Linuxhippy]

I really hope more distributions will use GCC's AutoFDO optimization in the future. Profile-guided optimization often leads to *way* better generated code, and with AutoFDO GCC makes the whole process relativly painless - yet most distributions simply compile with "-O2 -march=i686" and claim to be efficient.

A good example is the official firefox build distributed by mozilla, which outperforms distribution-specific binaries often by ~10-30%. Simply because mozilla is compiling with profile-guided optimization, while distribution-binaries are "-Os".

[DDF420]

Michael- Why is the ubuntu machine clocked at 4.0 GHz but the Clear machine is at 3.7 GHz? Would that have any effect on the results?

I'm referring to the comparison chart here: http://openbenchmarking.org/result/1...GA-CLEARUBUN61

Ubuntu is also using intel_pstate powersave whilst clear OS is using acpi-cpufreq performance. Guess that's what happens when you pull and old benchmark and compare to new one

[Michael]

Ubuntu is also using intel_pstate powersave whilst clear OS is using acpi-cpufreq performance. Guess that's what happens when you pull and old benchmark and compare to new one

It's not comparing old and new.... Both these tests are new. P-State is used by Ubuntu where as Clear Linux still uses CPUfreq. But yes that's why the frequency diff is shown since pstate advertises its max frequency as turbo rather than base via sysfs.

[Michael]

Michael, unfortunately they are not. You just list ext4 there -- or am I completely blind.

In addition doing pgbench on such different drives like Triton150 and Samsung 850 -- btw, was is Pro or EVO? -- is simply not correct.

Hmmm, I see. Since the SQlite and pgbench were the only disk related tests run, within their test profile XML they are set as 'System' rather than 'Disk', but currently the code is only showing the FS mount options if one of the tests in that result file has a 'Disk' test. So I guess I should also show the mount options if there is a 'System' test too. But anyhow, you can look at the other recent Clear Linux and Ubuntu results to see the mount options as a footnote on the system table.