Ubuntu 19.04 Is Offering Some Performance Improvements Over Ubuntu 18.10, Comparison To Clear Linux

[phoronix]

Phoronix: Ubuntu 19.04 Is Offering Some Performance Improvements Over Ubuntu 18.10, Comparison To Clear Linux

With the Ubuntu 19.04 "Disco Dingo" release less than one month away, we are getting ready for rolling out more tests of this next six-month installment to Ubuntu Linux. For those curious about the direction of Ubuntu 19.04's performance, here are some very preliminary data points using the latest daily state of Ubuntu 19.04 right ahead of the beta period. Tests were done on a high-end Intel Core i9 9900K desktop as well as a Dell XPS Developer Edition notebook when comparing Ubuntu 19.04 to Ubuntu 18.10 and also tossing in Clear Linux as a performance reference point.

http://www.phoronix.com/vr.php?view=27704

[treba]

Hm the MotionMark Firefox performance is really interesting (especially after https://www.phoronix.com/scan.php?pa...ebRender-67-68). What's happening there? It's surely worth some investigation.

[eydee]

Non-Intel distros should be benchmarked with all Meltdown patches off, and Clear Linux should be benchmarked with all of them on, set to highest level if there are levels. It's only fair like that, as Intel brought all that upon us.

[Flaburgan]

I'm wondering how ClearLinux set up Firefox and Selenium. As someone who launch many CI every day at work, I'm very interested if I can tune the setup and see it finishing quicker.

[Michael]

I'm wondering how ClearLinux set up Firefox and Selenium. As someone who launch many CI every day at work, I'm very interested if I can tune the setup and see it finishing quicker.

Clear doesn't have Selenium (or AFAIK) itself. PTS is using generic x86_64 Selenium Firefox/Chrome binaries and installing selenium support for Python3.

[andyprough]

Or, you could just turn off all of Intel's Meltdown/spectre mitigations and probably run circles around Intel's Clear Linux.

In an Intel (performance hacks) vs Intel (mitigation hacks) battle, disabling all Intel hacks is the likely winner.

[Flaburgan]

So if the binaries are the same, where does the perf difference comes from, in your opinion?

[elldekaa]

Nice to see the performance improvement between the two platforms, I am about to buy an core i9 9900k based desktop to replace my dell XPS15 9560 for programming: compiling LLVM/Clang is always a pain with *only* 4 cores.

[schmidtbag]

Non-Intel distros should be benchmarked with all Meltdown patches off, and Clear Linux should be benchmarked with all of them on, set to highest level if there are levels. It's only fair like that, as Intel brought all that upon us.

How exactly is showing the performance of a non-recommendable configuration a fair representation? Since (as far as I'm concerned) Clear is using the mitigations, really all that benchmark would show is misleading results.

[edwaleni]

I will have to go back and look, but it appears Clear Linux is doing better in VP9 with the latest roll.