DragonFlyBSD vs. FreeBSD vs. Ubuntu 20.04 On Intel's Core i9 10900K Comet Lake

[Royi]

If the goal is to test OS then everything but the OS must be equal.
Running different code from different compiler means we are testing the compiler mostly.

[aht0]

He is testing all the OSes on their default settings. Thats the only way to get some meaningful results. You may take it from there and personally tweak your favorite OS however you prefer but it doesn not reflect the results you'd be getting when comparing OSes Out Of the Box.

[tildearrow]

You should tweak the colors in the pie chart a bit. It's kind of hard to see.

[Michael]

You should tweak the colors in the pie chart a bit. It's kind of hard to see.

It's a never-ending goal, unfortunately, and something that I've been playing with for over a decade in trying to come up with a better automated color picking algorithm for PTS graphing code... Usually I end up hitting all sorts of corner cases in the end whether playing with auto color generation or trying to largely rely upon a static list of possible colors and then playing with alterations from those static pre-sets, but the current implementation is the best at present for most result files while yes this particular result file ended up getting pulled some odd color matches. I am a bit short on ideas for coming up with a better approach but if any color programming experts have other ideas happy to evaluate them as I am rather bad with colors/UI aspects.

[Royi]

He is testing all the OSes on their default settings. Thats the only way to get some meaningful results. You may take it from there and personally tweak your favorite OS however you prefer but it doesn not reflect the results you'd be getting when comparing OSes Out Of the Box.

We know the policy. It doesn't mean it makes sense.
Compilers are not part of the OS. They are bundled with the OS. Just like he installs other applications to test performance he can, where possible, install the same compiler and use the same compilation configuration. Otherwise he mostly tests the compilers and the chosen flags.

Moreover, we're talking about Linux. Users of Linux are usually more knowledgeable. I can see the logic testing the OS in its default on Windows / macOS, yet in Linux? This policy doesn't make sense. Use the same compiler, use the same flags, let's see the difference in the OS itself.

[aht0]

We know the policy. It doesn't mean it makes sense.
Compilers are not part of the OS. They are bundled with the OS. Just like he installs other applications to test performance he can, where possible, install the same compiler and use the same compilation configuration. Otherwise he mostly tests the compilers and the chosen flags.

Moreover, we're talking about Linux. Users of Linux are usually more knowledgeable. I can see the logic testing the OS in its default on Windows / macOS, yet in Linux? This policy doesn't make sense. Use the same compiler, use the same flags, let's see the difference in the OS itself.

Compilers also have to use OS's scheduler and capabilities of IO. Where particular idiosyncracies do affect the result.

Users of Linux are usually more knowledgeable - debatable. It's distrospecific. Arch or Gentoo users are very knowledgable, no doubt. Then comes next Linux user who goes to distrohopping because some minor GUI doo-dad is not available on distro A but is in distro B and he/she will prononunce whole distro A "sucks" because of that lack. There are tons of Linux users who can't conf shit through text editor and who's skill is limited to adding/removing custom repos, fucking up their OS and instead of fixing the problems, just re-installing from scratch. THAT ain't "knowledge", it's "baby steps". Average BSD user is not aftaid of text editor or console and is able to set up their desktop or server from scratch. Average Linux users across most distros ain't. "Shitty pseudo-graphical installers" do create kinda barrier where seekers of bling and GUI doo-dads go to Linux. Professionals among Linux users are minority. Arch and Gentoo users among Linux distros are minority. Not to speak of Slackware users. So your claim is debatable.

[uid0]

Michael, was powerd running on the FreeBSD systems? Without it, Intel CPU cores will not clock higher than the base frequency.

[aht0]

powerd prolly could also use configuration arguments in /etc/rc.conf

That'd go outside 'default' setup tho

[uid0]

powerd prolly could also use configuration arguments in /etc/rc.conf

There is no need to tweak anything; powerd just has to be enabled in /etc/rc.conf with powerd_enable="YES" -- that's it.
This should make the CPU select the turbo mode (a frequency among power states with '1' in the end, e.g. 3601).

[aht0]

There is no need to tweak anything; powerd just has to be enabled in /etc/rc.conf with powerd_enable="YES" -- that's it.
This should make the CPU select the turbo mode (a frequency among power states with '1' in the end, e.g. 3601).

Doesn't push my Ryzen 5 3600X into turbo frequencies without args. Or didn't look like it did.