DragonFlyBSD 6.0 Is Performing Very Well Against Ubuntu Linux, FreeBSD 13.0

[onlyLinuxLuvUBack]

Great benches !

[MadCatX]

The superior performance despite difference in manpower is easily explained: Superior architecture.

Linux has hit a wall, and won't magically get any faster overnight, no matter how much manpower is thrown into it.

Dragonfly absolutely has plenty of headroom. Imagine if the development team grew at all.

I doubt this is the case. I suspect that there is something going on with frequency scaling. Other tests Michael has done lately on the Intel platform show a huge performance difference between the powersave and performance mode of the intel_pstate governor. Maybe Intel messed something up in the recent kernels because I don't remember the performance gap to be this huge in so many tests. That's why I think that having AMD benchmarks for comparison would be interesting because they use a different governor.

[ayumu]

I doubt this is the case. I suspect that there is something going on with frequency scaling. Other tests Michael has done lately on the Intel platform show a huge performance difference between the powersave and performance mode of the intel_pstate governor. Maybe Intel messed something up in the recent kernels because I don't remember the performance gap to be this huge in so many tests. That's why I think that having AMD benchmarks for comparison would be interesting because they use a different governor.

While that'd explain Ubuntu's lackluster performance if true, it doesn't matter so much in practice; Ubuntu, due to its popularity, does work well as representative of Linux... and it is behind.

And so is FreeBSD, for that matter.

[MadCatX]

While that'd explain Ubuntu's lackluster performance if true, it doesn't matter so much in practice; Ubuntu, due to its popularity, does work well as representative of Linux... and it is behind.

If it's a regression, it should be investigated and patched. I doubt it's a Ubuntu-specific issue, Fedora 34 did just as badly as Ubuntu 21.04 in Michael's recent tests. The performance gap just seems too wide to be explained just by a different task scheduling implementation and definitely asks for a closer look at what is the Linux kernel doing here.

[ayumu]

If it's a regression, it should be investigated and patched. I doubt it's a Ubuntu-specific issue, Fedora 34 did just as badly as Ubuntu 21.04 in Michael's recent tests. The performance gap just seems too wide to be explained just by a different task scheduling implementation and definitely asks for a closer look at what is the Linux kernel doing here.

It should, sure. Still, while the "how it happened" is something worth investigating, the "what happened" is fact, and what the story is about:

For the most common hardware (Intel) and the most common distribution (current Ubuntu), Linux is now behind Dragonfly BSD.

[Linuxxx]

It should, sure. Still, while the "how it happened" is something worth investigating, the "what happened" is fact, and what the story is about:

For the most common hardware (Intel) and the most common distribution (current Ubuntu), Linux is now behind Dragonfly BSD.

Cool story, BSDbro, but that doesn't change the fact that all the BSDs are still horribly inadequate as a general desktop OS!

[ayumu]

Cool story, BSDbro, but that doesn't change the fact that all the BSDs are still horribly inadequate as a general desktop OS!

You mean the Year of Linux Desktop came and I was never made aware of it?

[Volta]

ayumu

Nice fairy tales. Change Ubuntu governor to performance and say goodbye to bad results. Linux is far from hitting a wall and currently it's beyond competition's range. P.S. Fedora 34 also uses schedutil, so expect more *BSD lead. We'll see how this looks on AMD.

[jb.1234abcd]

The Stress-NG test: context switching is a disaster for Ubuntu linux or Linux kernel.

https://www.phoronix.com/scan.php?page=article&item=corei9-freebsd13-dfly6&num=4

Could anybody try to explain that ? Or perhaps the test could be repeated for other general distro like Fedora or Arch ?

I googled for any discussions - this one sheds some light, but there is probably more as this is work in progress:

https://stackoverflow.com/questions/12111954/context-switches-much-slower-in-new-linux-kernels

[Volta]

The Stress-NG test: context switching is a disaster for Ubuntu linux or Linux kernel.

https://www.phoronix.com/scan.php?page=article&item=corei9-freebsd13-dfly6&num=4

Could anybody try to explain that ? Or perhaps the test could be repeated for other general distro like Fedora or Arch ?

I googled for any discussions - this one sheds some light, but there is probably more as this is work in progress:

https://stackoverflow.com/questions/12111954/context-switches-much-slower-in-new-linux-kernels

The most obvious culprits that comes to mind are vulnerabilities mitigations.