Ubuntu 18.04 LTS vs. Fedora 28 vs. Clear Linux Benchmarks

[Michael]

Clear Linux uses mq-deadline? Well, nice to see that I'm not alone in using it.

For SATA devices. For NVMe, last check used Kyber.

[alex79]

What we can see is that i7 8700k is much better and cheaper than R7 2700X...

[arjan_intel]

There are several reasons:

- Fedora 28 has just enabled some power management improvements (such as different drive power management policy), so those are possibly responsible for the differences in IO

At least Clear Linux has the same drive power management etc as well...

[Michael]

So, to be fair, I would suggest to repeat the benchmark with the following in mind:

- Test both with the same graphics system (either XOrg or Wayland).
- Use latest kernel and latest CPU firmware on Ubuntu as well, so we see the perf penalty of the Meltdown/Spectre mitigations there as well.
- Make sure power management is configured the same way on both systems

With the above in mind, my guess is that the difference will be substantially less, but of course due to the more security focused GCC flags Fedora would still be marginally slower.

As others have alluded to and I have mentioned dozens (or hundreds now?) of times in other threads, when doing a distro comparison it really only makes sense doing the defaults. Soon as you start tweaking things -- especially the notion of changing the kernel as well as the bundled firmware -- you really are running down a slippery slope of where to stop and where does it start/end for what is considered comparable to a given distribution/OS. It's just not a matter of say checking something different at the installer like a 'performance mode', but there are endless tunables.

Granted, I do run specialized articles where say comparing kernel versions or seeing what it's like building an Ubuntu kernel with Clear's Kconfig file, but for the purposes of doing a distribution/OS comparison, it doesn't make sense tweaking a bunch of things -- especially when most users probably do not go through all of those steps and all of those choices are choices knowingly made by each of the OS vendors.

[Vistaus]

There are several reasons:
- Fedora ships a later kernel with the latest firmware, both of which contain Meltdown/Spectre mitigations, thus it is not surprising that it is significantly slower on Intel systems.

So? I'm using 4.16 with the latest firmware on Solus and my laptop doesn't feel any slower than with previous kernels (4.9-4.15). And my system is Intel Broadwell, in fact, it's a 2015 Celeron so it's low-end. So if I'm not even noticing any difference with 4.16 on such low-end hardware, then 'nuff said.

[Royi]

Michael , What about checking other aspects of the system?
Like the memory consumption, CPU consumption at idle, Battery Life, etc...

Those are also factor to measure when reviewing a system.

[BaronHK]

Which, by the same token, is why Spectre/Meltdown should behave better on Fedora. Again, it's all about what a distro offers once installed. Obviously once you get into installing your own packages and changing the default settings, you can make any distro do anything you want, but that's not the point of a test like this one.

No, I think you missed what I said, so I'll back the truck up.

Ubuntu users have no protection from Spectre/Meltdown at the firmware level at all because they're running whatever CPU firmware their BIOS has unless they took an extra step to install intel-microcode. That's also not good because Intel fixes other problems with new firmware as well. I get _a lot_ of kernel messages saying "firmware bug" and "please update your firmware" in Ubuntu. Why don't they ship the fix? It can't be because they're opposed to proprietary software. They're trying to load up your entire hard drive with proprietary software if you take a look at their software center lately.

Phoronix benchmarks mean nothing because they don't test the same configuration. He should have switched Fedora over to GNOME on Xorg and installed the updated firmware into Ubuntu, tested the same kernel version and power management settings, and then Ubuntu's advantage would very likely disappear.

Benchmarks from Phoronix are really lousy clickbait. Michael has good reasons to post clickbait and especially stuff that spans multiple pages, because like two-thirds of the stuff he (tries to) loads on each page is ads/trackers. I tested the load time of the front page with and without Fanboy's Ultimate Adblock List loaded up, and it's 9 seconds without and 2 seconds with. Ouch.

[arjan_intel]

Phoronix benchmarks mean nothing because they don't test the same configuration. He should have switched Fedora over to GNOME on Xorg and installed the updated firmware into Ubuntu, tested the same kernel version and power management settings, and then Ubuntu's advantage would very likely disappear.

I will argue he tested the exact same configuration: Same hardware and 100% OS defaults.
That's the only way to compare linux distros to be honest. Sure you can make Fedora look like ubuntu (or other way around) by replacing half the content and settings, but then you're no longer comparing the operating systems.

[dungeon]

I will argue he tested the exact same configuration: Same hardware and 100% OS defaults.
That's the only way to compare linux distros to be honest. Sure you can make Fedora look like ubuntu (or other way around) by replacing half the content and settings, but then you're no longer comparing the operating systems.

Better to say 99% OS default since it is impossible to bench default in some cases, for example vblank is on by default for GPU drivers, etc... you see, as soon as you disable sync that is not default anymore

So to be able to bench something, some interevention is *really* needed

[perpetually high]

He should have switched Fedora over to GNOME on Xorg and installed the updated firmware into Ubuntu, tested the same kernel version and power management settings, and then Ubuntu's advantage would very likely disappear.

lol, sounds like you're saying Fedora made a mistake shipping with Wayland as the default instead of Xorg. Your words not mine. Let's just accept they're both great desktops, and you could argue Ubuntu 18.04 LTS won this round, no shame in that. It's a solid release.

And if it makes you feel better, we're fighting for second place. Clear Linux is ahead (if you wanna talk benchmarks)

FYI: If you're gonna attempt to diss Michael, keep in mind you're on _his_ website, not vice versa.