Even With A $199 Laptop, Clear Linux Can Offer Superior Performance To Fedora Or Ubuntu

[elatllat]

Has anyone actually tested the max RAM upgrade yet?

Popularity is making it more pricey;

https://youtu.be/0hMdQAjy43A

[r08z]

Ive set my makepkg.conf to -O3, -pipe, and installed my custom slimmed down kernel. Any more suggestions on how I can get more performance on my desktop??

Make sure you also have -march=native to tune to your CPU only.
Also visit their github page at: https://github.com/clearlinux-pkgs and search for packages that you care about performance for and apply any performance related patches only - make sure you understand what those patches do because some patches can break your system since you're not using Clear Linux... and in the .spec file you can see the standard flags used for each package including it's -m32 versions.

Generally, for any self compiled software you really need to consider how much you care about security vs raw performance. Not every package needs to be recompiled, but when recompiling a package, make sure you also recompile it's dependencies. You really can create a general system-wide speed up by recompiling gcc without assertions --disable-checking --disable-bootstrap and disabling it's security defaults by removing --enable-default-pie and --enable-default-ssp. And recompiling glibc with -march=native -O3 since these two pieces of software have massive influence on C/C++ programs.

Read read and read up on what settings and configurations mean and get acquainted with gcc and llvm switches because they put the software together and affect your experience.

[enigmaxg2]

Clear Linux looks like a very nice option. Only a couple questions: it does include up to date packages? (the ones which I'm most interested in are the Mesa ones, for gaming).

[cyclistefou]

Why am I the only one commenting about the very... bad title of this article? Michael, you're doing such a great job on Phoronix, but this title... come on.

Why would the ranking of Clear Linux vs other distributions, be different based on the price of the machine? If a distro is faster because of leaner / more optimized code, then it tends to be faster whatever the price tag of the computer it is running on... of course unless there are accelerators or specific features at different price tags that can shift performance for distributions taking advantage or not, of those accelerators...

Well, apart from this criticism, congrats for the wonderful job on Phoronix.

[AdamOne]

The reality is generic distributions don't tune for raw performance. They conservatively tune for various ranges of compatibility and security settings instead, hopefully without sacrificing too much performance. That's not an unreasonable goal when you have no control over what hardware your software is going to be deployed on.

Ive set my makepkg.conf to -O3, -pipe, and installed my custom slimmed down kernel. Any more suggestions on how I can get more performance on my desktop??

[AdamOne]

Does it have trash like SELinux and AppArmor disabled? They affect performance.

Minimal? It's even 100% slower in schbench, 90% slower in stress-ng, 12% slower in Apache:

This is why I generally trust the overly critical posts more, they end up being correct more often in the end, I find.

[sofar]

Does it have trash like SELinux and AppArmor disabled? They affect performance.

Neither: https://github.com/clearlinux-pkgs/l...r/config#L7979

[sofar]

tried to determine what version of the kernel and Mesa was currently shipping.

https://github.com/clearlinux-pkgs/linux/releases

https://github.com/clearlinux-pkgs/mesa/releases

[Etherman]

Clear Linux keeps kicking but. It's strange that they don't yet turn it into a commercial project yet by offering support services for it. It could be useful in specific cases where the best performance possible is required

[tinfoil]
Maybe it is a commercial project to shrink server CPU demand for a while. 🤔
[/tinfoil]

[Volta]

Did you not look at the recent benchmarks? The a̶f̶f̶e̶c̶t effect of SELinux is minimal.

Minimal? It's even 100% slower in schbench, 90% slower in stress-ng, 12% slower in Apache:

https://www.phoronix.com/scan.php?page=article&item=fedora-31-selinux&num=2