The Fastest Linux Distribution For Ryzen: A 10-Way Linux OS Comparison On Ryzen 7 & Threadripper

[untore]

including void without the musl target is not very interesting. Alpine does not count because its support for desktop usage is mediocre.

[gregzeng]

Clear OS is Intel's pride & joy for Linux.
https://01.org/clearlinux > Clear Linux* Project for Intel® Architecture

> DESCRIPTION
> The Clear Linux* Project for Intel® Architecture is building a Linux OS distribution targeted for various cloud use cases.
> The goal of the Clear Linux OS is to showcase the best of Intel Architecture technology, from low-level kernel features to more complex use cases that span the entire operating system stack.
It is difficult to install for most users, even though it is "standard" for a RPM-based operating system. First problem, is which files to install? On my Intel 64 bit system:
(1) Clear-17470-installer.iso.xz, 533 MB (559,573,460 bytes)
(2) Clear-20570-installer.iso.xz, 192 MB (201,429,560 bytes)
(3) Clear-20570-live.img.xz, 273 MB (286,367,256 bytes)

(1) would seem to be the choice, since it seems to have applications for a desktop operating system. The inside-file is: "Clear-17470-installer.iso". 558 MB (585,105,408 bytes). I have yet to install and test this.

Clear OS has been discussed in an older Phoronix forum thread. https://www.phoronix.com/forums/foru...n-gold-systems

I updated my message there, but repeating some relevant parts here:

Originally posted by Michael View Post
gregzeng 2-bit Joe Clear Linux has the GNOME desktop available currently, previously Xfce.


Yes. Distrowatch has it wrong again. https://distrowatch.com/table.php?distribution=clear
> " ... Desktop: No Desktop ... "

Checked Clear OS web site. It is a RPM-based distribution.
https://download.clearlinux.org/curr...4/os/Packages/

[sarfarazahmad]

If I ran(if i even could do that) clear linux's kernel on Arch would I get any benefits ?

[Templar82]

Clear Linux is very impressive, but I had in my head that it was a minimal server OS only. I will have to check it out after seeing this.

Personally I am using Void on my Ryzen PC.
As I don't do "Server Things" I'm happy with its overall performance, although it was interesting to see it fall over on a few of the tests.

A very good article.

[cj.wijtmans]

FDO / PGO is great stuff. I use GCC's -fprofile-generate on some code libraries I build at work and it's a 10 - 30% performance improvement. It's a time consuming build that requires a full test-set run and then a full rebuild.

That's why AutoFDO is really great. You can just run apps normally while collecting profile data and there's no need for automated test sets and slow compiler options (running code built with -fprofile-generate is slow).

I wonder if source code releases should not just come with a profile so it saves a lot of runs for a lot of people.

[Zan Lynx]

I wonder if source code releases should not just come with a profile so it saves a lot of runs for a lot of people.

That would be really interesting and could work.

Some problems I see is if the build systems are different in any way, it would screw things up. Different library versions, different compiler versions or different compiler options. Building for the "native" CPU instructions is very popular, but if the provider of the profile used a Ryzen and the builder is on Coffee Lake Intel, who knows how things may change?

[cj.wijtmans]

That would be really interesting and could work.

Some problems I see is if the build systems are different in any way, it would screw things up. Different library versions, different compiler versions or different compiler options. Building for the "native" CPU instructions is very popular, but if the provider of the profile used a Ryzen and the builder is on Coffee Lake Intel, who knows how things may change?

I think some kind of standard for source code releases so everyone can have some standard performance benefits. Not sure how the profiles work so i dont even know if it can work that way. Also if the librarie dependencies are so different that it would make different profile why would it compile in the first place. Though amd vs intel is a good point.

[Ragool]

Ok I want to use ClearLinux for my shiny new Ryzen 2400G build! But why was there not a Clearlinux benchmark for the OpenGL tests? Are these GPUs unsupported by Intels brand of Linux?

[Michael]

Ok I want to use ClearLinux for my shiny new Ryzen 2400G build! But why was there not a Clearlinux benchmark for the OpenGL tests? Are these GPUs unsupported by Intels brand of Linux?

Clear Linux doesn't currently ship with Radeon enabled with its Mesa build, that's why those tests weren't done. As far as whether Intel is planning to enable it or can request it, not sure.

[rogerdpack]

WAT? You are very confused. 32-bit distro builds do not affect the performance of 64 bit distro builds. 64-bit distro's are only held back by the fact that they sometimes limit themselves by targeting CPUs of the first gen x86-64 processors. Eliminated their 32-bit companions wouldn't magically make any of these distros any faster.

If x86 magically disappeared tomorrow, no distro would get any faster, but you would end up with buggier software because building software on diverse platforms helps catch bugs. It also makes software more portable, allowing projects like Raspbian and RaspBSD to exist with a minimum amount of work. You'd also give up the ability to spin up low memory VMs as x86-64 requires about 30% more RAM than identical x86 programs. That said, I'd love to see x32 ABI make a comeback, if x86 ever starts to fade.

I assume he's referring to the fact that gcc by default targets "-march=generic" which is some generic (oldish) processor. I think with gcc 4.8 they finally are targeting something more modern. But yeah allowing 32-bit versions doesn't actually *cause* that, unless it somehow subtly affect what -march=generic means within gcc for 64 bit, I guess...