AMD Ryzen 7 Performance On Windows 10, Windows Server 2016, Six Linux Distributions

[gregzeng]

Unexplained in these tests, and some comments: two beta-versions of the operating systems were used, because the very final release was not available, at the time of this report. "Fedora 28 Beta" & "Ubuntu 18.04 LTS" are both loaded with beta-software tricks, to assist standard beta testing.

Like all final release operating systems, these will auto-update, if given access to the internet. This auto-update will allow the latest "authorized" updates only, at the time of installation. The beta-program stuff will be removed, if all goes as expected. However the very, very latest updates will not be added, until these have been officially "approved".

Linux kernels are release, freshly updated, about every week. Ubuntu makes their version of these compiled-kernels available seconds after the source code is released by "The Linux Foundation". All the other Linux operating systems cannot immediately & quickly install these latest kernels, which the Debian-based operating systems can do so easily & simply.

There seems to be very widespread ignorance about Linux operating systems. There are about five main areas of computer operating systems: IoT, mobile, desktop, server & cloud. Linux dominates only FOUR (4) of these five types of computer operating systems. In the desktop systems, Linux is a very tiny third-place, after Microsoft Windows, and Apple.

Officially "Ubuntu 18.04 LTS" will be soon replacing "Ubuntu 16.04 LTS". This operating system is the foundation of more Linux operating systems than any other "foundation" operating system. "Red Hat" seems more prominent in the "server" areas than the Ubuntu-based operating systems, for the moment. But most creators of operating systems have good reason to avoid "Red Hat" (RPM) as the starting base for their operating system.

One well known web-site, checked just now, claims in its "309 living" Linux distributions:
RPM-based: 42
Debian-based: 141
Ubuntu-based: 65

That particular web site does strange tricks, such as trying to hide the true Ubuntu-statistics. "Mint" (for example) as just one "system", when it is actually has many versions of itself, all different. Hidden behind the brand-names are versions for: IoT, "core", server, cloud and many possible desktop environments. If you want to hide the truth about Linux, it is better to not let industry outsiders know these truths about Ubuntu's dominant influence on Linux. Influence is more important in my opinion, that raw-financial-measures.

In terms of "innovation" & influence, there is still so much more to be done, before Microsoft Windows can be "defeated". Linux needs to develop further its display managers (Wayland, X.org, etc), its package systems (Snap, Flatpak, etc), and its poorly featured & poorly performing partition systems on the desktop, where Microsoft's NTFS is still dominant.

This Phoronix report shows the superiority of the Unix-derived operating systems (Apple, BSD-types & others excepted). Now we all wait for the very major coders (Microsoft, Google, Red Hat and the very many Debian-types) to focus on befriending the last frontier: the human desktop, imho.

[edwaleni]

The Windows Server 2016 H.264 performance appears to be off quite a bit.

I assumed that the WS2016 image that was tested included Windows Desktop Experience? Many default codecs used by MSFT are not available in the Core version. (non-GUI)

Also, MSFT changed the codec priorities in Server 2016 to support new vGPU functionality in RDP 10. They use AVC444 which uses H.264 for video and some unknown proprietary codec for text transport. To support transcoding performance AVC444 looks for a DX11 compliant GPU to support performance offload. This is supposed to allow better OpenGL performance as well.

I can only speculate that Windows Server could not locate one on the test host and reverted to software acceleration.

VNC users have noticed similar slowdowns in screen performance in WS2016 as well as some Plex users.

[RaskAr]

Nice to see win not taking wind, but *glups* what is going on with debian-testing UDP perf ?_?

[andyprough]

Maybe a distro should ship with defaults that are oriented to the real world?

Lots of people love openSUSE with Btrfs and XFS. And anyone looking for performance does not use them. Phoronix tests are performance tests. No sense testing it if you aren't going to use the standard performance setting. Us users are perfectly capable of running PTS, no reason for Michael to waste his time.

[Space Heater]

I wonder what happened on the PostgreSQL read+write benchmark that made Windows so much faster there when it was much slower in the read-only benchmark. Almost feels like Windows cheats with fsync on writes.

It's almost certainly a flawed test.

[arjan_intel]

Michael -

Because anyone who installs openSUSE with performance in mind uses EXT4, just like the other distros. Only takes a mouse click at installation.

It's not that your tests are unfair. It's just that they never give us a proper comparison of how our actual boxes will compete with the other distros.

Which is no problem, we can just run PTS ourselves if we like. But I agree that you shouldn't waste your time on a useless comparison. I'm only interested in real world, not default.

Maybe a distro should ship with defaults that are oriented to the real world?

[renkin]

Ubuntu has some debugging trash enabled. Furthermore, it has non typical/non optimal Linux desktop kernel configuration.

Fedora 28 isn't even released yet. If anything, Fedora would have more debugging enabled.

[F.Ultra]

Is that because server software cares about stability first and foremost? Boy I wonder.

Yes, and?

[carewolf]

Or switch to Debian stretch or testing. Ubuntu 18.04 is based on Debian testing. It also means you don't have to change your habits oo much ;-)

No if 18.04 was based on testing it would work instead of just bricking your computer. Ubuntu are unfortunately doing their own things and breaking stuff all the time.

[angrypie]

And yet we are many who run Ubuntu on production servers without any problems at all.

Is that because server software cares about stability first and foremost? Boy I wonder.