Windows 10 May 2020 vs. Linux Performance On AMD Ryzen Threadripper

[phoronix]

Phoronix: Windows 10 May 2020 vs. Linux Performance On AMD Ryzen Threadripper

Given the recent release of the Windows 10 May 2020 Update, here are some fresh benchmarks showing how the latest Windows 10 software update paired with the latest AMD drivers performs against the latest 2020 Linux distribution releases. This testing was done on an AMD Ryzen Threadripper 3970X box given the interesting performance differences we have seen in the past to Linux's advantage with these HEDT processors. The Linux distributions tested against Windows 10 May 2020 Update were Ubuntu 20.04 LTS, openSUSE Tumbleweed, Arch-based Manjaro 20.0.2, Clear Linux 33250, and Fedora Workstation 32.

http://www.phoronix.com/vr.php?view=29250

[oleid]

I guess Windows vs Linux benchmarks are only interesting (as in: you don't know the result before you start reading the article) once you throw in a few GPU tests.

[andyprough]

Clear is not a supported desktop distro so its performance is irrelevant, and Tumbleweed on btrfs-xfs is slow -- anyone seeking performance would be using ext4. And I wouldn't use Manjaro, which is tuned for user friendliness, not for performance. If I didn't want to install Arch the hard way, there are a number of better installers than Manjaro which will give a vanilla Arch setup.

[ix900]

Well that was clear that Clear Linux is optimized for AMD

[milkylainen]

What are Windows binaries built with in general?
I could imagine that some binaries are generated with ICC on Windows?
Like Zstd f.ex? GCC windows binary or ICC? Microsoft?

[CochainComplex]

Clear is not a supported desktop distro so its performance is irrelevant, and Tumbleweed on btrfs-xfs is slow -- anyone seeking performance would be using ext4. And I wouldn't use Manjaro, which is tuned for user friendliness, not for performance. If I didn't want to install Arch the hard way, there are a number of better installers than Manjaro which will give a vanilla Arch setup.

I tend to support your statement in the past and still somehow today I m advocating Clear Linux. But their step backwards reducing the desktop user experience was a game changer for me. They have been just at the beginning of getting an influencing distro on the desktop field. And they brought a lot of momentum into distro world. But for now I have switched back to Debian derivates.
To much desktop stuff is missing and it can not all be replaced with flatpaks

[Volta]

Nothing new to see here and that's good. I wonder how Ubuntu performs with more desktop oriented kernel config? 1000Hz, low latency desktop etc.

[JustRob]

Really a cucumber to banana comparison.

While it's interesting to know how Windows compares to Linux, in general, too few distributions of Linux were used.

There is (for the most part) two distributions of Windows, each which has some ability to tune for specific use cases; while for Linux some distributions are constructed solely for security, performance, ease of use, or even to introduce a better way (with varying results).

This is much like choosing the benchmark which runs fastest and proposing that it's a metric; while it's speed relies on not testing many things that are important to various groups or use cases.

For example I like Operating System X because it's fast, well designed, secure by default and fairly popular (thus there's some support) - but instead I use a different operating system because it's not far behind in those features (even better in some cases) and far easier for me to understand (the philosophy behind its design) and use; so it's more productive for me. Either one could be configured to beat Windows or fall painfully behind; that fact isn't what's going to enable me to choose between the three.

[ColdDistance]

Really a cucumber to banana comparison.

While it's interesting to know how Windows compares to Linux, in general, too few distributions of Linux were used.

There is (for the most part) two distributions of Windows, each which has some ability to tune for specific use cases; while for Linux some distributions are constructed solely for security, performance, ease of use, or even to introduce a better way (with varying results).

This is much like choosing the benchmark which runs fastest and proposing that it's a metric; while it's speed relies on not testing many things that are important to various groups or use cases.

For example I like Operating System X because it's fast, well designed, secure by default and fairly popular (thus there's some support) - but instead I use a different operating system because it's not far behind in those features (even better in some cases) and far easier for me to understand (the philosophy behind its design) and use; so it's more productive for me. Either one could be configured to beat Windows or fall painfully behind; that fact isn't what's going to enable me to choose between the three.

• Fedora 32 Workstation is a general purpose desktop system.
• Ubuntu for desktop is a general purpose desktop system.
• Manjaro is a desktop system for end users and gaming.
• They are all bloated to give an easy desktop experience.

You are welcome.

[Azultra]

I guess Windows vs Linux benchmarks are only interesting (as in: you don't know the result before you start reading the article) once you throw in a few GPU tests.

Not really nice to say, and actually the recent Ryzen 4xxx benchmarks on laptops were surprising in that regard (Windows faster than Linux), so it's interesting to see that it's not a general trend.