Windows 10 vs. Ubuntu 18.04 LTS Performance On AMD Ryzen 9 3900X

[Michael]

If you used Windows 10 1903, why didn't you pit it against Ubuntu 19.04?
(I'll see myself out)

Because the Ubuntu 19.04 ISO doesn't boot on the system as ASUS has yet to release the patched BIOS.

[AsuMagic]

It would seem to me that there is something wrong going on with ffmpeg tests. Different CPU extension support built in?

[bug77]

Because the Ubuntu 19.04 ISO doesn't boot on the system as ASUS has yet to release the patched BIOS.

Actually, it's AMD who has to release the fixed microcode. They had a version out, but had to pull it because it was causing instability.
But my post was just tongue-in-cheek

[Dr. Righteous]

So, unless I missed it; this is 18.04 LTS running the "stock" Kernel which should be 4.15 correct? Impressive indeed. I would like to see the results running the latest Linux kernel.

[wizard69]

Interesting results! I’m wondering if anybody out there has any idea why Windows is so much slower with PyBench and other interpreter tests? It just seems strange that both Python and PHP failed so badly compared to Linux.

[reavertm]

It would seem to me that there is something wrong going on with ffmpeg tests. Different CPU extension support built in?

If someone knows ffmpeg buildsystem specifics:

https://salsa.debian.org/multimedia-team/ffmpeg/blob/debian/master/debian/rules

[AsuMagic]

If someone knows ffmpeg buildsystem specifics:
https://salsa.debian.org/multimedia-...r/debian/rules

But are phoronix results using the distro's package or a local build from PTS?

[rbmorse]

Because the Ubuntu 19.04 ISO doesn't boot on the system as ASUS has yet to release the patched BIOS.

Ubuntu has patched systemd in the 19.04 repository so the Dingo can boot on Ryzen3000/x570 although the installer .iso hasn't been updated. Now that 18.10 is dead, you can change software sources to notify for all upgrades, apt update, then run:

sudo do-release-upgrade

to upgrade an existing 18.04 to 19.04.

Probably not useful for testing/comparing purposes as it won't be a "clean" install, but for those of us with Ryzen 3000 series CPUs and a need to run a 5.x series kernel, this will work.

[reavertm]

But are phoronix results using the distro's package or a local build from PTS?

You are right.
https://openbenchmarking.org/innhold...3b4c29125791de

Which means it relies upon what ffmpeg buildsystem detected vs what was overriden (not a lot, see table on first page of article).
And actually uses different archive for Windows, it could be prebuilt for Windows as no Windows compilation flags are listed on first page. Therefore this test might be misleading, as it might neither compare out of the box performance (if Ubuntu packages are not used) nor real OS performance (in which case software would be compiled with the same flags on both OSes).

[tildearrow]

Michael, I can't believe it. Please don't use BLAKE2 for cross-OS comparisons, and instead PLEASE do use it in cross-processor ones.

"Cycles Per Byte" means it is supposed to compare processor architectures, NOT operating systems. If the results vary, it's caused by kernel differences or compiler differences (unless the whole program has been written in assembly, which is likely).