AMD Ryzen 7 3700X + Ryzen 9 3900X Offer Incredible Linux Performance But With A Big Caveat

[phoronix]

Phoronix: AMD Ryzen 7 3700X + Ryzen 9 3900X Offer Incredible Linux Performance But With A Big Caveat

After weeks of anticipation, we can now share how the AMD Ryzen 7 3700X and Ryzen 9 3900X performance is under Linux. These first Zen 2 processors do indeed deliver a significant improvement over Zen/Zen+ processors and also battle Intel's latest 14nm CPUs but for Linux users there is one big, unfortunate issue right now.

http://www.phoronix.com/vr.php?view=28040

[mlau]

Some of the MKL-DNN results (and others) seem a bit odd.. like there's a part that detects a non-intel CPU and hits a non-optimized path, or march-native
is not yet making sensible decisions for zen2. Maybe test some of these with march=skylake?
AMDs newest is an exceptionally great all-round CPU, no doubt.

[monte84]

Michael Can we expect gaming benchmarks in the near future? Also, on the Intel system, like the 9900K, was MCE (Multi Core Enhancement) enabled in the BIOS?

[Nille_kungen]

Very impressive but i still was expecting some game benchmarks.

[Michael]

Very impressive but i still was expecting some game benchmarks.

Game benchmarks are coming, ran short on time due to spending easily dozens of hours trying to figure out the Linux systemd/boot issue.

[leipero]

Page 11:
The Ryzen 7 3700X showed an 10% improvement or so over the Ryzen 7 3700X.

Very impressive CPUs, whatever the issue is I'm sure it would be easy to solve since it works on some distributions.

[Raka555]

Michael, can you maybe try a distro like void linux or Devuan that does not use systemd ?
Maybe its not the kernel...

[Adarion]

Yes, I was also thinking that - if you already reach init state - it might not (entirely) be the kernel that is hampering boot. Still, one would expect things to just work out. After alls these CPUs should even be able to run ancient CP/M or DOS code.

[cyring]

Thank you for this review.
Is it Arch Linux ready ?

[Michael]

Yes, I was also thinking that - if you already reach init state - it might not (entirely) be the kernel that is hampering boot. Still, one would expect things to just work out. After alls these CPUs should even be able to run ancient CP/M or DOS code.

Right as said in the article, using Linux 5.2 Git + Ubuntu 18.04 LTS does work, so it's some weird kernel<->user interaction going on.