Comment 381 for bug 1690085

(In reply to Dennis Schridde from comment #295)
> I am using an AMD Ryzen 5 2400G on an Asus ROG Strix B350-F (firmware
> version 3805 and 3803 before that) and was also experiencing freezes, e.g.
> after about one hour into `rsync -a /home ...`, with Gentoo Linux (Linux
> 4.15.10), Arch Linux 2018.03.1, Fedora 27, Fedora 28 nightly, Fedora 28
> beta. Setting rcu_nocbs=0-7, processor.max_cstate=5 or
> intel_idle.max_cstate=5 did not seem to have an effect -- the freezes still
> occurred. The hardware itself (mainboard, CPU, RAM) was cross-checked by
> the supplier, who found no fault in it. Since I disabled CPU C-states in
> the mainboard firmware about a day ago, I am no longer able to reproduce the
> freeze.

Ryzen 2xxx are almost Ryzen 1xxx but with lower TDP. Meaning lower CPU frequency and thus lower power.
And since all available solutions and workarounds are about not to let the chipset give less power to the system (cpu, memory...) or to explicitly give more power to the system, this is not surprising (to my opinion) that the bug is still here.
What will be more interesting is to see if the bug will still be here for the next generatation of Ryzen...

Also, since now this is clear like rock water that this bug is not a software bug (thus not a kernel bug), but clearly a bug from AMD (most probably in the full design of Ryzen both CPU and chipsets), please let me propose you all to consider again to trick your motherboard voltage. Just giving it very few more and you'll have a system fully working. The CPU will still be able to lower its power, to scale its frequency... This is totally a win-win.