Taichi X370 with Ubuntu idle lock ups/idle freeze

Hi!

Ehh, it's rather complicated in Your case I think. See, MCE reboot/freeze is completely different from segfaults, at least that's what I found out and what seems to be the case described in the forum link I posted in my previous posts.

Regarding "reboot in BIOS", that's a new thing for me. It kinda suggests You may have thermal problems or MCE "works" in BIOS as well :)
See in BIOS my CPU have always much higher temps than in OS (which happens to be Ubuntu). Like, when idling even with my overclock (from 3.0 -> 3.6) it's about 40C, but in BIOS it's always in high 50s, meaning 55-59. Why is that, I do not know, but it is.
By default there is this temp protection switched on (which I don't recommend disabling) which may interfere, but 70+ in BIOS, that's high, maybe it's ok for 1800x, but seems quite high.

Also, I have default cooler which came with my CPU, paste already pre-applied, You, obviously, don't. It might be a good idea to check how fast Your fans are going and what's the temp under load. One of the easiest You can do, just clean the thermal paste off and put new one on, I use grizzly paste, I apply the paste in very thin amount all over the CPU (this is what grizzly recommends) and in addition I put pea sized point in the middle, that works fine. I think that made my CPU 5 degrees cooler than default paste.
Set fans to default setting (by default they are in default), but just make sure they are.

I think my overclock is considered "mild", I overclock only pstate0, the rest are set to auto, so it works like this: highest is 3600, next is 2550 (or so, I don't remember exactly and I have no access to Ryzen at the moment to check) and lowest is 1550 (or so), which, to me, seems reasonable, I only want the max to be overclocked not the rest.

So, reapply thermal paste, run stress -c 16 (command in ubuntu) or prime95, monitor Your temps. Monitor syslog, if there are HW errors, it may have smth to do with voltage to be too low. If kill-ryzen is giving You segfaults but stress or prime95 not, You may have to RMA Your CPU. I had run memtest86 for a long time with my old (faulty) CPU, no issues, it was kill-ryzen which segfaulted. When voltage was low on my old overclocked CPU, I was getting MCE errors in syslog, increasing voltage helped (as expected, I was stupid to hope overclock @ default voltages to work :))
BUT, if You're running all default and kill-ryzen gives You segfaults, just RMA Your CPU.

Br, Kirurgs

MisterJ wrote: HxwRXP, are you still without a solution?  I have not kept up with your thread.  I cannot answer questions you have about Linux and Ryzen.  I do know that the Linux kernel has been modified for new AMD processors.  Please be sure you have the latest.  I also know that AMD was required to release a HW change (processor, I think) for the segfault problem.  I strongly recommend that you open a AMD support ticket and have a discussion with them.  It seems like AMD is making the updated HW available rather sparingly.  Good luck and please keep us informed here.  Thanks enjoy, John.

Kirurgs, Ryzen/Threadripper have a rather well know CPU temperature reporting oddity.  At one time AMD added 20 C to the CPU readings and programs like AIDA64 subtracted 20 C.  In my  system the BIOS always shows about +20 C in the BIOS.  I wish AMD would clearly tell us what the various CPU temperatures mean and how they are used internally and what we should really believe.

Thanks much, HxwRXP for your update.  It is great you are stable now.  I am not into OCing so know almost nothing about C-States. 
Some have had trouble with the BIOS 3.20 update and I have avoided it (3.10 on my board).  As near as I can determine it fixes only obscure problems (not defined).  The most important thing to remember is to use the Instant Flash in the BIOS.  Do not use the Windows method - Don't understand why ASRock has started providing that!  Enjoy, John.