AMD Ryzen 9 3900X stability issue

[falstaff]

Hi,

My almost two year old AMD Ryzen 9 3900X lately (since about 2-3 months) started to crash more often. At first it was around weekly, today the system crashed twice already, so it seems its getting worse. Usually the kernel starts to print BUG messages like this:

BUG: Bad page map in process gsd-xset:shlo6 pte:13813b025 pmd:180adb067
page:00000000b86924b4 refcount:2 mapcount:-1 mapping:00000000c9bb564a index:0x3a pfn:0x13813b

Usually no special/heavy workload at the time of the crashes, mostly idling Desktop system with a CPU at ~55°C. After a reboot, all things seem to work normally. I am using Arch Linux, and I have that problem since several kernel versions as well as with LTS kernels. To me a software issue seems unlikely.

I added the second 2x16GB memory set a bit later. But I run memtest86+ for a whole night without any reported memory issue. I still feel that its somehow memory related, I once "played" a bit with memory settings in the BIOS and felt it improved, but hard to tell since the crashes are very spaced out.

My hardware:

• AMD Ryzen 9 3900X (be quiet! Dark Rock Slim cooler)
• Corsair Vengeance LPX 2x2x16GB
• ASUS ROG STRIX X570-E
• Sapphire Radeon RX 570 Nitro+ (4GB)
• Samsung SSD 970 EVO Plus 1TB
• ADATA SX8200 2TB
• 2x6TB WDC
• Seasonic Focus Plus Platinum (850W, 4 years old)

I am a bit lost what the exact problem might be. Any tips/ideas what to tweak, how to diagnose, or what to replace first?

Thanks!
falstaff