Ryzen 5950X: "Uncorrected, software containable error" events after upgrade to kernel 7.0.x

[harrydus]

Hello,

we are seeing a number of MCE/RAS events on several Proxmox VE 9.2 hosts after upgrading from the 6.17 kernel series to the new 7.0 kernel series and would like to know whether others are observing something similar.

• Hetzner dedicated servers
• AMD Ryzen 9 5950X
• ZFS
• Multiple independent hosts affected
• Proxmox VE 9.2.x
• Kernel: 7.0.6-2-pve

CPU: Model name: AMD Ryzen 9 5950X 16-Core Processor cpuid=0x00a20f10 microcode=0x0a20102e

Since upgrading to kernel 7.0.x we have seen MCE events on multiple hosts.

Example event:
2026-06-01 09:53:50 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=7), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffff8dad806f7b80, misc=0x10000000000000, walltime=0x6a1d56ae, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x00000007, microcode=0x0a20102e

Log:
Jun 01 09:53:50 proxmox02 kernel: mce: [Hardware Error]: Machine check events logged
Jun 01 09:53:50 proxmox02 kernel: [Hardware Error]: System Fatal error.
Jun 01 09:53:50 proxmox02 kernel: [Hardware Error]: CPU:1 (19:21:0) MC7_STATUS[Over|UE|MiscV|AddrV|PCC|SyndV|-|Poison|Scrub]: 0xffff8dad806f7b80
Jun 01 09:53:50 proxmox02 kernel: [Hardware Error]: Error Addr: 0x0000000000000000
Jun 01 09:53:50 proxmox02 kernel: [Hardware Error]: IPID: 0x0000000000000000, Syndrome: 0x0000000000000000
Jun 01 09:53:50 proxmox02 kernel: [Hardware Error]: Bank 7 is reserved.
Jun 01 09:53:50 proxmox02 kernel: [Hardware Error]: cache level: RESV, tx: INSN

All affected systems (5 so far) have the same CPU model (Ryzen 5950X) using the same microcode version. We are running this machines for quite a while and hadn't this kind of errors with kernel 6.x

Did anyone experience similar issues?

 

[j.theisen]

Hi @harrydus

thanks for posting in the forum!

Can you please confirm that this behavior started after just changing the kernel version or did it start after the upgrade from PVE 8 to 9?
If latter, please try rolling back the kernel to 6.17 with no other software changes and confirm the issue doesn't persist.

Are you using a custom performance governor or CPU frequency scaler?
Please also provide the output of the following commands

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver

Are these log events accompanied by any other symptoms / outages or is it currently mainly a cosmetic issue?

EDIT:
Also if available, please provide logs from additional events to compare the error messages / addresses.

Yours sincerely
Jonas

 

[harrydus]

Hello,

yes I can confirm, that this kind of problems started after switching to kernel 7.0. We already switched back to 6.17 to test it and confirmed that with 6.17 this kind of problem didn't happen. However: Also with Kernel 6.17 there were MCE errors, but they were always correctable or deferred, so nothing as serious as a "Uncorrected, software containable error".

We don't use a custom performance governor or CPU frequency scaler.

The log events weren't accompanied by other symptoms, however we had some random reboots in the past on this and other machines at Hetzner which were not related to this errors.

Here are the MCE messages and contents of scaling_governor and scaling_driver of all affected hosts.
Note: Host 4 is the one on which we switched back to kernel 6.17. You can see that after event 3 2026-05-07 00:57:32 +0000 error: Uncorrected, software containable error. only 2 other noncritical errors of the type Corrected error, no action required did occur

# Host 1:
scaling_governor: performance
scaling_driver: amd-pstate-epp
MCE events:
1 2026-04-27 21:58:07 +0000 error: Corrected error, no action required., CPU 2, bank Unified Memory Controller V2 (bank=14), mcg mcgstatus=0, mca DCQ SRAM ECC error. Ext Err Code: 6, mcgcap=0
x0000011c, status=0x80000002d386d163, walltime=0x69efdbef, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x0000000e, microcode=0x0a20102e
2 2026-06-01 09:53:50 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=7), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffff8dad806f7b80, misc=0x10000000000000, walltime=0x6a1d56ae, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x00000007, microcode=0x0a20102e


# Host 2:
scaling_governor: performance
scaling_driver: amd-pstate-epp
MCE events:
1 2026-03-13 02:46:41 +0000 error: Deferred error, no action required., CPU 2, bank Unified Memory Controller V2 (bank=25), mcg mcgstatus=0, mci Task_context_corrupt, mcgcap=0x0000011c, statu
s=0x9090909090909090, walltime=0x69b37a91, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x00000019, microcode=0x0a20102e
2 2026-05-30 05:09:12 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=13), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt
Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffff8ccac1108b20, misc=0x10000000000000, walltime=0x6a1a70f8, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x000
0000d, microcode=0x0a20102e


# Host 3:
scaling_governor: performance
scaling_driver: amd-pstate-epp
MCE events:
1 2026-04-24 06:41:23 +0000 error: Deferred error, no action required., CPU 2, bank Unified Memory Controller V2 (bank=20), mcg mcgstatus=0, mci UECC Poison consumed, mcgcap=0x0000011c, statu
s=0x9066ffd8cf9de9ff, walltime=0x69eb1093, cpu=0x00000001, cpuid=0x00a20f12, apicid=0x00000002, bank=0x00000014, microcode=0x0a201211
2 2026-05-06 09:07:52 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=15), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt
CECC Task_context_corrupt, mcgcap=0x0000011c, status=0xffffd3dcc06c8b10, misc=0x10000000000000, walltime=0x69fb04e8, cpu=0x00000001, cpuid=0x00a20f12, apicid=0x00000002, bank=0x0000000f, micr
ocode=0x0a201211

# Host 4:
scaling_governor: performance
scaling_driver: amd-pstate-epp
MCE events:
1 2026-03-29 22:49:01 +0000 error: Deferred error, no action required., CPU 2, bank Unified Memory Controller V2 (bank=21), mcg mcgstatus=0, mci Task_context_corrupt, mcgcap=0x0000011c, statu
s=0x9090909090909090, walltime=0x69c9ac5d, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x00000015, microcode=0x0a20102e
2 2026-04-30 06:07:38 +0000 error: Deferred error, no action required., CPU 2, bank Unified Memory Controller V2 (bank=16), mcg mcgstatus=0, mci Error_overflow CECC, mcgcap=0x0000011c, status
=0xc931d231c0315d5c, misc=0x10000000000000, walltime=0x69f2f1aa, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x00000010, microcode=0x0a20102e
3 2026-05-07 00:57:32 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=25), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt
UECC Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffffffffa514b640, misc=0x10000000000000, walltime=0x69fbe37c, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=
0x00000019, microcode=0x0a20102e
4 2026-05-26 13:53:19 +0000 error: Corrected error, no action required., CPU 2, bank Unified Memory Controller V2 (bank=26), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt Task
_context_corrupt, mca DRAM ECC error. Ext Err Code: 0 Memory Error 'mem-tx: generic write, tx: instruction, level: L2', memory_channel=0,csrow=0, mcgcap=0x0000011c, status=0xdead000000000122,
 misc=0x10000000000000, walltime=0x6a15a5cf, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x0000001a, microcode=0x0a20102e
5 2026-05-29 14:36:55 +0000 error: Corrected error, no action required., CPU 2, bank Unified Memory Controller V2 (bank=21), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt Task
_context_corrupt, mca DRAM ECC error. Ext Err Code: 0, memory_channel=0,csrow=0, mcgcap=0x0000011c, status=0xdead000000000300, misc=0x10000000000000, walltime=0x6a19a487, cpu=0x00000001, cpui
d=0x00a20f10, apicid=0x00000002, bank=0x00000015, microcode=0x0a20102e

# Host 5:
scaling_governor: performance
scaling_driver: amd-pstate-epp
MCE events:
1 2026-05-11 19:40:54 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=13), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt
UECC Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffffffff8620b000, misc=0x10000000000000, walltime=0x6a0230c6, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=
0x0000000d, microcode=0x0a20102e
2 2026-05-13 09:38:16 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=17), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt
UECC Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffffffff8620b000, misc=0x10000000000000, walltime=0x6a044688, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=
0x00000011, microcode=0x0a20102e
3 2026-05-29 14:02:40 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=9), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt P
oison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffff8eba6eb1ca28, misc=0x10000000000000, walltime=0x6a199c80, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=0x0000
0009, microcode=0x0a20102e
4 2026-05-31 14:11:24 +0000 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=10), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt
UECC Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffffffff9f60b000, misc=0x10000000000000, walltime=0x6a1c418c, cpu=0x00000001, cpuid=0x00a20f10, apicid=0x00000002, bank=
0x0000000a, microcode=0x0a20102e

As a side note: We already communicated with Hetzner support, but they won't do anything about this because

"Proxmox is not something we support or guarantee on our server models"

 

[leesteken]

My 5950X with microcode 0x0a201213 works fine with all the recent 7.0 Proxmox kernels. That in itself is not useful to you but maybe you can do (or request) a BIOS update?

EDIT: I also have the 3.20251202.1~bpo13+1 version of the amd64-microcode package installed. Do your systems have it installed?

 

[harrydus]

My 5950X with microcode 0x0a201213 works fine with all the recent 7.0 Proxmox kernels. That in itself is not useful to you but maybe you can do (or request) a BIOS update?

EDIT: I also have the 3.20251202.1~bpo13+1 version of the amd64-microcode package installed. Do your systems have it installed?

Yes, we have 3.20251202.1~bpo13+1 of amd64-microcode installed. I'm very skeptical that Hetzner support will do a BIOS update just for us, since their policy is "We don't officially support Proxmox".

 

[Domino]

No errors here on a 5950x.

amd64-microcode is already the newest version (3.20251202.1~bpo13+1)

Motherboard is an Asrock x570m pro4 with bios P5.65
Memory is M391A4G4AB1-CWE

 

[leesteken]

I'm very skeptical that Hetzner support will do a BIOS update just for us, since their policy is "We don't officially support Proxmox".

But is sounds like a Linux kernel, is mostly Ubuntu 26.04(?), compatibility issue with the BIOS/firmware. Do they support Ubuntu? Would it maybe be worth your trouble to install Ubuntu on one server instead of Proxmox? I think a request for an up-to-date BIOS is not unreasonable (regardless of operating system) with the recent firmware security fixes in AMD's AGESA.

Motherboard is an Asrock x570m pro4 with bios P5.65

This is AGESA Combo V2 PI 1.2.0.E and I have AGESA ComboV2 1.2.0.F. @harrydus, what AGESA version does the Hetzner BIOS currently have? I think the amd64-microcode package from Debian is typically older.

 

[egeekial]

I am seeing what appears to be a similar issue on my local Proxmox (PVE 9.2.4, kernel 7.0.14-5-pve) host running a 5950X (microcode: 0xa201211).

kernel: [Hardware Error]: System Fatal error.
kernel: [Hardware Error]: CPU:1 (19:21:2) MC9_STATUS[Over|UE|MiscV|AddrV|PCC|SyndV|UECC|Deferred|Poison|Scrub]: 0xffffffffffffffff
kernel: [Hardware Error]: Error Addr: 0x0000000000000000
kernel: [Hardware Error]: IPID: 0x0000000000000000, Syndrome: 0x0000000000000000
kernel: [Hardware Error]: Bank 9 is reserved.
kernel: [Hardware Error]: cache level: L3/GEN, tx: RESV

CPU frequency scaling configuration:

# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
powersave

# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
amd-pstate-epp

The host did not reboot or freeze, and I did not observe any VM or service interruptions.

 

[Stage3600]

Adding a data point with two things I don't think are in the thread yet: non-ECC consumer RAM, and a phantom event on microcode 0x0a201213 — which argues against the newer microcode being protective.


System: Ryzen 9 5950X (stepping 2, cpuid 0x00a20f12), ASUS TUF Gaming X570-Plus WiFi, BIOS 5044 (01/04/2026), 96GB non-ECC DDR4 (two mixed consumer kits), ZFS root, PVE 9.2.4. Governor performance, driver amd-pstate-epp.


Timeline:

• Jun 1, 21:14 — installed proxmox-kernel-7.0.6-2 (built 2026-05-29). I'd deliberately held off on 7.0 until then. First boot 22:18 the same evening.
• Jun 6 — first "Uncorrected, software containable error." UMC bank 18, status 0xffffffffffffffff.
• Jun–Aug — six more, all UMC banks, all with 0xffffffff in the upper half of the status word. No panics, no VM or service interruption. Five fell inside a single 30-day uptime.
• Eleven months of rasdaemon history before June: never an uncorrected event.
• Aug 9 — BIOS 5021 → 5044, microcode 0x0a201211 → 0x0a201213.
• Aug 14 — another one, still on 7.0.14-11-pve, now at 0x0a201213. UMC bank 27, status 0xffffffff980d9d10.

Linux journal entry for most recent MCE event:

Aug 14 03:23:15 pve kernel: mce: [Hardware Error]: Machine check events logged
Aug 14 03:23:15 pve kernel: [Hardware Error]: System Fatal error.
Aug 14 03:23:15 pve kernel: [Hardware Error]: CPU:1 (19:21:2) MC27_STATUS[Over|UE|MiscV|AddrV|PCC|SyndV|UECC|Deferred|Poison|Scrub]: 0xffffffff980d9d10
Aug 14 03:23:15 pve kernel: [Hardware Error]: Error Addr: 0x0000000000000000
Aug 14 03:23:15 pve kernel: [Hardware Error]: IPID: 0x0000000000000000, Syndrome: 0x0000000000000000
Aug 14 03:23:15 pve kernel: [Hardware Error]: Bank 27 is reserved.
Aug 14 03:23:15 pve kernel: [Hardware Error]: cache level: RESV, tx: INSN
Aug 14 03:23:15 pve rasdaemon[2243]: rasdaemon: mce_record store: 0x556485a1ea28
Aug 14 03:23:15 pve rasdaemon[2243]:            <...>-202230 [001] .....     0.002229 mce_record 2026-08-14 03:23:15 -0500 Unified Memory Controller V2 (bank=27), status= ffffffff980d9d10, Uncorrected, software containable error., mci=Error_overflow Processor_context_corrupt UECC Poison consumed Task_context_corrupt, cpu_type= AMD Scalable MCA, cpu= 1, socketid= 0, misc= 10000000000000, addr= 0, synd= 0, mcgstatus=0, mcgcap= 11c, apicid= 2, microcode= a201213

The status words are the tell. 0xffffffffffffffff decodes every flag as set — PCC, poison consumed, overflow — which would have panicked the host. Instead I had 30 days of uptime containing five of them.

Sample of MCE Events:

2026-06-06 22:55:04 -0600 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=18), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt UECC Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffffffffffffffff, misc=0x10000000000000, walltime=0x6a24f9a8, cpu=0x00000001, cpuid=0x00a20f12, apicid=0x00000002, bank=0x00000012, microcode=0x0a201211

11 2026-08-14 03:23:15 -0500 error: Uncorrected, software containable error., CPU 2, bank Unified Memory Controller V2 (bank=27), mcg mcgstatus=0, mci Error_overflow Processor_context_corrupt UECC Poison consumed Task_context_corrupt, mcgcap=0x0000011c, status=0xffffffff980d9d10, misc=0x10000000000000, walltime=0x6a7ed073, cpu=0x00000001, cpuid=0x00a20f12, apicid=0x00000002, bank=0x0000001b, microcode=0x0a201213

Now pinned to 6.17.13-21-pve and watching. Will report back either way. Happy to post the full rasdaemon log or dmidecode if useful.