Proxmox VE 9.2 host randomly hard-freezes with no logs or errors (MSI X299 / i9-7920X)

[EvanF]

I'm trying to track down random, complete freezes on my Proxmox host, and I've run out of useful leads.

Hardware

• Motherboard: MSI X299 SLI PLUS (MS-7A93), BIOS 1.H0 (2020)
• CPU: Intel Core i9-7920X (12 cores / 24 threads)
• RAM: 128 GB DDR4-2400 (8 × 16 GB, non-ECC)
• GPUs: 2 × RTX 3060
• Storage: ZFS mirrors for the host and guest storage, plus a SAS2308 HBA passed through to a TrueNAS VM
• Network: USB 2.5 GbE adapter for the main bridge; onboard Intel I219-V available separately
• PSU was replaced earlier in 2026.

Software

• Proxmox VE 9.2.x
• Currently running kernel 6.17.13-21-pve
• NVIDIA 550 driver; pinned kernel due to compatibility concerns with newer kernels.

The problem

The entire host randomly freezes. Fans and power remain on, but the host, web interface, SSH, networking, and guests become unreachable. It never recovers without a physical power cycle.

I've documented six freezes since installing this motherboard on September 24. They occur after anywhere from roughly 5 hours to 8 days of uptime, with no obvious pattern. The host is not idle when it happens.

There are no useful kernel errors, MCE reports, PCIe AER errors, or storage errors before the freezes. Persistent external telemetry records temperatures, fan speeds, CPU load, memory use, Vcore, and the 3.3 V rail every second. These readings remain normal right up to the freeze. Remote kernel logging also stops without recording an obvious error.

What I've tried

• Disabled CPU C-states in the BIOS and verified the setting; the freezes continue.
• Ran MemTest86+ 7.20. An initial parallel run reported errors, but two full single-core passes and 900 additional passes over the reported region produced zero errors.
• Monitored temperatures, voltages, and system load; nothing abnormal has shown up.
• Set up persistent logging and remote kernel logging to capture the final moments.
• Checked PCIe link status and investigated the possibility of network-driver issues.

I'm not sure whether this is a motherboard/CPU stability problem, a kernel or driver issue, or something involving PCIe or power management.

Any help or suggestions are appreciated!

Thanks!

 

[xybo]

Check driver that your nic is using, if its using this e1000e than this is know issue. You can google how to resolve it.

 

[EvanF]

View attachment 100323
Check driver that your nic is using, if its using this e1000e than this is know issue. You can google how to resolve it.

The onboard Intel I219-V wasn't being used when the crashes started. I only started using it for crash telemetry after the problem began. The USB Realtek 2.5GbE adapter is the primary NIC and carries all guest traffic.

I've also checked for the suspected e1000e hardware-hang errors and haven't found any. Do you know of a specific e1000e bug that can cause a complete host freeze without logging a hardware-unit hang?

 

[xybo]

no, not sure about that RT USB card, but, can you remove it and assign in /etc/network/interface your primary bridge(vmbr0) to use intel onboard nic so you can eliminate RT as cause of this problem.