Why am I experiencing a kernel halt with Debian Trixie Linux before “/init” runs on a ASUS M5A99FX motherboard?

I had an ASUS M5A99FX Pro 2.0 motherboard running Debian Trixie Linux in CLI mode on a server. The motherboard failed after a power outage. I had another motherboard of the same model running the same version of Linux, except using the KDE desktop, and with completely different peripheral cards.

I pulled the unit off the desktop and installed it into the server. Not too surprisingly, I had to clear the CMOS to get the unit to POST, which unfortunately means the BIOS settings were lost. I reset the settings to the best of my ability to remember. The unit loads the kernel, but no matter what I try, it will not start the /init. Apparently it halts just after the last kernel command concludes. There is no error… it just halts.

I have tried an earlier version of the kernel and every BIOS switch of which I can think. I have tried booting from different drives, both SSD and spinning hard drives, and both over SATA and USB.

I am loading the kernel in GRUB using:

linux /boot/vmlinuz-6.12.107+deb13-amd64 root=UUID=884a0f09-f70f-497c-9bd4-c2192c2868d6 ro earlyprintk=efi,keep log

Here is the last page of the boot:

[    0.558898] pci 0000:02:00.1: extending delay after power-on msec
[    0.558993] pci 0000:02:00.1: D0 power state depends on 0000:
[    0.559307] PCI: CLS 64 bytes, default 64
[    0.559447] Trying to unpack rootfs image as initramfs...
[    0.559932] PCI-DMA: Disabling AGP.
[    0.560051] PCI-DMA: aperture base @ b4000000 size 65536 KB
[    0.560103] PCI-DMA: using GART IOMMU.
[    0.560154] PCI-DMA: Reserving 64MB of IOMMU area in the AGP a
[    0.560911] software IO TLB: tearing down default memory pool
[    0.563129] LUT offset 0 assigned for vector 0x400
[    0.563302] perf: AMD IBS detected (0x000000ff)
[    0.563634] amd_uncore: 4 amd_nb counters detected
[    0.564311] Initialise system trusted keyrings
[    0.564374] Key type blacklist registered
[    0.564497] workingset: timestamp_bits=36 max_order=23 bucket_o
[    0.564557] zbud: loaded
[    0.564956] fuse: init (API version 7.41)
[    0.565307] integrity: Platform Keyring initialized
[    0.565365] integrity: Machine keyring initialized
[    0.583131] Key type asymmetric registered
[    0.583182] Asymmetric key parser 'x509' registered
[    0.681940] Freeing initrd memory: 36080K

I tried adding nitramfs_async=0 to the kernel parms and got this:

[    0.178127] Mountpoint-cache hash table entries: 65536 (order: 7, 524288 s, linear)
[    0.289204] smpboot: CPU0: AMD FX(tm)-8350 Eight-Core Processor (family: , model: 0x2, stepping: 0x0)
[    0.289598] Performance Events: Fam15h core perfctr, AMD PMU driver.
[    0.289670] ... version:                0
[    0.289746] ... bit width:              48
[    0.289796] ... generic registers:      6
[    0.289844] ... value mask:             0000ffffffffffff
[    0.289895] ... max period:             00007fffffffffff
[    0.289946] ... fixed-purpose events:   0
[    0.289995] ... event mask:             000000000000003f
[    0.290134] signal: max sigframe size: 1776
[    0.290610] rcu: Hierarchical SRCU implementation.
[    0.290661] rcu:     Max phase no-delay instances is 1000.
[    0.290770] Timer migration: 1 hierarchy levels; 8 children per group; 1 cr snode level
[    0.292960] NMI watchdog: Enabled. Permanently consumes one hw-PMU counter.
[    0.293315] smp: Bringing up secondary CPUs ...
[    0.293518] smpboot: x86: Booting SMP configuration:
[    0.293569] .... node  #0, CPUs:      #1 #2 #3 #4 #5 #6 #7
[    0.310023] smp: Brought up 1 node, 8 CPUs
[    0.310023] smpboot: Total of 8 processors activated (64212.72 BogoMIPS)

I really think this must be a BIOS setting, but I cannot think what. I have tried changing:

OS Type: Other OS
EHCI Hand-off: Enabled
Core C6
C1E
Launch CSM: Enabled.

The BIOS refuses to let me change the CSM setting. It automatically changes it back complaining about the graphics card. The CPU is an AMD FX-8350.