Skip to content
Oznog

1.3 · hardware failures and rmas · as-is

A settings dump that skips a whole menu

date
20260903
what happened
Supermicro's own configuration utility captured 282 BIOS settings on the X10DRU-i+ boards, including PCIe bifurcation, but only 201 on the X10DRH-iT board, with the entire bifurcation submenu absent from the dump even though the physical BIOS screen has it and it works.
what it cost
The fleet came close to buying PCIe-switch NVMe carriers to work around a limit the board did not actually have.
what changed
A setting's absence from a vendor dump is no longer treated as proof a board lacks the capability. That board's gold file cannot carry the setting, so it is reset by hand in Setup after any CMOS reset.
the check now
Confirm on the physical BIOS screen before concluding a board lacks a capability; roughly 200 settings on a dual-socket board is itself the tell that the dump under-reports.

A tool that reports a subset of the truth is more dangerous than one that fails outright, because its output reads as complete.

On 20260903, while building the gold BIOS files that let this fleet apply firmware settings as code, the same vendor utility was run against two Supermicro board families. On the X10DRU-i+ boards it dumped 282 settings as commented text, including the IIO bifurcation submenu that decides whether a PCIe slot presents itself as one device or four (which is what makes a four-drive NVMe carrier work). On the X10DRH-iT board it dumped 201 settings and the bifurcation submenu was simply not there. No warning, no comment, no error. The file just ends up shorter.

The board can bifurcate. The physical BIOS screen has the menu and it works. Reading the dump alone, the obvious conclusion was that this board family could not split a slot, and the obvious remedy was to buy carriers with an onboard PCIe switch, which cost real money and would have solved nothing.

Two things came out of it. First, the gold file for that board cannot carry the setting, so it has to be set by hand in Setup after any CMOS reset, and that is written down where the next person will look. Second, a count is now a signal: a dual-socket server board with about two hundred exposed settings is under-reporting, and the physical screen is the tiebreaker before any money moves.

Source: node0 lessons v0.1, lesson 1.3. Sanitized: checklist v0.1, 20260921; names pass only; voice pass 20260921. Part of oznog.com/node0.