Dead CPU or Dead Motherboard The Swap Test and What the Parts Cost

On LGA sockets the pins live on the board, they bend if you look at them with intent, and a bent pin imitates a dead CPU almost perfectly — dead memory channels, no POST, cold processor.

1. Remove the CPU and inspect the socket under bright light and magnification, at a low angle. Bent pins break the regular grid pattern — scan row by row for any pin out of line or flattened.
2. A CPU that 'died' immediately after a reseat, a heatsink change, or a transport knock is a pin problem until proven otherwise.
3. Straightening pins is jeweller's work with a fine needle and patience — possible on a few pins, hopeless on a crushed field. Price that honesty into the repair decision.
4. Also inspect the CPU's underside pads for dirt or a fingerprint film; clean with isopropyl alcohol and a lint-free wipe. Contact faults are free to fix and commonly missed.

Five minutes with a magnifier has saved more 'dead CPUs' than any other step in this guide family.

Substitution has an asymmetry: some swaps risk the known-good component. Sequence matters.

1. If the suspect board shows any VRM suspicion — heat marks, bulged caps, hot MOSFETs, a short reading at the CPU power input — do NOT fit your known-good CPU into it. A regulator that fails high can destroy the replacement processor in seconds.
2. In that situation the safe direction is CPU-into-known-good-board only: move the suspect processor to a board you trust. If it boots there, it is innocent and the original board is convicted without endangering anything.
3. If the board's VRM passed inspection cleanly (no marks, no short, sane resistance), either direction is acceptable — choose whichever spare you actually have.
4. Match compatibility exactly: same socket, same generation family, and a BIOS on the test board that supports the test CPU. An unsupported pairing produces a fake 'dead' result.

The rule in one line: never feed a good CPU to an unproven regulator.

One variable per test, everything recorded. Sloppy swapping manufactures mysteries.

1. Change exactly one component, then test fully: minimum boot configuration, five minutes of run time, the heat check, and the POST reading.
2. Note the exact configuration and result for each trial. Memory of 'I think it booted before' is how good parts get binned.
3. If the suspect CPU boots in the good board, let it run and warm properly — intermittent processors exist, and a 30-second boot is a weaker acquittal than a warm, stable run.
4. If neither swap is possible (no spare CPU, no test board), the honest status is 'board or CPU, unseparated' — the repair decision then has to be made on price and probability, not on a fake certainty.

Probability, stated honestly: between the two, the board fails far more often than the processor. But 'far more often' is a betting line, not a verdict — the swap is the verdict.

Prices move and vary by generation, so think in structure rather than numbers:

1. Older-generation server CPUs are usually the cheaper half of the pair on the used market — processors are plentiful, pulled from decommissioned fleets.
2. The matching server motherboard is typically the scarcer, pricier part, and a like-for-like board preserves everything else: RAID configuration, drivers, licensing tied to the platform, and the OS installation.
3. A different (newer) board is rarely a drop-in: it can demand a new CPU family and new RAM generation, converting a board purchase into a platform purchase. Price the whole cascade before choosing it.
4. A donor machine of the same model, bought whole, sometimes undercuts a bare board and donates PSU modules, fans and rails besides. For ageing fleets this is often the smartest spend.

Decide with the full invoice in view — part, downtime, and the migration work a platform change drags behind it.

Processors almost never die of old age alone. A replacement fitted without asking the question can be a second victim.

1. The usual upstream killers: a VRM that overshot, a heatsink mounted badly enough to cook it over months, or a PSU rail event. All three leave the replacement exposed to the same death.
2. Before the new CPU goes in, the board's VRM must pass the inspection in the companion guide — no heat marks, no bulged caps, sane resistance at the CPU power input.
3. Fit with fresh thermal paste, correct heatsink torque (even, cross-pattern, per the manual), and verify temperatures under load through the BMC sensors once it boots.
4. If the CPU died with no upstream evidence anywhere — it happens, rarely — accept it, document it, and move on.

A swap that ends the story is a fixed server. A swap that ignores the cause is a sequel.