A PC that only boots on the second power-on because of memory training

A PC that only boots on the second power-on because of memory training. The supporting cases describe the following observed symptom pattern:

• Cold boot failed while a second power cycle booted successfully.

This page treats the reported successful action as a bounded repair path. It does not claim that the same visible symptom always has the same cause.

Searchable Windows error codes and messages

No numeric Windows error code was present in the supporting records. Search by the exact symptom and displayed message instead of inventing a code.

Use this procedure when:

• Windows 10 or Windows 11 shows the same symptom, code, affected component, and trigger described above;
• the proposed repair can be performed on the exact affected device, package, update, disk, or setting;
• BitLocker is unlocked and the Windows installation is assumed correct unless the case explicitly identifies installation damage; and
• important data and the current configuration have been backed up.

An exact Windows build may remain unstated when it does not change the repair. Exact model, firmware, driver, update, disk, or partition information becomes mandatory when it selects the branch or file.

Do not use this page merely because one error code or a broadly similar symptom appears. Stop and choose another diagnostic path when the named component, trigger, boot stage, or demonstrated cause differs.

Do not apply firmware, storage-controller, partition, registry, driver-removal, Reset, or reinstall steps speculatively. RAID and SCSI-specific branches are retained for the small number of systems that actually use them, but they are not the default path.

The original discussion may omit the exact Windows build, hardware revision, driver package version, update subrevision, or complete causal mechanism. Those omissions do not invalidate a narrowly demonstrated outcome.

The following cannot remain unknown before making a potentially destructive change: the exact target, why it matches this page, the current value or version, the backup or rollback path, and the condition that will stop the procedure.

• A verified backup of important files
• Administrator access or Windows Recovery Environment as required
• A way to record screenshots, commands, versions, and before/after state
• Stable power for firmware, storage, update, or installation work
• A second working device or recovery media when the affected PC may not restart
• The exact computer or motherboard model and hardware revision, stable power, and the manufacturer’s recovery instructions.
• The exact update or target Windows version, complete result/extend code, and preserved Setup or Windows Update logs.
• A known-good comparison component or a repeatable isolation test; do not purchase parts solely from one stop code.

1. Record the exact displayed error, stop code, message, and time of failure.
2. Record what changed immediately before the problem: update, driver, firmware setting, hardware, application, sleep/resume, clone, or installation attempt.
3. Back up accessible personal files. If Windows does not start, use WinRE, installation media, or another trusted environment without modifying partitions.
4. Photograph or export the current configuration that will be changed.
5. Disconnect only nonessential external devices unless a device is the subject of the test.
6. Perform one material repair branch at a time so success remains attributable.

Before applying the repair:

1. Photograph the current firmware settings and boot order before changing anything. Partition layout should be detected from the disk; firmware mode is needed only when it changes the repair branch.
2. Record the complete code, update identifier, phase, and operation. For feature-upgrade rollback, preserve Panther/Rollback logs and run SetupDiag before retrying.
3. Reproduce the failure with one variable changed at a time. Record which component follows the fault into a known-good configuration.
4. Compare the exact symptom and trigger with the cases listed in the Issue section.
5. If the diagnosis depends only on a generic code, stop and collect the missing logs, device identity, disk layout, or update information.

Choose only the branch whose diagnosis and starting state match the machine.

Implementation checklist

1. Identify the exact computer or motherboard model and hardware revision from the chassis label or current firmware screen.
2. Photograph every relevant firmware page and the current boot order. Download only the exact-model vendor instructions and recovery file.
3. Connect stable AC power. If the repair changes one setting, change only that setting; if it flashes firmware, follow the vendor’s required intermediate-version order.
4. Save, restart, and allow any memory training or firmware recovery cycle to finish without interrupting power.
5. Return to firmware setup, verify the intended value and Windows Boot Manager entry, then test Windows startup.

Branch 1 — demonstrated resolution path

1. Enter firmware setup and disable Memory Fast Boot.
2. Set the DDR4 memory speed manually to 3600 rather than relying on the failing training behavior.
3. Save, cold-shutdown, and retest first boot.
4. Restart or repeat the original operation.
5. Confirm the reported success condition: The author marked the problem solved after these memory-training changes.

Risk recorded for this path: Low.

If a displayed menu, package name, disk letter, or firmware option differs, do not substitute a guessed value. Detect the actual value or obtain the exact-model instructions first.

• The exact symptom stops after one controlled change: preserve the before/after evidence and repeat the relevant restart, update, device, or workload test.
• Operation succeeds but the underlying cause remains unknown: record this as a successful recovery, not a proven root-cause diagnosis.
• The symptom changes or a new code appears: stop. Treat it as a new diagnostic state rather than repeating the old action.
• The repair has no effect: restore the previous state when possible and move to the next evidence-supported branch.
• The disk disappears, hardware testing becomes unstable, or recovery no longer starts: stop repair attempts and protect the data first.

• Do not flash firmware for a similar model, interrupt a flash, or change storage-controller mode without a tested rollback path.
• Do not repeatedly retry an upgrade while overwriting the logs from the previous failure.
• Do not infer failed hardware from a single blue-screen code without repeatable testing or health evidence.
• Do not disable BitLocker, Secure Boot, Defender, Windows Firewall, or storage-controller features as generic experiments.
• Do not use registry cleaners, generic driver updaters, or commands copied for a different partition layout.
• Do not call a reinstall, Reset, factory recovery, or hardware replacement a component-level repair.
• Do not expose internal database identifiers on the public page.

1. Restore the recorded firmware value or use only the exact-model vendor recovery procedure.
2. If the change was an update removal, reinstall only after Windows is stable and the corrected package or driver is available.
3. Return the original component only if it is known safe; otherwise keep it disconnected and restore the last stable configuration.
4. Reconnect devices or re-enable services one at a time and retest.
5. If rollback is unavailable because the action was destructive, stop and restore from the verified backup or system image.

1. Restart Windows twice and perform one full shutdown/start cycle when the repair affects startup, firmware, drivers, or storage.
2. Repeat the exact operation that previously failed.
3. Confirm the original code or message is absent; do not count a different failure as success.
4. Check Device Manager, Windows Update history, Event Viewer, SetupDiag, disk health, or the affected application as relevant.
5. Record the final Windows build, driver/firmware/package version, and changed setting.
6. Preserve the evidence until the system completes ordinary use, shutdown, sleep/resume, and update checks appropriate to the issue.

Confidence is approximately 80% when the symptom, trigger, component, and action all match; the action is the only material change; and the original operation succeeds repeatedly. The supporting records contain explicit successful outcomes, but no independent TechXplored lab reproduction is claimed.

Stop when the target cannot be identified exactly, a required backup or rollback is unavailable, the procedure would require guessing a partition or firmware value, the system becomes less stable, or the result cannot be repeated. Single-source pages remain narrower than corroborated pages and should not be generalized beyond their demonstrated configuration.