Startup & Boot — No POST, Boot Loops, Safe Mode and the Recovery Options

1 The boot sequence, step by step

When you press the power button, the motherboard's power supply delivers voltage and the processor starts executing a small piece of firmware. That firmware is either UEFI or a legacy BIOS — the same job described with two generations of technology. Its first task is to check that the basic hardware is present and sane, and its second task is to decide where to load the operating system from and to start that hand-off. Everything after the hand-off is the operating system's job, not the firmware's.

There are two boot paths, and knowing which one your machine uses changes the commands that are valid later in this guide. On a UEFI system, the firmware looks at a small special partition on your drive (the EFI System Partition), finds a file called the Windows Boot Manager, and that manager reads a database of boot entries (the BCD) to find and load Windows. On a legacy (BIOS/MBR) system, the firmware reads the first sector of the drive (the MBR), which points to a boot manager that then reads the same kind of boot configuration to start Windows. Both paths end the same way — Windows loading its kernel — but the first few files live in different places, which is why a "fix" written for one path can do nothing on the other.

1. Power on — the power supply starts and the CPU begins running firmware.
2. POST — firmware tests the CPU, memory and core hardware.
3. Video — firmware switches on basic display output.
4. Firmware hand-off — the firmware finds the boot device and starts its boot manager.
5. Boot manager — reads the boot configuration (BCD) to find Windows.
6. Windows loader — loads the Windows kernel and early drivers.
7. Windows — starts services, drivers, then shows sign-in.

Keep one detail in mind from here on: the word "boot" is used loosely to mean several different steps. When a guide or an error message says "boot failed", ask which step. "No bootable device" is a firmware problem finding the drive; "Windows Boot Manager" with an error code is the boot-manager/BCD step; "INACCESSIBLE_BOOT_DEVICE" is Windows failing to reach its own drive after it has already started. The same two words describe three different hand-offs.

2 What POST actually does

POST stands for Power-On Self-Test. It is the firmware's quick check that the essential hardware is present and functional before it trusts that hardware with the operating system. At a minimum it verifies that the CPU is executing, that memory is readable, and that core input/output paths respond. It is deliberately brief — it is not a full diagnostic and it does not deeply test your storage drives or run benchmarks. Its purpose is to catch "the machine is fundamentally broken" problems early, before anything else runs.

The important thing to understand is that POST happens before any operating system exists in memory. That is why you cannot "run a Windows repair" on a machine that is failing POST — there is no Windows to repair yet. This is the clearest rule in the whole guide: if a problem is a POST problem, it is a hardware problem. Windows tools only become available once POST has passed and the firmware has successfully handed off to a drive.

How a failed POST announces itself varies by manufacturer. Some boards emit a series of beeps from a small speaker, some light up a set of small diagnostic LEDs (often labelled CPU, DRAM, VGA and Power), and some simply show no video at all. Because the exact meaning of a beep pattern or LED depends on the board, treat generic "beep code" lists as a rough guide only — the specific board manual is authoritative. The next two sections deal with these signals and with the safe physical checks they usually point to.

Two practical notes. First, POST does not test everything: a drive that passes POST can still be slowly failing at the file level, and a GPU can pass the basic POST check and still produce a broken image later. Second, some symptoms that look like POST failures (for example, a monitor that only works on one output) are really display problems, not memory problems — the "no POST vs no display" section keeps those apart.

3 No POST vs no display vs no boot

These three phrases are often used interchangeably, but they describe three different points on the pipeline and they send you to three different places. Keeping them straight is the single biggest time-saver in this guide.

The deciding question is always can you open the firmware menu? If yes, POST has passed and the problem is a display, a boot-device, or a Windows problem. If no — no logo, no menu, just beeps or darkness — the problem is at the hardware/POST level, and no amount of Windows recovery will reach it. That single check routes you to the right half of this guide immediately.

A common trap: a "no display" case is sometimes a RAM case in disguise. A failing memory stick can stop the firmware before video initialises, which looks identical to a dead GPU from the front of the machine. That is why the safe first move for "fans spin, no screen" is to reseat and test the memory and to try a different display, rather than to assume one or the other. The RAM section and the display section give the exact checks.

4 The PC has power but nothing appears

This is one of the most stressful symptoms because the machine is "on" in some sense but gives you nothing to work with. First, split it into two sub-cases that mean very different things. If there is truly no sign at all — no LED, no fan, no click — you are at stage A (power) and the check is almost entirely external: the outlet, the cable, and the power switch on the back of the supply. If, however, the fans spin and a light is on but the screen stays dark, the board is receiving power and the problem has moved to POST or the display path (stages B and C).

For the "fans spin, dark screen" case, the checks below are the safe first moves, in order. Each one is quick and reversible, and each rules out a common cause before you move to the next. Do not skip to opening the case until you have tried the external checks — a loose monitor cable or the wrong display output is far more common than a failed component.

1. Confirm the monitor is on, set to the correct input, and try a known-good cable.
2. If the PC has both a motherboard (integrated) and a graphics-card output, try the other one.
3. Wait a full minute — after a power cut some boards retry POST a couple of times.
4. Power off, unplug, and reseat the memory modules one at a time (see RAM).
5. Remove all but one memory stick and one display connection, then power on.
6. Listen for beeps and watch for diagnostic LEDs (see beep codes).

If, after reseating the memory and trying a different display, the machine still gives no logo, no beeps and no firmware menu, you are most likely looking at the power supply, the CPU, or the motherboard. These are the components where "open it and keep guessing" stops being a good plan, for two reasons: they are the hardest to test without proper equipment, and a wrong move (a shorted cable, a dropped screw, a bent pin) can turn a one-part fault into two. That is a legitimate point to stop and call in help — see professional service. What is still useful at this stage is to record exactly what you see (which fans, which lights, any beeps), because that record is what a technician needs to skip ahead.

5 Beep codes and diagnostic LEDs

A beep code is the firmware's way of reporting a POST failure when it cannot trust the display. The pattern — how many beeps, whether they are long or short, and how they repeat — is a small code that maps to a category of failure. The crucial caveat: the meaning of a pattern is set by the firmware vendor, and it is not universal. The classic schemes (AMI, Award, Phoenix) differ from each other, and many modern boards use their own tables. So a generic online list is a hint, not an answer. If your board has a manual, its beep/LED table is the one to trust.

Even without the exact manual, the broad themes are consistent across vendors, and they are useful for triage. Clusters of short beeps very frequently point at memory; a continuous or long repeating tone often points at video. Boards with diagnostic LEDs remove the guessing entirely — a lit "DRAM" LED means memory, a lit "VGA" LED means video, and so on. When you have LEDs, read them first; they are the firmware telling you exactly which of its tests failed.

Two things to avoid. First, do not treat a single beep you have never seen before as proof of a specific part — confirm it against your own board's table before replacing hardware. Second, do not interpret beeps on a machine that otherwise shows a normal logo; some boards beep once on a perfectly successful POST, which is not an error at all. If in doubt, the safest read is to run the minimum-configuration test, which identifies the failing component regardless of what any beep chart says.

6 Basic hardware isolation

When a machine will not POST, the most reliable diagnostic is to reduce it to the minimum set of parts the firmware needs and see if that works. This is called a minimum-configuration test, and its logic is simple: if the stripped-down system POSTs, add parts back one at a time until it stops working — the last part you added is the culprit. If the stripped-down system still does not POST, the problem is in the core set (board, CPU, memory, power), which is usually where professional testing takes over.

Before you begin, be honest with yourself about comfort and safety. This involves opening the case and handling components, which is where the guide draws its line for casual users. If opening the case is not something you are at ease doing, that is a perfectly reasonable stopping point — the steps below are provided so you know what is involved and what to ask a technician to do, not as a pressure to attempt them.

1. Unplug and open the case; ground yourself.
2. Leave only the CPU, one stick of memory, and the power supply connected.
3. Disconnect all storage drives, and any add-in cards you do not need.
4. If the CPU has integrated graphics, remove the separate graphics card; otherwise keep one GPU with its power connected.
5. Power on and watch for POST (logo, beep, or a change in LEDs).
6. If it now POSTs, add one component back at a time, power-testing after each.

Two interpretations. If the minimum system does POST, you have your answer: whatever you re-added last was the failure, and you can then clean-reseat or replace just that part. If the minimum system still will not POST even with a single known-good memory stick, you have narrowed the fault to the board, CPU, or power supply — the three parts that are genuinely hard to test at home. At that point, the minimum-configuration result is exactly the information a technician wants: "It does not POST even with one known-good stick and no add-in cards." That single sentence saves them an hour of work.

7 RAM and memory-related startup problems

Memory is the single most common cause of a no-POST / no-video symptom, and the good news is that it is also the easiest to diagnose at home because it is modular: you can test sticks and slots individually. A machine with a bad memory stick often shows exactly the symptoms that cause the most alarm — fans spinning, no logo, sometimes a set of beeps — and the fix is as gentle as reseating the module.

Memory problems also have a "soft" face that appears later in the pipeline rather than at POST: a system that boots but randomly restarts, throws stop errors, or is unstable under load can be suffering from marginal memory that passes the quick POST test but fails under real use. If your machine POSTs fine but behaves unreliably, memory is still a prime suspect, and the same one-stick-one-slot method applies.

1. Power off, unplug, ground yourself.
2. Remove all memory and start with a single stick in the slot the manual recommends.
3. Power on. If it POSTs, that stick and slot are fine.
4. Add the second stick; if it now fails, the newly added stick is suspect.
5. Repeat per stick and per slot to map which combination fails.

Read the results carefully. If only one stick ever fails, that stick is the culprit and can be replaced. If a stick works in one slot but not another, the slot (or that region of the board) may be the issue. If no combination POSTs and you are confident the sticks are good, the fault has moved to the board or CPU. One caution: some boards will not start at all with certain memory configurations (for example, populating the wrong pair of slots), so if your manual specifies a slot order, follow it — a "dead" board that starts happily once you move the stick to the correct slot is a configuration issue, not a failure.

8 CPU, motherboard and power-related POST problems

These three components account for the remaining no-POST cases, and they are the ones where home diagnosis has honest limits. They are soldered to the board or central to it, they are expensive, and testing them properly calls for equipment (a known-good replacement part, a multimeter, a POST card) that most people do not keep. The practical approach is therefore to recognise these failures, confirm the cheaper causes are ruled out first, and then hand the diagnosis to someone equipped to test them.

Each has a characteristic flavour. A failing power supply tends to cause power that is present but unreliable: random shutdowns, fans that start then die, or a system that only powers on with the side panel open (a classic supply-under-load sign). A failing motherboard often presents as a total no-POST with no beeps and no LEDs that change, even with known-good memory. A CPU failure is genuinely rare for a casual user to hit, and when it does it is usually accompanied by thermal problems (the fan spins wildly then the system shuts down to protect the chip) or a clear physical event such as an overheating incident.

Before you conclude any of these is the cause, make sure the cheaper causes are genuinely out of the picture: good power at the outlet, a retested single memory stick in the correct slot, and a clean minimum configuration. A surprising number of "dead motherboard" calls turn out to be a single bad stick of memory or a loose 8-pin power connector to the board. Once those are ruled out and the minimum system still will not POST, you have done the useful part of the diagnosis — you can state with confidence that the fault is in the board/CPU/power group — and the next step is professional, not further guessing.

9 GPU and display-output problems

Display problems are the ones most often misread as "the computer is dead" when in fact the computer is fine and only the video path is broken. The tell is that the system may be POSTing perfectly — you might even hear it complete its startup sound — yet the monitor stays dark or shows a garbled image. This is stage C failing while stage B succeeds, and it is frequently the cheapest thing in the whole guide to fix because it is often a cable, an output, or a setting rather than a part.

There is one important branch to check first: does your machine have both a graphics-card output and a motherboard (integrated) output? If the discrete graphics card is failing, switching the monitor to the motherboard's video output — which uses the CPU's built-in graphics — will get you a working image, and it also confirms the GPU as the culprit. Not every board has integrated graphics, and on some systems the motherboard output is disabled whenever a discrete card is fitted, so this is worth confirming rather than assuming.

1. Try a different cable, then a different monitor, to rule out the display path.
2. Switch to a different video output on the same card.
3. If available, switch to the motherboard's integrated output.
4. If the card is loose, power off and reseat it, and confirm its power cable is connected.
5. Compare: if integrated output works but the card does not, the GPU is the failure.

Interpreting the outcome: if a different monitor or cable fixes it, you have a display-path part to replace and the computer was never at fault. If integrated output works but the discrete card never does, the graphics card is the failure and can be replaced (or you can run on integrated graphics if your CPU supports it and you do not need the card). If no output works and there are beeps or a VGA LED, the failure may actually be memory stopping the board before video initialises — so if the display checks all come up empty, loop back to the memory test before declaring the GPU dead. That ordering — display path first, memory second — avoids the expensive mistake of replacing a GPU that was never the problem.

10 Storage drive detection problems

A "no bootable device" message, or a firmware menu that simply does not list your drive at all, means the firmware cannot find the storage it is supposed to hand off to. This is stage E, and it splits into two very different situations. If the drive is not listed in the firmware's storage/boot device list, the firmware cannot even see it, which points to a connection or a failed drive. If the drive is listed but the system still will not boot from it, the drive is present but the boot configuration or the operating system on it is the problem — that is a BCD/Windows issue, covered later.

The connection checks come first because they are the common and the cheap. A SATA data cable or a power cable can work loose, a drive can be seated in a failing port, or a drive can have quietly begun to fail and dropped off the bus. The safe sequence is to reseat both cables, try a different port or cable, and only then consider the drive itself.

1. Open the firmware and check whether the drive appears in the storage or boot device list.
2. If it is absent: reseat the drive's data and power cables.
3. Try a different SATA port (for an internal drive) or a different cable.
4. For an M.2 drive, reseat it and confirm it is fully clicked into place.
5. Check the drive's health (S.M.A.R.T.) if it does appear but behaves badly.

Two further branches worth naming. If a drive is present but was installed on a different partition scheme than the firmware is currently using (a drive formatted as GPT attempted in legacy mode, or vice versa), the firmware may see it and still refuse to boot it — that is a firmware setting, not a broken drive, and the firmware section covers it. And if the drive is listed, boots, but Windows then cannot reach it (an "INACCESSIBLE_BOOT_DEVICE" type error), the drive's presence is not the issue — the operating system's ability to talk to it is, which is usually a driver or a file-system problem handled in the recovery sections. The through-line: absent from the firmware is a connection/drive problem; present but won't load is a configuration/OS problem.

11 Incorrect boot device or boot order

The firmware does not "boot Windows" — it boots whatever device is first in its boot order, and then trusts that device to take it further. When the order is wrong, the firmware dutifully hands off to the wrong thing and the result looks like a failure even though every component is healthy. The most common real-world versions of this are a USB stick or external drive sitting in a port (the firmware tries to boot from it, fails, and moves on or errors), a network/UEFI-network entry ranked above the internal drive, or an optical drive that used to contain an install disc.

The fix is almost always to make the correct internal drive the first boot device and to remove any removable media. Because the exact menu layout differs between firmware vendors — the boot order may live under "Boot", "Boot Options", "Boot Priority" or a "Boot Override" list — the guidance here is about what to look for rather than which button to press. Find the list of bootable devices, confirm your internal drive is present, and move it to the top.

A useful habit: once you have fixed the order, leave removable media out of the machine until you are done. A single forgotten USB installer can make a perfectly good computer "suddenly" fail to boot, and it is the kind of cause that is embarrassing to find but takes seconds to fix. If changing the order does nothing, the problem has moved on from "which device" to "the device won't load", which is where the BCD and recovery sections take over.

12 UEFI/BIOS configuration checks

The firmware is where boot order, boot mode and a few safety switches live, and several startup failures are really firmware-setting failures. Getting into it is the first step: the key to open the firmware varies by manufacturer (commonly Del, F2, or F12, sometimes shown on the very first moment of the screen), so if you are not sure, try the one your maker lists or use the operating system's "restart into firmware" option, which the recovery sections describe. Because the menus are laid out differently on every board, this section tells you what to check, not where it is.

The two settings most worth verifying are boot order (covered above) and the boot mode. A drive is installed in one of two modes — UEFI or legacy (CSM) — and it will generally not boot in the other. If a machine that previously booted fine suddenly cannot, and hardware changes are ruled out, a firmware reset (or an accidental toggle of the boot mode, often after a BIOS update or a cleared CMOS battery) can put the firmware in a mode that no longer matches the drive. The safe move is to note the current values, and if nothing is clearly wrong, use the firmware's "load default settings" option and re-set the boot order — this resets a stray change without you having to hunt for which setting drifted.

One setting deserves a specific caution: Secure Boot. It is a signature-checking feature and, in most cases, you should leave it at its default. There are narrow situations where a boot-mode or signature mismatch makes it relevant, but toggling it is the kind of change that can both fix and break things depending on the setup, so treat it as a last-resort lever with a note of what you changed, not a first move. The general principle for this whole section: the firmware is a place to verify and, if needed, reset to known defaults, not a place to experiment broadly. If you cannot confidently read the settings, resetting to defaults and re-choosing the boot order is the safe and usually sufficient action.

13 Boot loops and repeated restarts

A boot loop is a system that keeps restarting before it reaches a usable state, and the most diagnostic detail about any loop is where in the sequence it restarts. A restart that happens before the Windows logo (during firmware/early hand-off) points toward the boot device, the boot configuration, or a failing component. A restart that happens after the Windows logo has appeared — often after a period of spinning dots — points toward the operating system: a bad update, a driver, corrupt system files, or a failing drive. The same visible symptom ("it just keeps rebooting") has two different halves, and the timing tells you which half you are in.

The first question to answer, before any fix, is the classic one: what changed right before this started? A loop that begins the morning after a Windows update is, overwhelmingly, update-related, and the fix is to uninstall that update from the recovery environment (covered in the update section) rather than to start rebuilding boot files. A loop with no obvious trigger is more likely to be a failing drive, corrupt system files, or — less commonly — heat or power. Noting the timing of the loop and what changed is the cheapest diagnostic in this section and it should always come first.

One pattern deserves special mention because it is so common and so fixable: the loop that begins right after an update and lands on "Preparing to configure Windows" or a percentage that hangs. That specific behaviour is the operating system applying an update that cannot complete, and the standard remedy is to interrupt into the recovery environment and uninstall the update. If a loop has no connection to a change, the priority shifts to getting into recovery to run a system-file check and to verifying the drive's health, because a physically failing drive is the one cause of looping that no software fix will cure and that you do not want to keep booting against.

14 Windows Automatic Repair

Automatic Repair is Windows' built-in first response to a boot that keeps failing. When Windows detects that it has not started successfully a few times in a row, it stops trying to boot normally and instead launches its repair routine, which scans for a small number of common, automatically-fixable problems — usually missing or corrupt boot files — and attempts to correct them. You recognise it by the "Automatic Repair" screen and, when it cannot fix the problem, the "Your PC didn't start correctly" screen with an "Advanced options" button.

It is worth understanding what Automatic Repair is not, because people sometimes expect more of it. It is a narrow, automated sweep, not a full diagnosis, and it will not resolve a failing drive, a bad update it doesn't recognise, or a deeper file-system problem. When it succeeds, great — the machine boots and you move on, ideally checking what changed. When it fails, it hands you exactly what you want: the "Advanced options" menu that is the front door to every manual recovery option in the rest of this guide. In other words, Automatic Repair failing is not a dead end; it is the point at which the machine has done its automatic part and is offering you the manual controls.

If you find yourself in a loop of Automatic Repair itself (it runs, fails, reboots, runs again), there is a useful thing to do: interrupt the cycle and go straight for "Advanced options" instead of waiting for each automated attempt. Repeatedly letting it run its automatic pass can take a long time without changing the outcome. From "Advanced options" you can choose Startup Repair, System Restore, Uninstall Updates, or a Command Prompt — the same set of tools you would use if you had booted into the Recovery Environment on purpose. The next section explains that environment, which is where you will do most of your actual work.

15 The Windows Recovery Environment

The Windows Recovery Environment, usually called WinRE, is a lightweight recovery operating system that Windows uses to fix itself when the main system cannot start. Everything in the next sections — Startup Repair, System Restore, Uninstall Updates, the Safe Mode choices, and the Command Prompt — is reached from WinRE's menu. Understanding WinRE as a hub rather than a single tool makes the rest of the guide much easier to navigate, because you are not learning many separate entry points; you are learning one menu.

There are several ways to get there, and which one you use depends on how far the system will go. If Windows reaches a sign-in screen, you can trigger an advanced restart from the power button (hold the modifier key and choose Restart) or from the settings recovery area. If Windows is in the "Automatic Repair" loop, the "Advanced options" button drops you straight into the same menu. If Windows will not start at all and you cannot reach either of those, the reliable method is to boot from Windows installation or recovery media and choose "Repair your computer" — that path is covered in the recovery-when-Windows-won't-start section. There is also a well-known trick of force-interrupting startup a few times in a row to coax Windows into the recovery menu, which is useful when the automatic repair is not triggering on its own.

Once you are in WinRE, the layout is consistent: a choice between "Troubleshoot", "Start up settings", and a few direct options, and "Troubleshoot" leads to "Advanced options", which is the collection of tools you will use. Because the exact wording shifts slightly between Windows versions, read the menu for intent rather than memorising labels. The important orientation is this: WinRE gives you access to your drive and to the boot tools without needing Windows to have started. That is precisely what makes it the right place to work — you can inspect and repair the boot configuration and the system files even though the system itself is the thing that is broken.

16 Entering and using Safe Mode

Safe Mode starts Windows with only the minimum drivers and services it needs, deliberately leaving out most graphics drivers, third-party drivers, startup programs and background services. The value of that minimal state is diagnostic: if a machine that will not start normally can start in Safe Mode, you have proven that the operating system itself is intact and that the failure is caused by something Safe Mode excludes — usually a driver, a recently installed program, or a startup item. That single fact narrows a huge space of possibilities down to a manageable few.

You can enter Safe Mode in a few ways depending on how far Windows gets. From the Recovery Environment, it is reached through "Troubleshoot" → "Advanced options" → "Startup Settings" → Restart, then choosing the Safe Mode option (the exact key is shown on the screen that follows). If Windows does reach a sign-in screen, some versions let you restart directly into Safe Mode from the power options. Once you are in Safe Mode, the desktop looks stripped-back and lower-resolution, which is normal and expected — that is the minimal configuration working as designed, not a new problem.

What to do once you are in: the goal is to identify and then remove or roll back whatever Safe Mode is excluding. The common wins are uninstalling a program or driver that was installed just before the problem began, disabling a startup item, or restoring a driver to its previous version. Because Safe Mode is a diagnostic and repair state, not a permanent one, remember that the "fix" is to resolve the underlying cause so the machine can return to a normal start — you do not want to be living in Safe Mode. And if the machine cannot even enter Safe Mode, that is itself information: it suggests the problem is deeper than a single excluded driver or program, which points toward the system-file and boot-configuration tools below, or toward a hardware cause you should revisit.

17 Safe Mode with Networking

Safe Mode with Networking is the same minimal environment as plain Safe Mode, with one deliberate addition: the network stack. Everything else is excluded exactly as before, so it is just as good at isolating a driver or startup problem — but because the network is present, it is also the variant to choose when the fix you need requires an internet connection.

The situations that call for it are specific and practical. You may need to download a correct or previous driver that is causing the failure, run an online security scan to rule out or remove malicious software, or fetch a repair tool you do not already have on the machine. In each case the logic is identical to plain Safe Mode — boot minimal, fix the excluded cause — with the network added purely so you can retrieve the fix. If you do not need the network, use plain Safe Mode, because a smaller environment isolates a little more cleanly.

One caution that applies here but not to plain Safe Mode: because networking is active, be deliberate about what you download and run while in this state. You are already in a troubleshooting posture, and the last thing a failing machine needs is a second variable introduced by an untrusted download. Prefer official sources for drivers and tools. As with all Safe Mode work, the objective is to correct the underlying cause so the machine returns to a normal boot, not to make Safe Mode with Networking a permanent way of working.

18 Startup Repair

Startup Repair is the automated tool inside the recovery environment, and it is the right first manual attempt for most "Windows won't start" situations. It scans the boot configuration and the critical boot files, looks for the specific, known problems it is designed to handle, and attempts to correct them — typically missing or damaged boot files and boot-manager entries. It is a good first move because it is safe, it is automated, and it fixes a genuinely common set of failures without you needing to understand the boot internals.

The honest framing is that Startup Repair is a targeted sweep, not a universal fix. It will not repair a failing physical drive, and it will not fix every possible form of file corruption. When it succeeds, the machine boots and you should note what changed beforehand so the cause does not recur. When it reports that it could not repair the startup problem, that is not a failure of your effort — it is the tool telling you that the cause is outside its narrow set, and it is the correct signal to move on to the next options rather than to run it repeatedly in hope.

The sensible order is: try Startup Repair first (it is quick and safe), and if it does not resolve the issue, proceed to Uninstall Updates (if an update is a plausible trigger), then System Restore (if a restore point exists from before the problem), and only then to the command-line tools. Running the options in that order means each more involved step is only attempted when the simpler, safer ones have been ruled out — which is the same principle the whole guide follows, just applied within the recovery menu.

19 System Restore

System Restore rolls the operating system back to a restore point — a saved snapshot of system files, installed programs, drivers and settings taken at an earlier time. Its particular value is for the "it worked before, then something changed" situation: if a restore point exists from before the problem began, restoring to it can undo the change that broke the boot. It is worth trying early when you suspect a recent change is responsible, because it can reverse a class of problems that Startup Repair does not touch.

It is equally important to understand what System Restore does not do, because the name invites a misunderstanding. It is not a backup of your files: it does not roll back documents, photos or other personal files, which is reassuring. But it does undo programs, drivers and updates that were installed after the chosen restore point, so anything you installed in that window will be removed. Choose the restore point that is closest to (but before) the time the problem started, and be aware of what was installed since then.

The practical limit is that System Restore is only as useful as the restore points you have. If the machine never created a restore point, or the relevant one was deleted, the option will show no usable points and you must move on. Where a suitable point does exist, it is a low-risk, high-value step — it is reversible (Windows keeps the ability to undo a restore), it targets exactly the "a change broke things" cause, and it requires no command-line knowledge. That makes it one of the best options in the recovery menu, provided a point is actually available to restore to.

20 Uninstalling a problematic update

If the failure began shortly after a Windows update, uninstalling that update is very often the correct and the fastest fix, and it is one of the most satisfying resolutions in this whole guide because it is targeted and reversible. A bad update can corrupt the boot path, install a conflicting driver, or leave an update transaction half-applied — all of which present as "Windows will not start" — and removing the offending update returns the system to the state it was in before the update ran. This is the option to reach for first when the timeline points at an update.

From the recovery environment, the path is "Troubleshoot" → "Advanced options" → "Uninstall Updates", which typically offers the choice to remove recent updates or to remove a specific quality (security) update. Choose the update that matches the timing of when the problem started. The machine will apply the removal and restart; if the update was the cause, it should boot normally. If it does not, the update was not (or not only) the cause, and you continue with the other options.

One nuance: not every update-related failure is a simple "remove the last update" situation. An update that left a failed transaction in progress sometimes needs to be let finish or cleaned up differently, and a feature update (a major version change) is handled differently from a routine monthly update. The general rule, though, holds: when the timeline says "an update happened, then it broke", try removing that update before moving to more invasive steps. It is safe, it is targeted, and it is the fix that matches the most common real-world cause of a sudden boot failure.

21 Startup Settings and Advanced options

This section is about the menu itself, because it is where all of the previous options physically live and where a couple of extra, less-often-needed tools sit. From the recovery environment, "Troubleshoot" → "Advanced options" is the collection of Startup Repair, System Restore, Uninstall Updates, the Command Prompt, a link back to the firmware settings, and the "Startup Settings" entry. "Startup Settings" is specifically the place that offers the Safe Mode choices and a few other boot behaviours, presented as numbered options you select after a restart.

The Startup Settings list is worth reading rather than memorising, because the numbers can shift between versions and the screen always shows what each does. The Safe Mode entries (plain, with Networking, with Command Prompt) are the ones you have already met, but the list also carries a couple of specialised options that are useful in narrow situations. One is the option to disable the automatic restart on failure, which is genuinely helpful for diagnosing a loop: instead of the machine hiding a stop error by immediately rebooting, it will display the full error, which often names the offending driver. The other, disabling driver signature enforcement, allows an unsigned driver to load — a last-resort diagnostic, not something to leave on, because it weakens a protection.

The point of mapping the menu is that you stop treating "recovery" as one amorphous thing and start treating it as a small toolbox with a sensible order. For most people the working sequence is: try Startup Repair, then (if an update is suspected) Uninstall Updates, then (if a restore point exists) System Restore, then (if you need to isolate a driver or read a stop error) the Startup Settings options, and finally the Command Prompt for the manual repairs. Keeping that order means you always try the safe, automated, targeted options before reaching for the ones that require more care.

22 The recovery Command Prompt

The Command Prompt inside the recovery environment is the manual layer beneath the menu buttons: the menu's "Startup Repair" and the like are, at their core, running commands like the ones below. Reaching for it directly is appropriate when the automated tools have not resolved the problem and you want to run a specific check or repair. It is also the right place for the diagnostics that have no button at all, such as checking the health of the drive or the status of the recovery environment itself.

Two ground rules before any command. First, back up your data first if at all possible — see the data-protection section — because some of these commands touch the boot and system partitions. Second, understand each command before running it and run them one at a time, reading the output. The commands below are grouped from "safe to run, informative" to "changes the boot configuration", and each carries a note on what it does and its risk. None of them is a magic fix; each is a specific tool for a specific kind of problem.

reagentc /info

bcdedit

sfc /scannow

DISM /Online /Cleanup-Image /RestoreHealth

chkdsk C: /f /r

bootrec /scanos

bootrec /rebuildbcd

bcdboot C:\Windows

The through-line for all of this: the command line is where you go when the buttons have failed, and it is used in a specific order — check status, repair system files, then (and only then) rebuild the boot configuration, scanning before you rebuild. Used in that order, with a backup made first, it is a powerful and safe set of tools. Used out of order or without understanding, it is where people make the mistakes this guide is trying to prevent.

23 Boot configuration and BCD problems

The BCD — the Boot Configuration Data store — is the database the boot manager reads to work out where Windows is and how to load it. When the BCD is missing, incomplete, or corrupted, the failure has a very characteristic look: the system reaches the boot manager but then shows a "Windows Boot Manager" screen with a status code, or it simply cannot find a bootable operating system even though the drive is present. These are stage-F failures, and they are among the most fixable problems in this guide precisely because they are a data problem (a broken list of instructions) rather than a hardware problem.

The good news is that most BCD problems are resolved by the automated Startup Repair, because rebuilding a damaged boot configuration is one of the specific things it is built to do. If Startup Repair has already been tried and did not resolve it, the manual path is the bootrec and bcdboot commands in the Command Prompt section, used in order: confirm the tool can see the installation (/scanos), add it to the configuration (/rebuildbcd), and, if needed, recreate the boot files from a known-good Windows (bcdboot). Each of these is explained there with its risk; this section is about recognising the BCD as the cause and understanding why those commands are the fix.

Two things help you confirm the BCD (rather than the drive or the firmware) is the actual cause. First, the drive is detected by the firmware — you can see it in the firmware's device list — so stage E passed. Second, the error text points at the boot manager or at a missing boot store rather than at a missing device. A drive that is absent from the firmware is a connection/drive problem (see drive detection); a drive that is present but whose boot store is broken is a BCD problem, and that is the situation the rebuild commands address. Keeping that distinction clean prevents the common error of rebuilding a BCD on a machine whose real problem is a loose or failing drive.

24 Telling hardware failures from software failures

Much of startup troubleshooting is really this one decision: is the failure at the hardware level (before or during POST) or at the software level (once the firmware has handed off)? Getting it wrong wastes time in the wrong half of the guide — running Windows repairs on a machine that is failing POST, or reseating memory on a machine that is failing to load Windows. The reliable test is the firmware: if you can open the firmware menu and it responds, the core hardware has passed its self-test and the problem is almost certainly in the boot device or the operating system. If you cannot reach the firmware — no logo, no menu, just beeps or darkness — the problem is at the hardware/POST level.

Use the table as a score rather than a single decisive row. A machine that cannot reach the firmware, beeps, and has no drive in the device list is unambiguously a hardware story. A machine that reaches the firmware, sees its drive, but then shows "Windows Boot Manager" with an error is unambiguously a software/boot-configuration story. The genuinely ambiguous cases — the ones where a single clue does not decide it — are exactly why the guide is ordered the way it is: you work the stages in sequence, and the first stage that does not produce its "working" sign is your answer. When a case is unclear, default to checking the hardware basics (power, memory, display) first, because they are the cheaper and more common causes, and only move to the software tools once the firmware is confirmed to be reachable.

25 Protecting your data before advanced steps

Before you move from the safe, reversible checks into the more consequential territory — rebuilding boot files, restoring a system state, or anything that could require a reinstall — it is worth spending ten minutes thinking about the data on the machine. The reason is not that every recovery step destroys data (most do not touch your files), but that the last resorts in this guide do: a clean reinstall of Windows wipes the system drive, and a drive that is physically failing can lose everything if you keep booting it and stressing it. Getting your important files off first converts a potential total loss into a recoverable one, and it is the single highest-value "step" in the whole guide even though it repairs nothing.

There are a few ways to pull data off a machine that will not start normally. If you can reach the Recovery Environment, you can open its File Explorer and copy files to an external drive. If the machine will not reach recovery at all but the drive is healthy, you can often remove the drive, connect it to another computer (in an enclosure, or as a secondary drive), and copy the files directly. If the drive is making concerning noises or is clearly failing, the priority flips entirely: stop using it, get the most important files off as fast as you can, and treat the drive as time-limited. In every case the principle is the same — get the irreplaceable files somewhere safe before you do anything that could make the drive harder to read.

A practical rule of thumb: the further along you are in the recovery ladder, the more you should confirm the data is safe. At the "try Startup Repair" stage, a quick backup is prudent. At the "rebuild the BCD" stage, it is important. At the "reinstall Windows" stage, it is mandatory. Matching your backup effort to how irreversible the next step is is the whole point of this section.

26 Recovery when Windows cannot start at all

Everything in the recovery sections assumes you can get into the Recovery Environment. But there is a case where you cannot: Windows will not start at all, the automatic repair is not triggering, and there is no sign-in screen to trigger an advanced restart from. When that happens, the built-in recovery is simply unreachable from the machine, and the answer is to bring it in from outside — using Windows installation or recovery media.

The method is to create or obtain a bootable Windows USB (or DVD) on a working computer, then boot the failing machine from it and choose the "Repair your computer" option rather than "Install". That drops you into the very same Recovery Environment you would have reached from within Windows, with the full set of Advanced options — Startup Repair, System Restore, Uninstall Updates, Command Prompt — available against the drive. This is the most reliable way into recovery, and it is the one to use when the internal path is not available. Because it boots from external media, it also sidesteps some boot-order problems: you are not asking the broken internal boot to work, you are starting from known-good media.

Two practical notes. First, the media must be bootable in the mode your machine uses — UEFI media for a UEFI machine — so if a UEFI-created USB is not booting, that mismatch is a clue about the machine's firmware mode (and possibly the cause of its original failure). Second, if you are using this path because the built-in recovery would not come up, it is worth checking the recovery environment status from the Command Prompt once you are in — a disabled WinRE is a known reason the internal recovery menu is missing, and it is a fixable setting rather than a sign of deeper damage.

27 When the recovery options do not work

There comes a point in any startup problem where you have tried the automated repair, uninstalled the likely update, restored a system state, run the file checks, and rebuilt the boot configuration, and the machine still will not start. That is not a failure of the process — it is the process working as intended, having ruled out every fixable software cause and narrowed the remaining possibilities. When you reach this point, the honest set of remaining explanations is small, and it is worth naming them clearly so you can decide the next step rather than cycling through the same options in hope.

The remaining possibilities, in rough order of likelihood, are: a physically failing drive that no amount of file repair can cure; a deeper corruption that requires a clean reinstall of Windows; or a hardware fault (memory, board, power) that was masked until now. The way to tell which is to lean on the diagnostics you already have: if the drive's health check is reporting serious errors, the drive is the cause and the fix is replacement, with data recovered first. If the drive is healthy and the firmware is fine but every software repair has failed, a clean reinstall is the last software resort — and it is the point at which the data-backup step becomes mandatory, because a clean install resets the system drive. If the firmware itself is not reliable, the problem has been hardware all along and belongs in the professional column.

The discipline here is to stop re-running the same options and to make a decision based on the evidence. Reinstalling Windows is a legitimate and often successful final step, but it is the one action in this guide that erases the system drive, so it is only appropriate once the data is safe and the cheaper causes are genuinely excluded. And if the evidence keeps pointing at hardware rather than software, the correct and efficient thing is to stop at the software boundary and hand the diagnosis to a professional — the guide has already done its job by telling you precisely what is and is not working.

28 When professional hardware service is appropriate

Throughout this guide, a line has been drawn between the checks a careful user can make and the work that belongs to a qualified technician. That line is not about being cautious for its own sake; it is about where the tools, the parts, and the risk genuinely change. There are specific situations where continuing at home is either unlikely to help or actively likely to make things worse, and recognising them is part of doing startup troubleshooting well.

The clearest cases for professional help are: a machine that will not POST even with a single known-good memory stick and a clean minimum configuration (pointing at the board, CPU or power supply); a burning smell, visible burn marks, bulging components, or any sign of a short; water or liquid damage of any kind; and a failing drive that holds irreplaceable data. Each of these is either beyond safe home testing or carries a real risk of converting a single fault into a larger one — and in the data-recovery case, the risk is the loss of the data itself, which no amount of careful home work can undo.

The most useful thing you can do before you hand the machine over is to hand over the diagnosis, not just the machine. A short, factual summary saves the technician real time and helps you avoid being charged for re-discovering what you already know. It is worth stating which stage the machine reaches (does it POST? does the firmware open? is the drive detected?), what you have already ruled out (a known-good memory stick tested, the display path tried, an update uninstalled), and what changed right before the failure began. That summary is the direct output of working through this guide in order, and it is exactly the information that lets a professional skip ahead to the part that actually needs their tools.