GPU Boots Fine But Goes Black After Windows Login? Driver Load Point, Desktop Composition & Post-Login Load Diagnosis Guide

Key Takeaways
A normal boot screen followed by a black screen after login points to the driver handover, desktop composition, or a post-login program rather than a dead card. This guide separates those three stages with safe-mode, clean-boot, single-display, and external-display checks, so you can narrow the fault before replacing any part.
Where the Boundary Sits
The symptom described in this guide has a very specific shape. The machine powers on, the motherboard or manufacturer brand screen appears, and the early boot display looks normal. Then, at or shortly after the Windows login screen, the picture disappears. The monitor may report no signal, show a black screen with a visible mouse cursor, or flicker briefly before going dark. The machine itself may still sound as though it is running: fans spin, drive lights blink, and the login sound may even play.
That sequence matters because it draws a line through the display chain. The firmware display path works. Before Windows loads its full graphics driver, the system uses a basic display path provided by the firmware and by Windows itself. If that early picture is present and stable, the cable, the monitor input, and the card's basic output path are at least capable of producing an image. The fault therefore sits further along the chain, at one of three later stages: the handover from the basic display driver to the full graphics driver, the takeover by desktop composition, or the load imposed by programs that start after login.
This distinction changes the whole diagnostic approach. Replacing a cable or a card at this point would be guessing, because the early picture already shows that the basic path can work. The task now is to find which later stage breaks the picture, and to do so with tests that change one variable at a time.
The Three Stages After the Boot Screen
Stage One: The Driver Handover
When Windows moves from its basic display driver to the full graphics driver, several things change at once. The driver takes control of the card's clocking, memory, and output configuration. It applies the resolution, refresh rate, and colour settings stored from the previous session. It also negotiates the link with the monitor again, which can involve a different display mode than the one used during boot.
A black screen at this stage often appears exactly at the moment the login screen should appear, or during the transition from the login screen to the desktop. The picture may vanish cleanly, with no error message, because the display mode negotiation itself has failed or because the driver has stopped responding while taking over. A driver that crashes during load can leave the monitor without a signal even though Windows itself continues to boot in the background.
It is important to understand what a normal pre-login display does and does not prove. It proves that the basic display path works. It does not prove that the full driver loads correctly. The basic driver and the full driver exercise the card differently: the full driver uses more of the card's features, applies stored settings, and manages power states that the basic driver never touches. A card that displays the boot screen perfectly can still fail at the handover, and that failure does not by itself mean the card is dead.
Stage Two: Desktop Composition Takeover
Once the full driver is loaded, Windows hands the visible desktop to its composition system. Composition is what draws windows, transparency, animations, and the taskbar as a single composed picture. It also manages multiple displays, scaling, and high dynamic range settings where applicable.
A fault at this stage tends to appear slightly later than a driver handover fault: the login screen may appear, the password may be accepted, and then the desktop either never appears or appears for a moment before going black. The mouse cursor may remain visible on a black background, which is a useful clue. A visible cursor means the display pipeline is still producing an output and the composition layer is at least partly alive, even though the desktop itself is not being drawn.
Composition faults are often tied to stored display settings rather than to the card itself. A resolution or refresh rate that the monitor cannot display, a scaling setting carried over from a different monitor, or a multi-display arrangement that no longer matches the connected hardware can all leave composition drawing a picture that never reaches the screen. This is why a machine that worked yesterday can fail today after a monitor swap, a cable change, or a settings change that seemed harmless at the time.
Stage Three: Post-Login Startup Load
The third stage begins once the desktop is up. Programs that start automatically at login begin to load: system utilities, display-related tools, overlay software, screen capture tools, and third-party services. Some of these programs interact directly with the graphics driver or with the display pipeline. A conflicting program can crash the driver, force a display mode change, or seize the display output for itself.
A fault at this stage has a distinctive timing signature. The desktop appears normally, and then, seconds later, the screen goes black. The delay corresponds to the startup program loading. If the screen goes black immediately at login, the driver handover or composition is more likely. If the desktop is visible for a while first, the startup load deserves closer attention.
Second displays add a further variation to this stage. A program or setting that moves the desktop, a window, or the primary display designation to a second monitor can make the main screen look dead while the picture is actually being shown elsewhere. This is not a hardware failure at all; it is a display arrangement problem, and it is easy to miss if the second display is powered off, disconnected from view, or sitting on a different input.
Test One: The Safe Mode Display Check
Safe Mode is the single most informative test for this symptom, because it changes the display stack in a controlled way. In Safe Mode, Windows loads a minimal set of drivers and services, uses a basic display driver, and does not load the normal startup programs. If the picture is stable in Safe Mode, the basic display path and the card's basic output are working, and the fault lies in something that Safe Mode leaves out: the full driver, a startup program, or a stored display setting.
To run this test, boot into Safe Mode and observe the display for several minutes. Note whether the picture is stable, whether the resolution looks low, and whether the desktop appears at all. A stable Safe Mode picture is a strong result. It does not prove the card is healthy in every respect, but it does rule out the simplest explanations: a completely dead output, a cable that cannot carry any signal, and a monitor that cannot display anything.
If Safe Mode is also black, the diagnosis changes direction entirely. A black screen in Safe Mode means the fault is not confined to the full driver or to startup programs, because Safe Mode bypasses both. At that point this guide no longer applies. The problem may sit earlier in the chain, and the reader should turn to a guide covering no display from power-on, including the boot screen itself. Forcing the tests in this guide onto a machine that is black even in Safe Mode will waste time and can mislead the diagnosis.
Record the Safe Mode result before moving on. Write down whether the picture appeared, how long it remained stable, and whether anything unusual happened, such as flickering or a resolution that looked wrong. These notes become the reference point for every later comparison.
Test Two: Clean-Boot Comparison
If Safe Mode displays correctly, the next step is to separate the full driver from the startup programs. A clean boot does this by starting Windows normally, with the full driver loaded, but with non-essential startup items and third-party services disabled. If the picture is stable in a clean boot but fails in a normal boot, a startup program is implicated. If the picture fails in a clean boot as well, the driver or a stored display setting is more likely.
The value of this test lies in its control. Between a clean boot and a normal boot, the driver is the same and the hardware is the same. The only difference is the set of programs that load after login. A change in outcome across that single difference points at the startup load, not at the card.
When running this comparison, keep everything else identical: the same monitor, the same cable, the same display input, and the same login account. Changing the monitor at the same time as disabling startup items would destroy the comparison, because two variables would have changed at once and the result could not be attributed to either one.
If a clean boot points at a startup program, the follow-up is methodical rather than dramatic. Re-enable startup items in small groups, rebooting after each group, and note which group brings the black screen back. Display-related utilities, overlay tools, and screen management programs are sensible early candidates, but the grouping method matters more than guessing the culprit in advance. The goal is an observation — this group changes the outcome — not a theory about which program is at fault.
Test Three: Single-Display Login
Multi-display arrangements are a common and underappreciated cause of a post-login black screen. The machine may be outputting a picture, but to a display that is not the one being watched. A second monitor that is powered off, set to the wrong input, or disconnected from view can silently receive the primary desktop while the main screen shows nothing.
The test is simple: disconnect every display except one, and log in with that single display connected. Use the display and cable that are known to show the boot screen, so the early path is already verified. If the picture returns with a single display, the fault lies in the multi-display arrangement, not in the card or the driver. The follow-up then involves reconnecting the second display, checking the display arrangement settings, and confirming which display Windows treats as primary.
If the picture remains black with a single display, the multi-display arrangement is ruled out, and the diagnosis returns to the driver handover, composition, or startup load. Either way, the test has removed a whole category of explanations with a single controlled change, and that is progress even when the result is negative.
Pay attention to display inputs during this test. A monitor with multiple inputs can sit on the wrong input while the machine outputs correctly on another. Cycling through the monitor's inputs, and confirming the input matches the connected cable, is part of the test, not an afterthought.
Test Four: External Monitor or Television Comparison
When the machine is a laptop, or when a spare display is available, an external monitor or television provides an independent display path. Connect the external display, and observe whether it shows the login screen and desktop while the built-in or primary display does not, or vice versa.
This comparison separates the display device from the machine's graphics output. If the external display works while the primary does not, the fault is more likely tied to the primary display, its cable, its input setting, or a stored setting that applies to that display. If both displays go black at the same point, the fault is more likely in the shared driver or composition layer, because two independent display devices failing at the same stage points at the common cause upstream of both.
For a desktop machine, the same logic applies with a second monitor or a television. The point is not to find a display that works and stop there; it is to compare two displays under identical conditions and note where their behaviour diverges. Divergence is information. Identical failure is also information, because it points upstream.
Test Five: Event Viewer Entries as Clues Only
Windows Event Viewer can contain entries relating to driver crashes, display driver restarts, and application faults around the time of the black screen. These entries can be useful clues. They can suggest which driver or program was involved, and they can confirm the timing of a crash relative to login.
They must never be treated as proof on their own. An event entry records that something was logged, not that it caused the black screen. Driver entries can appear as a consequence of a crash rather than its cause. Application entries can coincide with the black screen without being related to it. Treating a single log entry as a verdict is one of the most common diagnostic errors in display troubleshooting.
Use Event Viewer the way a detective uses a witness statement: as a lead to check against the controlled tests above. If an entry points at a particular driver version, that is a reason to examine the driver situation carefully, not a reason to replace hardware. If an entry points at a startup program, that is a reason to include that program in the clean-boot grouping, not a reason to declare it guilty. The controlled tests decide; the logs merely suggest where to look.
Reading the Results Together
No single test in this guide gives a complete answer. The diagnosis comes from the pattern across tests. The table below summarises how the main patterns map to the most likely stage, and what to examine next in each case.
| Test pattern | Most likely stage | What to examine next |
|---|---|---|
| Boot screen normal, Safe Mode stable, clean boot stable, normal boot black | Post-login startup program | Startup groups, display-related utilities, overlay tools |
| Boot screen normal, Safe Mode stable, clean boot black | Driver handover or stored display setting | Driver state, stored resolution and refresh rate, composition settings |
| Boot screen normal, single display works, multi-display black | Display arrangement | Primary display designation, second display input and power state |
| External display works, primary display black at the same stage | Primary display path or its stored setting | Primary cable, input selection, per-display settings |
| Both displays black at the same stage | Shared driver or composition layer | Driver state, composition settings, startup load |
| Safe Mode also black | Outside this guide | No-display-from-power-on diagnosis, including boot screen faults |
| Desktop visible briefly, then black after a delay | Post-login startup load | Startup programs that load after the desktop appears |
A pattern that does not fit the table neatly is still useful. It means the fault may involve more than one stage, or that a variable changed between tests. In that case, repeat the tests with stricter controls: same monitor, same cable, same account, one change at a time, and written notes for each result.
Common Misconceptions
Several assumptions lead this diagnosis astray, and each is worth naming plainly.
The first is that a black screen after login means a dead card. It does not. A dead card would generally fail to produce the boot screen as well. When the boot screen is normal and Safe Mode is stable, the card is demonstrably capable of producing a picture. The fault lies in a later stage, and replacing the card would not address that stage.
The second is that the answer is automatically a new cable or a new card. Cables and cards are easy to blame because they are easy to replace, but the tests above exist precisely to avoid that kind of guessing. If the boot screen is stable on the existing cable, the cable is carrying a signal. A cable fault that appears only after login, and only with the full driver loaded, is possible but far less likely than a driver, composition, or startup explanation, and it should only be considered after the controlled tests point that way.
The third is that a normal pre-login display clears the full driver. It does not. The basic display driver and the full driver are different software exercising the card differently. Clearing the basic path says nothing about the handover, and the handover is exactly where this symptom lives.
The fourth is that Event Viewer provides a verdict. It provides clues. A log entry without a matching controlled test result is a hypothesis, not a conclusion.
When to Stop Using This Guide
This guide has a defined scope, and recognising its edge is part of using it well. If Safe Mode is also black, stop here and divert to a no-display diagnosis that covers the boot screen itself. If the boot screen is missing, intermittent, or garbled from power-on, the fault sits before the stages this guide addresses, and the tests here will not isolate it. If the picture fails only under heavy graphical load, such as during games or rendering, that is a different symptom with a different diagnostic path, and it should not be forced into this guide's framework.
Staying inside the guide's scope is not a limitation; it is what makes the guide useful. A diagnostic method that claims to cover every black screen covers none of them well. The boundary — firmware display works, post-login display fails — is the condition that makes these five tests meaningful.
A Sensible Order of Work
For a reader facing this symptom now, the order of work is straightforward. First, confirm the exact symptom: boot screen normal, black at or after login. Second, run the Safe Mode check and record the result. Third, if Safe Mode is stable, run the clean-boot comparison. Fourth, run the single-display login. Fifth, compare with an external display where possible. Sixth, check Event Viewer for clues that match the test pattern, and treat those clues as leads rather than verdicts. At each step, change one variable, record the result, and let the pattern across results guide the next step.
Patience matters more than speed here. Display faults invite quick part swaps because swapping feels like progress. The controlled tests in this guide are slower, but they produce an answer that a part swap cannot: not just a working machine, but an understanding of which stage failed, and therefore which change actually fixed it.
Frequently Asked Questions
Why does the screen work during boot but go black after login?
Because the boot screen and the desktop use different display paths. The boot screen uses the firmware and basic display path, while the desktop uses the full graphics driver, desktop composition, and startup programs. A fault in any of those later stages leaves the boot screen untouched and breaks only the post-login picture.
Does a normal boot screen mean the graphics card is fine?
No. It means the card's basic output path works. The full driver exercises the card differently and applies stored settings that the basic path never uses. A card can display the boot screen and still fail at the driver handover. The Safe Mode and clean-boot tests narrow down whether the fault sits in the driver, composition, or startup load.
What does it mean if Safe Mode works but normal mode does not?
It means the basic display path works and the fault lies in something Safe Mode leaves out: the full driver, a startup program, or a stored display setting. The clean-boot comparison is the next step, because it separates the full driver from the startup programs while keeping everything else the same.
What does it mean if Safe Mode is also black?
It means the fault is not confined to the full driver or startup programs, and this guide no longer applies. The diagnosis should move to a no-display guide covering faults from power-on, including the boot screen itself, rather than forcing the tests here onto a symptom they were not designed for.
Could a second monitor be the cause even if it is switched off?
Yes. Windows can designate a powered-off or wrongly-input second display as the primary display, sending the desktop there while the watched screen shows nothing. Disconnecting all but one display and logging in again is the direct test for this, and it costs nothing to run.
Should Event Viewer entries be treated as proof of the cause?
No. Event Viewer entries are clues, not verdicts. A driver entry can be a consequence of a crash rather than its cause, and an application entry can coincide with the black screen without being related. Use log entries to guide the controlled tests, and let the test pattern decide.
Is replacing the cable or the card a reasonable first step?
Not for this symptom. When the boot screen is stable, the existing cable is already carrying a signal, and the card is already producing a picture. Replacing parts before running the Safe Mode, clean-boot, and single-display tests is guessing, and it can mask the real fault while adding unnecessary cost and confusion.
What should I record during these tests?
Record the display used, the cable and input, the login account, and the result of each test in plain language: whether the picture appeared, when it disappeared, and whether a cursor remained visible. Consistent notes across tests are what turn isolated observations into a pattern that points at one stage.