GPU Keeps Rebooting or Black-Screening? How to Tell a Failing PSU from a Faulty Graphics Card
Key Takeaways
A graphics card that reboots the PC or black-screens under load is not automatically faulty. An ageing or undersized power supply can produce identical symptoms when its +12V output sags under sudden load spikes. This guide explains PSU power delivery, how PSU-side and card-side symptoms tend to differ, and which safe checks to run before sending a card for repair.
Introduction: The Game Crashes, and the Card Gets the Blame
The scene is familiar: a game is running, the screen suddenly goes black, and the whole PC restarts. Or the display cuts out the moment a 3D application loads, while the desktop was perfectly fine a minute earlier. The first suspect is usually the graphics card — and sometimes that suspicion is correct. But a failing or undersized power supply unit (PSU) can produce exactly the same symptoms, and replacing or sending the card for repair will not fix a problem that lives in the power supply.
This guide explains how a PSU feeds a graphics card, why power supplies weaken with age, and how PSU-side and card-side symptoms tend to differ. The goal is not to turn the reader into a repair technician. It is to help the reader narrow down which part is suspect before spending money on a repair that may not be needed — and to know which checks are safe to perform at home and which must be left alone.
How a PSU Powers a Graphics Card
A computer PSU converts mains alternating current into the direct current that components use. It produces several output rails, principally +12V, +5V, and +3.3V. The graphics card draws the great majority of its power from the +12V rail, so the health of that rail matters more to the card than any other output.
A graphics card receives power from two sources:
| Power source | What it is | Rated delivery |
|---|---|---|
| PCIe slot | Power delivered through the motherboard slot the card sits in | Rated up to 75 W under the PCIe specification |
| Auxiliary connectors | Dedicated cables from the PSU that plug directly into the card | 6-pin rated up to 75 W; 8-pin rated up to 150 W; 12VHPWR / 12V-2x6 rated up to 600 W |
These figures are specification ratings, not measured values from any particular card or PSU. A card that needs more power than the slot alone can provide must have its auxiliary connectors fitted; a missing or loose auxiliary cable is one of the simplest causes of load-related instability.
Load Spikes: Why a PC Can Be Fine at Idle and Fail in Games
GPU power draw is not steady. At the desktop, a card may draw only a small amount of power. The moment a game or 3D workload starts, current demand can rise sharply within a very short time. These sudden rises are called transient load spikes, and they are the hardest part of the PSU's job.
A healthy PSU holds its +12V output steady through those spikes. A weak one lets the voltage dip momentarily. If the dip is deep enough, the graphics card — or the whole system — can reset, producing a black screen, a reboot, or a crash that looks exactly like a card fault. This is why a PC can be completely stable at idle and fail within seconds of entering a game: idle never asks the PSU the question that load asks.
Why a PSU Loses Its Ability to Deliver
A PSU does not usually fail in one dramatic moment. It degrades gradually:
- Capacitor ageing. Sustained high temperatures inside a PSU age its internal capacitors. Aged capacitors hold their charge less effectively, so output voltage becomes less stable and ripple — small unwanted fluctuations on top of the DC output — grows larger.
- Heat and dust. A dust-clogged PSU fan, a case with poor airflow, or a PSU mounted where it breathes hot air all raise internal temperatures and accelerate ageing.
- Insufficient headroom. A PSU sized with little margin above the system's peak demand has to work near its limits whenever the GPU is loaded. Running near the limit continuously leaves no room for transient spikes and shortens component life.
- Cables and connectors. A cable that has been bent sharply, pinched, or partially pulled out of its socket adds resistance and heat at the connection. The 12VHPWR connector family is particularly sensitive to incomplete seating, which is why connector condition is checked before any other conclusion.
Two common PSU designs deserve a brief note. A single-rail PSU provides one +12V output for the whole system. A multi-rail PSU splits the +12V output into several rails, each with its own over-current protection. Each design has trade-offs in protection behaviour and load sharing, and neither is categorically better than the other. The design label alone does not diagnose a fault.
Telling Them Apart: PSU Problem vs Graphics Card Problem
The table below compares symptom tendencies. Every entry is a tendency, not proof. The two categories can overlap, and both faults can exist at the same time.
| Symptom | Tends to point toward the PSU | Tends to point toward the graphics card | How to confirm |
|---|---|---|---|
| Whole PC restarts or powers off under load | Yes — the system loses power, not just the picture | Less typical | Check whether the restart happens only when load rises |
| Screen goes black but fans and lights stay on | Possible, if the card resets on a voltage dip | Possible, if the card or its driver stops outputting | Note whether the system itself is still running (sound, network activity) |
| Visual artifacts, coloured blocks, garbled textures | Unlikely as the primary cause | Yes — damaged output or memory tends to corrupt the picture | Take a screenshot: corruption present in the screenshot points toward the card |
| Stable at idle, fails the moment a game loads | Yes — matches transient-spike behaviour | Possible — load also stresses the card itself | Test with a known-good PSU if one is available |
| Failure follows a specific picture or scene | Less typical | Yes — repeatable corruption tied to content suggests the card | Check whether the same scene fails every time |
| Burning smell or visibly scorched connector | Possible at the connector or cable | Possible at the card's connector | Stop using the system and inspect the connector and cable before powering on again |
A useful rule of thumb: if the picture itself is corrupted, the card is the more likely suspect. If the whole machine loses power or resets with a clean picture right up to the failure, the PSU is the more likely suspect. Neither pattern is conclusive on its own.
Safe Pre-Repair Checks
These checks involve no disassembly of the PSU and no contact with live internal parts:
- Reseat every power connector. With the PC powered off and unplugged, unplug and firmly replug each auxiliary power cable at both the card end and the PSU end (on modular PSUs). A connector that clicks fully home is a connector that is seated.
- Inspect cables and connectors. Look for discolouration, melted plastic, a burnt smell, bent pins, or a connector that feels loose. Any of these is a reason to stop using that cable and seek advice.
- Match PSU wattage against the card's recommended supply. Check the graphics card maker's published recommended PSU wattage for the card, and compare it with the PSU's rated output and the rest of the system. A PSU well below the recommendation is a suspect, not a certainty.
- Cross-test if a known-good PSU is available. Swapping in a PSU already known to be healthy is the most decisive home test. If the symptoms vanish with a different PSU, the original PSU is the likely cause.
- Reduce the load temporarily. Running the card at a lower power limit or with the game at lighter settings is a diagnostic observation, not a fix. If symptoms ease when load drops, the power delivery chain — PSU, cables, or the card's own power circuitry — is implicated.
Never open the PSU casing. A PSU contains high-voltage capacitors that can hold a dangerous charge long after the unit is unplugged. There is no user-serviceable part inside. Do not open it, modify it, or attempt to short-start it. If the PSU itself is suspected, it is replaced or serviced as a sealed unit.
Prevention
- Leave power headroom: choose a PSU whose rated output sits comfortably above the system's peak demand, including transient spikes, rather than exactly matching it.
- Keep the case ventilated and clean dust from intake filters and fans so the PSU runs cool.
- Avoid running the card long-term through splitters, adapters, or one cable split across multiple connectors. Each adapter is an extra point of resistance and a potential failure point.
- Seat connectors fully when installing or moving hardware, and re-check them after transporting the PC.
FAQ
Q: The PC only restarts after I installed a new graphics card. Is the PSU too weak?
A: It is a strong candidate. A new, more powerful card raises peak demand and transient spikes, so a PSU that coped with the old card may not cope with the new one. Check the card maker's recommended PSU wattage, verify all auxiliary connectors are seated, and cross-test with a known-good PSU if possible before concluding the card is faulty.
Q: Is a higher-wattage PSU always better?
A: Not necessarily. A larger PSU provides more headroom, which is useful, but wattage alone does not by itself ensure stable output. Build quality, voltage stability under transient load, and correct cabling matter as much as the headline wattage figure. An oversized but poor-quality unit is not automatically an improvement.
Q: Why is the PC fine at idle but restarts the moment a game starts?
A: Idle places almost no demand on the PSU. A game raises current demand suddenly, and if the PSU's +12V output dips far enough during that spike, the card or the system resets. This pattern points toward the power delivery chain, but it does not by itself distinguish a weak PSU from a card-side power fault — cross-testing is what separates them.
Q: Can a failing PSU damage other parts?
A: An unstable supply can make the system unreliable and can add stress to other components over time. That is a reason to investigate PSU-side symptoms promptly rather than ignoring repeated restarts. It is not a reason to assume damage has already occurred: repeated instability should be diagnosed, not panicked over.
Q: My card's connector looks slightly discoloured. What should I do?
A: Stop using that cable and connector combination and have it inspected. Discolouration or a burnt smell at a power connector indicates overheating at that connection. Do not continue running the system on a damaged connector in the hope that it will hold.
Q: When should I send the card for repair instead?
A: If the PSU, cables, and connectors have been checked or cross-tested and the symptoms persist — especially picture corruption, repeatable artifacts, or failures that follow specific content — the card itself becomes the primary suspect. At that point, describe the checks already performed when arranging the repair, since that information helps narrow the diagnosis.