Diagnosing a GPU or PSU fault is one of those problems where the symptoms overlap just enough to send you in completely the wrong direction. Black screens, sudden shutdowns, driver errors, the PC rebooting mid-game. All of those can come from either component. The difference is in the detail, specifically in what Windows logs at the moment of failure, and in how the crash actually behaves. This guide walks through every layer of that diagnosis, from a two-minute Event Viewer check right through to hardware isolation.
What Causes a GPU or PSU Fault?
The reason this diagnosis is tricky is that both components fail under the same conditions: high load, sustained power draw, and heat. So the system that crashes during a gaming session could be telling you one of several very different things.
The most common cause of GPU-side crashes is a driver fault or timeout. Windows logs these as Display driver stopped responding and has recovered, or as LiveKernelEvent 141 or 117 in the System log. The GPU essentially stops responding to the OS, Windows tries to recover the display driver, and either succeeds (you see the screen go black briefly then come back) or fails (the system reboots or you get a blue screen). These are software faults a lot of the time, but they can also indicate failing VRAM, a dodgy PCIe connection, or a GPU that's genuinely on its way out.
PSU faults behave differently. When a power supply can't sustain the load, it doesn't give Windows time to log a graceful shutdown. The machine just cuts out, as if someone pulled the plug. Windows notices this on the next boot and logs a Kernel-Power Event ID 41, which means the system rebooted without a clean shutdown. Event ID 6008 is the companion entry, logging the unexpected shutdown time. These are the two IDs you're looking for when you suspect the PSU.
Overheating is a third possibility that gets misread as either a GPU or PSU fault. A GPU that's thermal throttling or hitting its shutdown temperature will produce WHEA-Logger entries and sometimes Display 4101 events. Clean the fans, check GPU temperatures with a tool like GPU-Z, and rule this out before spending money on replacement hardware.
And then there's the catch-all: RAM, motherboard instability, or OS corruption that mimics GPU or PSU symptoms. SFC and DISM exist precisely to catch this category. Run them before you do anything with hardware.
GPU or PSU: Quick Diagnostic Check
Check Device Manager and Event Viewer Easy
- Open Device Manager
PressWin + Xand choose Device Manager. Expand Display adapters. Look for a yellow warning triangle, a Code 43 or Code 31 error, or a missing discrete GPU. If your dedicated GPU has disappeared from the list entirely and only shows the integrated graphics, that's a significant GPU hardware signal. Double-click the GPU and read the Device status on the General tab. - Open Event Viewer
PressWin + R, typeeventvwr.msc, press Enter. Go to Windows Logs then System. Scroll to the time of your last crash. You're looking for three things: Kernel-Power 41 or 6008 (unexpected power loss, PSU suspect), Display 4101 (GPU driver timeout), and WHEA-Logger or PCIe errors (hardware-level fault on the GPU or board). - Observe the next crash
Trigger the fault deliberately if you can. Open a GPU-heavy game or run a benchmark. Watch what happens. If only the application crashes or the screen goes black for a few seconds and then recovers while the PC stays on, that's a GPU or driver fault. If the entire machine instantly powers off with no warning, no blue screen, nothing, that's a PSU or power hardware fault. The distinction is that clean.
This quick check takes about five minutes and already narrows the field significantly. Most people who've been chasing this problem for days haven't looked at Event Viewer at all, which is a shame because it usually tells you exactly what happened. For context on how Microsoft defines these events, Microsoft's own documentation on Event ID 41 explains the conditions that trigger it and what the BugcheckCode values mean.
GPU or PSU: Intermediate Fixes
Clean Reinstall the GPU Driver Intermediate
- Download the driver first
Go to NVIDIA's driver page, AMD's support site, or Intel's download centre depending on your GPU. Download the latest stable driver package for your exact GPU model and save it somewhere you can find it after the reboot. - Uninstall the current driver
Open Device Manager, expand Display adapters, right-click your GPU and choose Uninstall device. Tick the box that says Delete the driver software for this device if it appears. This removes the installed driver files, not just the device entry. Reboot when prompted. - Install the downloaded driver
Run the installer you downloaded. Choose a clean install option if offered (NVIDIA's installer has this explicitly). This overwrites any corrupted driver state. Reboot again after installation. This one often needs a second reboot before the GPU is fully stable, so don't test immediately after the first restart. - Retest under load
Run the same workload that previously caused the crash. Check Event Viewer afterwards for any new Display 4101 or Kernel-Power 41 entries. If the crash is gone, the fault was a software driver issue. If it persists, the driver is not the root cause and you need to go deeper.
Compare Behaviour Under Different Loads Intermediate
- Run a CPU-only workload
Copy a large folder of files between drives, or run a CPU benchmark. This keeps the GPU mostly idle. If the system crashes or shuts off during this test, the fault is not GPU-specific. It points to the PSU, RAM, or motherboard. If it stays stable, the fault is load-dependent on the GPU. - Run a GPU-heavy workload
Now run something that hammers the GPU: a demanding game, a 3D render, or a GPU stress test. Note the crash behaviour carefully. Does the display go black and recover? Does the whole machine cut out? Does it blue screen? Each of these has a different implication for whether it's a GPU or PSU fault. - Check power plan settings
Right-click Start and open Power Options. Make sure the plan is set to Balanced or High Performance. Aggressive power-saving settings can interfere with PCIe power states and cause the GPU to drop out in ways that look like hardware failure but aren't. Set sensible display and sleep timers so you can distinguish a real crash from a wake-from-sleep failure.
If you're getting crashes that seem to happen randomly, even at idle, that changes the picture entirely. Random idle crashes with Kernel-Power 41 can indicate a PSU that's failing at a basic level, not just under load. That's a more urgent situation. For related instability problems that can look similar, see our guide to random Windows restarts which covers RAM and motherboard causes in more detail.
Advanced GPU or PSU Diagnosis
Run DISM and SFC to Rule Out OS Corruption Intermediate
- Open an elevated Command Prompt
Search for Command Prompt in Start, right-click it, and choose Run as administrator. - Run DISM first
Typedism /online /cleanup-image /restorehealthand press Enter. This contacts Windows Update to download and replace any corrupted system files. It can take 10 to 20 minutes. Let it finish completely before doing anything else. - Run SFC
Once DISM completes, typesfc /scannowand press Enter. SFC checks protected system files and replaces any it finds corrupted using the repaired component store that DISM just fixed. Reboot after this completes, then retest under load.
Filter Event Viewer for Precise Fault Identification Advanced
- Open Event Viewer and filter the System log
Openeventvwr.msc, right-click Windows Logs then System, and choose Filter Current Log. In the Event IDs field, enter41, 6008, 4101. This filters to only the events that matter for a GPU or PSU fault diagnosis. - Add WHEA-Logger as a source
In the same filter dialog, you can also filter by Event source. Add WHEA-Logger to catch hardware-level PCIe and memory errors. These often appear just before a Kernel-Power 41 entry and indicate the GPU or its PCIe slot is producing hardware errors, not just driver timeouts. - Correlate timestamps
Reproduce the fault under load, then immediately check the filtered log. Match the timestamp of the crash to the nearest event entries. The sequence matters: a Display 4101 followed by a Kernel-Power 41 suggests the GPU hung first and the system then lost power control. A Kernel-Power 41 with no preceding Display events suggests the PSU cut out independently.
Hardware Isolation: GPU and PSU Swap Testing Advanced
- Reseat the GPU and PCIe power cables first
Power off completely and unplug from the mains. Remove the GPU, clean the PCIe slot contacts gently, and reseat the card firmly. Reconnect both ends of the PCIe power cables. A loose cable at the PSU end is a surprisingly common cause of load-dependent shutdowns that look exactly like a PSU failure but aren't. - Test the GPU in another system
If you have access to another PC with a compatible PCIe slot and a PSU of sufficient wattage, fit your GPU there and run the same load test. If the crashes follow the GPU to the new system, the GPU hardware is faulty. If the new system runs fine, the fault is in your original machine, pointing back to the PSU, motherboard, or PCIe slot. - Test with a known-good PSU
Borrow or source a quality PSU of adequate wattage for your GPU. Tom's Hardware's GPU hierarchy and power draw data is useful for checking what wattage your GPU actually needs under load. Fit the replacement PSU and run the same load test. If shutdowns stop, the original PSU was the fault. - Reduce to minimal configuration
If swapping isn't an option, disconnect non-essential drives, remove extra PCIe cards, and test with a single RAM stick. This reduces total system power draw and can expose a PSU that's marginal rather than fully failed. A system that stabilises with fewer components attached is a strong indicator the PSU is undersized or degraded.
If you're not comfortable doing hardware swaps yourself, or you don't have access to spare components to test with, this is exactly the kind of diagnosis we handle remotely for the software side and can advise on in detail for the hardware side. For blue screen errors that appear alongside these crashes, our BSOD diagnosis guide covers the stop codes that commonly appear with GPU and PCIe faults.
Preventing GPU or PSU Faults
Most of this is unglamorous but it works. Start with the PSU. An undersized or cheap power supply is the single most common cause of load-dependent shutdowns in systems with mid-range to high-end GPUs. Size your PSU to your GPU's actual power draw, not the marketing spec on the box. Add 20 to 30 percent headroom above the GPU's TDP plus the rest of your system's draw, and buy from a brand with a decent reputation for consistent rail voltage.
Keep GPU drivers current, but don't chase every new release. NVIDIA and AMD both occasionally push driver versions that introduce instability, and rolling back one version is sometimes the right call. Stick to official sources, never third-party driver sites, and use the clean install option when updating.
Dust is underestimated. A GPU running 20 degrees hotter than it should because the heatsink is clogged will produce WHEA errors and thermal shutdowns that look like hardware failure. Clean the system every three to six months, especially if it's in a dusty environment or on a carpet.
Run dism /online /cleanup-image /restorehealth and sfc /scannow every few months as routine maintenance. OS corruption building up quietly over time is a real thing, and catching it early prevents it from being misread as a GPU or PSU fault later. Use surge protection on the mains feed, particularly if you're in an area with variable power quality. Voltage spikes stress PSU components and shorten their lifespan noticeably.
GPU or PSU: Summary
Separating a GPU or PSU fault comes down to three things: how the crash behaves (display recovery versus full power cut), what Event Viewer logs at the moment of failure (Display 4101 versus Kernel-Power 41), and whether the fault follows the component to another system. Start with the quick Event Viewer check because it costs nothing and often gives you a clear answer. Clean reinstall the GPU driver before touching any hardware, because a software fix is always preferable to a component swap. If software changes nothing and the logs consistently show Kernel-Power 41 under GPU load, the PSU is the most likely culprit. If Display 4101 and LiveKernelEvent entries dominate with no power events, focus on the GPU. And if you've exhausted software options, hardware isolation is the only way to close the case on a GPU or PSU fault definitively.