How to Check If a GPU Is Working Properly

We will show you how to check if a GPU is working properly by letting you watch three things: its temperature, how it behaves under load, and what it puts on screen. A graphics card is one of the most expensive parts of a PC, and a big part of a laptop's price too, so when something looks wrong it pays to confirm whether the card is actually faulty before you spend on a replacement.

A faulty GPU can cause game crashes, show strange visual artifacts, or could be running too hot. Spotting the signs early lets you rule out a simple fix before you assume the card is dead. A working card sits at 140-185°F (60-85°C) while gaming, runs benchmarks without stuttering, and renders everything clean.

You can check your graphics card health using Windows Task Manager to see GPU usage. Use MSI Afterburner or HWMonitor to watch temperatures. Run stress tests like 3DMark to check for problems. A card that's working properly will show steady temps, no visual glitches, and benchmark scores within 10% of what other cards like yours get.

The most common GPU failure is no POST (Power-On Self-Test), which means your computer turns on but shows nothing on screen. Heat problems are trickier to spot. The system starts up fine, but once the GPU gets hot, you might see visual glitches, crashes, or a blank screen.

1. How to Check GPU Health Quickly

Before running detailed tests, start with the basics. Make sure your GPU's power cables are pushed in all the way, as loose cables are a very common cause of crashes and blank screens. If you're having display problems, try a different video cable or port before deciding the card itself is broken.

It's also worth checking for driver updates on NVIDIA, AMD, or Intel's website, since old drivers cause problems that look like the hardware is broken.

You can also open Device Manager (right-click Start, select Device Manager) and expand Display adapters to confirm Windows sees your GPU; a yellow warning icon or a missing entry there points to a driver or connection problem before you go any further. To start, expect 86-104°F (30-40°C) at idle and 140-185°F (60-85°C) while gaming.

2. Using System Monitoring Tools

Built-in Windows tools can quickly show if your computer sees your GPU and if it's working right.

Windows Task Manager GPU Monitoring

Press Ctrl + Shift + Esc to open Task Manager, then click the Performance tab and select your GPU. Start a game or another graphics-heavy program and watch the 3D graph. Microsoft explains that Task Manager reads GPU activity from the Windows graphics scheduler, so a rising graph shows that Windows is sending work to that GPU.

Detailed GPU Monitoring Tool Setup

These free programs let you watch your GPU's health in detail. Here's how to use each one:

MSI Afterburner (Recommended for All Graphics Cards)

1. Download and Install: Go to MSI's website and download the latest version of Afterburner. It comes with RivaTuner, which shows stats on your screen while gaming.
2. Initial Setup: Open MSI Afterburner and click the gear icon to open settings.
3. Enable GPU Monitoring: Click the Monitoring tab and check the boxes for GPU temperature, GPU usage, video memory usage, clock speed, and fan speed.
4. On-Screen Display: Turn on "Show in On-Screen Display" for the stats you want to see while gaming.
5. Key Numbers to Watch:
   • GPU Usage: Should reach 95-100% during heavy gaming
   • Clock Speed: Should match what your card is rated for
   • Power Draw: Steady power use means your GPU is working well

GPU-Z (Detailed Graphics Card Information Tool)

1. Download GPU-Z: Get it free from TechPowerUp's website.
2. Graphics Card Tab: Make sure all the specs match what your GPU should have.
3. Sensors Tab: Watch your GPU stats in real time:
   • GPU Load shows how hard your card is working
   • Memory Controller Load shows how busy the video memory is
   • PerfCap Reason shows what is limiting performance; 'Pwr' under heavy load is normal and means the card is hitting its power limit; 'Temp' means thermal throttling is active and needs attention
4. GPU Validation: Use GPU-Z validation or compare the reported specs against TechPowerUp's GPU database to make sure your graphics card isn't fake.

HWiNFO64 (Complete System Monitoring)

1. Installation: Download from HWiNFO.com and install the program.
2. Sensors-Only Mode: Open HWiNFO and pick "Sensors-only" to see live stats.
3. Find Your GPU: Look for your graphics card in the sensor list.
4. Key Numbers to Track:
   • Temperature Sensors (check the core, hotspot, and memory temps)
   • Power Use (compare to what your card is rated for)
   • Thermal Throttling (should say "No" during normal gaming)
   • Memory Errors (should always be zero)
5. Temperature Logging: Turn on logging to track temperature changes over time.

3. How to Make Sure Your Dedicated GPU Is Being Used

A computer with integrated and dedicated graphics may show both in Windows. Seeing the dedicated card in Device Manager only proves that Windows detects it. Use Task Manager to confirm that a game or program is sending work to that card.

1. Find the GPU numbers: Open Task Manager, select Performance, and note the name beside GPU 0, GPU 1, or each later number.
2. Check the program: Select Processes, right-click a column heading, and turn on the GPU and GPU engine columns. Start the program you want to test. The GPU engine entry identifies the GPU number and the engine that the program is using. Microsoft's Task Manager guide explains how these per-process columns work.
3. Choose the high-performance GPU if needed: In Windows 11, open Settings, System, Display, and Graphics. Select the program, choose Options, select High performance, save the change, and restart the program. These steps follow Microsoft's graphics preference instructions.
4. Check the desktop display cable: Connect the monitor to a port on the graphics card, not a motherboard video port. AMD also recommends the discrete card's display output for systems that have both integrated and dedicated graphics.

4. How to Check If Your GPU Fans Are Working

A stopped fan at the Windows desktop does not always mean it is broken. Some graphics cards use a zero-RPM mode that stops the fans at low temperatures and starts them when the card gets hotter. ASUS describes this behavior in its graphics card fan troubleshooting guide. Test the fans under load before judging them.

1. Watch the sensors: Open MSI Afterburner, GPU-Z, or HWiNFO and find the fan speed reading. Leave the sensor window open.
2. Add a safe load: Start a game or one of the benchmarks in the next section. Watch the temperature and fan reading as the card warms up.
3. Look at every fan: Each fan should start smoothly when the card reaches its own fan-start point. The exact point depends on the model, so use the card maker's specifications instead of one temperature for every GPU.
4. Try manual control: If the card supports it, briefly raise the fan setting in Afterburner and then return it to Auto. MSI's Afterburner instructions show where to enable custom fan control. MSI notes that laptop GPU fan curves are usually controlled by the manufacturer and cannot be changed there.

5. How to Test a Graphics Card with Benchmarks

Test your graphics card's performance and stability with these benchmark programs.

Recommended GPU Benchmarking Tools

3DMark is the most widely used benchmark for gaming performance and GPU stress testing. For free stability testing, Heaven and Valley from Unigine are solid choices; run them for at least 30 minutes to bring out heat problems. PassMark lets you compare your card's score against a large database of results for the same GPU model. FurMark is the most intense stress test out there, but use it carefully, as it pushes power use hard and can expose PSU weaknesses.

Understanding Your Results

6. Physical GPU Inspection and Maintenance

Many GPU problems come from physical issues that software can't find. Here's how to check your graphics card by hand, just like you would with the other parts of a computer.

Checking the PCIe Slot and Connector

Unscrew the bracket, push the release latch on the PCIe slot, and carefully remove the card. Look at the gold edge connector for rust, color change, dust, bent pins, or any burn marks. If it looks dirty, rub the gold pins gently with a clean white pencil eraser until they brighten, then wipe off the residue with a lint-free cloth. While the card is out, check the motherboard slot itself for dust buildup or pin damage. When you put the card back, push it firmly until you hear the latch click. This same once-over is worth running on any second-hand card.

GPU Power Connector Issues

For standard 8-pin or 6-pin connectors, make sure the cables are pushed in all the way with no gaps. If you have an RTX 40 or 50 series card using the 12VHPWR connector, push it in until it clicks firmly and avoid bending the cable within 1.4 inches (35mm) of the connector, as that connector has a known history of melting when stressed. Check for any color change or burnt smell around the connector. As a general rule, check your power supply headroom and keep at least 20% more wattage than your system draws under load.

GPU Sagging and Support

Look at your GPU from the side; the far end should be level, not drooping. If it's sagging, install a support bracket or anti-sag stand. A vertical GPU mount with a riser cable is another option, and some people use cables tied from above to keep the card in position.

GPU Thermal Paste and Thermal Pad Inspection

This section is for advanced users who are okay with taking apart a graphics card. Note that removing the GPU cooler usually cancels the warranty, so only do this on older cards or ones already out of warranty.

Graphics cards older than 3-4 years often do better with fresh thermal paste on the GPU chip; the signs are temperatures slowly climbing over months or throttling that wasn't there before. Thermal pads on the VRAM and VRM chips also wear out over time. Picking the right thickness is critical (usually 0.5mm, 1mm, 1.5mm, or 2mm depending on the chip); measure the old pads or look up a teardown guide for your card. Using the wrong thickness means poor contact and overheating, no matter how good the pads are. For materials, Arctic MX-4 or Thermal Grizzly Kryonaut are good paste choices, and Gelid or Thermalright make quality pads.

GPU Cooling System Maintenance

Use compressed air to clean the heatsink fins and fan blades, holding each fan still while blowing so you don't spin the bearings too fast. Also make sure your case has clear airflow paths reaching the GPU; a card with good thermal paste but poor case airflow will still run hot.

7. How to Test a GPU for Faults and Rule Out Other Parts

One crash or failed benchmark does not prove that the graphics card is bad. Drivers, changed clock settings, heat, power, and other hardware can cause similar symptoms. AMD's system stability checklist recommends testing at factory settings and using a known-working GPU or computer when possible.

1. Make the problem repeat: Write down the program, action, and time that caused the failure. Run the same test again so you know whether each change helps.
2. Return to factory settings: Remove GPU, CPU, and memory overclocks or undervolts. Test with the card's default clock, voltage, and power settings.
3. Rule out software: Update the problem program and use a clean graphics driver installation. If the trouble began after a driver update, check that driver's release notes and try the last version that worked.
4. Test more than one workload: Run a benchmark and a game at default settings. A failure in only one program needs more software testing. The same artifacts or crashes in different workloads give you stronger evidence of a system or hardware problem.
5. Cross-test the hardware: If possible, put a known-working card in this PC or test the suspected card in another suitable PC. ASUS also recommends using integrated graphics or another PCIe slot during graphics card fault checks.

8. Common Problems and Solutions

Here are the most common graphics card issues and how to fix them.

No Display on Startup (Most Common Failure)

Start with the display signal. Select the correct monitor input, then try a different display cable and port. Listen for motherboard beep codes and check the board manual for their meaning.

If the basics check out, look at power. Loose power cables that aren't pushed in all the way are a common cause, as is a PSU that's not powerful enough for the card. Try swapping to different cables from the power supply. For connection issues, remove the GPU and push it back in firmly, clean the edge connector with a soft pencil eraser (the same method used to clean RAM contacts), and try a different PCIe slot if you have one. Then use the cross-test steps above before deciding which part has failed.

Heat-Related Failures (Problems That Get Worse Over Time)

Heat-related failures are recognizable by their pattern: everything works fine at first, then problems appear after 10-30 minutes of gaming, and disappear again once the system cools down. Temperatures climbing above 185-194°F (85-90°C) and symptoms getting worse in a warm room are strong signs. The most common causes are dried thermal paste on GPUs over 3 years old, dust blocking the heatsink, worn cooling fans, wrong thermal pad thickness on the memory or VRM chips, and poor case airflow. Clean the heatsink first, then check fan operation. If temperatures are still high, consider replacing the thermal paste. Undervolting through MSI Afterburner can also lower heat a lot without making the card slower.

Visual Artifacts and Corruption

Artifacts show up as random colored dots or lines, checkerboard patterns, stretched triangles across the screen, or colors looking wrong. Overheating is the most common cause, so fix cooling first. If you're running an overclock, turn it down or shut it off entirely. Artifacts that keep showing up at stock settings with normal temperatures point to VRAM failure; run stress tests like OCCT with its VRAM test to confirm. Driver corruption can also cause artifact-like symptoms; use DDU (Display Driver Uninstaller) to do a clean removal, then install the latest version fresh. If artifacts continue after all of that, the card may be failing.

Random Problems (Hardest to Find)

Random issues are the hardest to find because they don't happen when you want them to. Run stress tests for 2-4 hours with temperature monitoring on the whole time to try to trigger the problem. After any crash, check Windows Event Viewer for GPU-related errors and compare the time with your temperature log. The main cause is usually heat that only shows up under certain loads, bad solder joints reacting to the parts expanding and contracting from heat, shaky power delivery, VRAM errors when hot, or driver problems with certain games.

Performance Getting Worse

Screen flickering is almost always a driver issue; update to the latest version or roll back to a previous one that was stable. If your frame rates are lower than expected, check whether the GPU is throttling due to heat (most cards begin throttling at 87°C / 189°F or higher, though the exact point depends on the model) and confirm that the game is using the dedicated GPU. If its 3D load remains low, your CPU may be the bottleneck rather than the GPU, and it is worth confirming the processor itself is working properly. A slow performance drop over time is usually dust buildup causing thermal throttling, and a deep cleaning usually fixes it.

We put together a video version of this.