How to Check GPU Health Before Buying Used -- Complete Guide 2026

Why Testing Matters -- The Real Risk

Pakistan's used GPU market is full of legitimate deals but also full of traps. A seller on OLX or at Hall Road can hand you a GPU that worked fine for five minutes in their system but has been running at 95°C for two years, has degraded VRAM from mining duty, or has a fan bearing that will fail within a month. By the time you discover the problem, the money is gone and the seller is unreachable.

The good news: most GPU problems reveal themselves within 30-60 minutes of proper stress testing. The tools required are all free. The process takes less than an hour. This guide walks you through every test in the correct order, what to look for, and what the red flags mean.

Always insist on testing before payment. Any legitimate seller will allow this. If a seller refuses to let you test the GPU in a working system before purchase, walk away. That refusal is itself a red flag.

What You Need -- Free Tools

Download and install these before you go to test a GPU:

• GPU-Z (techpowerup.com) -- identifies the GPU, shows specs, monitors sensors
• FurMark (geeks3d.com) -- GPU stress test that pushes the card to maximum load
• HWiNFO64 (hwinfo.com) -- detailed hardware monitoring including temperature history
• OCCT (ocbase.com) -- includes VRAM error test, excellent for finding memory issues
• Any demanding game -- GTA V, PUBG, or similar for real-world testing

Step 1 -- Verify the GPU Identity with GPU-Z

Before running any test, open GPU-Z and verify that the card in front of you is actually what the seller claims it is. This step catches one of the most common scams in used GPU markets: a card being sold under a different model name.

In GPU-Z, check these fields against the claimed model:

• Name: Should match exactly (e.g., "GeForce GTX 1050 Ti")
• Memory Size: For GTX 1060, verify it shows 6144 MB, not 3072 MB
• Memory Type: GT 1030 should show GDDR5, not GDDR4 (performance difference is 30-40%)
• Bus Width: Should match the card's spec (e.g., 128-bit for GTX 1050 Ti)
• Shader Processors: Should match the official spec

If any of these do not match the claimed model, you are looking at a different -- and likely inferior -- card. Walk away or renegotiate the price based on what it actually is.

Step 2 -- Check Physical Condition

Before plugging the GPU into anything, inspect it physically:

• Fan check: Spin each fan blade gently with your finger. They should spin freely and smoothly with no grinding, scraping, or resistance. Any grinding sound indicates worn bearings that will fail soon.
• PCB inspection: Look for burn marks, corroded contacts, bent capacitors, or any signs of physical damage. A small amount of dust is normal -- heavy caked-on dust that packed the heatsink fins is a sign of poor maintenance.
• Thermal paste: If you can see the heatsink area, fresh thermal paste (white or grey, still slightly moist-looking) is a good sign. Dried, cracked, or orange thermal paste indicates the card has been running hot for a long time.
• PCIe connector pins: The gold fingers on the PCIe edge should be clean and bright. Dark or corroded contacts can cause instability.

Step 3 -- FurMark Stress Test (15-20 Minutes)

FurMark is a GPU stress test that pushes the card to 100% load continuously. It is designed to find thermal and stability issues quickly. Run it for at least 15-20 minutes while watching GPU-Z's sensor tab or HWiNFO64 alongside it.

What you are watching for:

• Temperature: Most GPUs should settle between 70°C and 85°C under FurMark. Above 90°C is a warning sign. Above 95°C consistently is a red flag -- the thermal solution is degraded and the card will throttle performance and degrade faster under gaming loads.
• Clock speeds: Watch the GPU clock in GPU-Z. It should hold steady at its boost clock. If the clock drops significantly and keeps dropping over time, the card is thermal throttling -- it is overheating and reducing performance to protect itself.
• Artifacts: Any flickering, colored pixels, geometric shapes, lines, or screen tears during FurMark are a critical failure sign. Stop the test immediately -- this card has hardware damage.
• Crashes: Any crash, black screen, or driver reset during FurMark is an automatic fail.

Step 4 -- VRAM Test with OCCT

VRAM errors are one of the hardest GPU problems to detect because they do not always show up as visible artifacts. Instead, they cause random crashes, game freezes, or corrupted textures that appear intermittently and are easy to blame on other causes.

Run OCCT's GPU VRAM test for at least 10 minutes. This test writes and reads patterns to every memory cell on the GPU and reports any errors. Even a single VRAM error is a significant red flag -- healthy GPU VRAM should show zero errors in this test.

One or two errors might still be okay depending on where they occur in VRAM, but any errors at all should reduce what you are willing to pay significantly, and a card with many VRAM errors should be avoided entirely.

Step 5 -- Real Game Test (30 Minutes)

After the stress tests, run a real game for at least 30 minutes. This tests the card under the actual conditions it will spend most of its time in -- running real game shaders, handling actual VRAM patterns, and interacting with a real game engine rather than a synthetic benchmark.

GTA V is ideal for this because it has a built-in benchmark tool that runs a consistent scripted sequence. Run the GTA V benchmark three times in a row. The FPS numbers should be consistent between runs -- within 5-10% of each other. If the third run is significantly slower than the first, the card is thermal throttling under sustained load.

During the game test, watch for:

• Any visual artifacts that did not appear in FurMark (some artifacts only appear with specific game shaders)
• Random texture corruption or flickering textures
• Unexplained frame drops that do not correspond to heavy scene changes
• Any driver crash or game crash

Mining GPU Red Flags -- Specific Checks

Mining GPUs are a particular concern in Pakistan's used market. Ethereum mining ran cards at 100% GPU load continuously, often in poorly ventilated setups, potentially for years before the mining crash. Here is how to identify one:

• HWiNFO64 sensor history: Check the GPU's max recorded temperature. Mining GPUs often show higher historical maximums than gaming-used cards.
• Fan wear: Mining rigs ran fans at high speed continuously. Fan bearings on mining GPUs wear out faster. The grinding fan test is especially important here.
• Throttling pattern: Mining GPUs were often run with power limit reductions to improve efficiency. Check if the card throttles even at moderate loads, which can indicate changed power delivery components.
• VRAM condition: GDDR5 and GDDR6 VRAM degrades under continuous high-bandwidth mining workloads faster than under gaming workloads. Run the OCCT VRAM test with extra scrutiny on any card you suspect was mined.

How to Negotiate Based on Test Results

Use the test results as leverage. If the card passes all tests cleanly, pay the asking price if it is fair. If the card has high temperatures (85-90°C) but no artifacts or crashes, negotiate a discount -- the card needs thermal paste replacement and potentially new fans, which costs time and money. If the card fails FurMark or shows VRAM errors, either walk away or offer a deeply discounted price that reflects the risk.

Quick Test Checklist -- Print and Bring

• GPU-Z: Verify model name, VRAM amount, memory type, bus width
• Physical: Fan spin test, PCB inspection, thermal paste condition
• FurMark 15-20 min: Temperature stays below 90°C, no artifacts, no crashes
• OCCT VRAM 10 min: Zero errors preferred
• Real game 30 min: Consistent FPS, no artifacts, no crashes
• Ask for 24-hour return window before paying