GPU Running Hot? Here's How to Diagnose and Fix It Yourself

Your frame rate suddenly tanks mid-game, your PC sounds like a jet engine, or your game crashes and exits back to the desktop without warning. These are the classic signs of a GPU running hot — and before you assume the graphics card is dead or start pricing replacements, the reality is that the vast majority of GPU overheating problems can be identified and resolved without spending a cent on new hardware. The causes are usually predictable and the fixes are straightforward once you know what to look for.

This guide walks through every step of the process: reading your temperatures accurately, identifying which cause is most likely, and applying the right fix in the right order. Work through these systematically and there is a good chance you will walk away with a significantly cooler GPU, better performance, and a computer that sounds a lot quieter under load.

Step One: Read Your GPU Temperature Properly

You cannot diagnose an overheating problem without accurate temperature data, and not all monitoring tools show the same readings. Before doing anything else, install a monitoring tool and establish what your GPU is actually doing under different conditions.

The best free options are GPU-Z, HWiNFO64, and the vendor overlay tools built into the Nvidia App or AMD Adrenalin Software. GPU-Z gives you a clean read on the core temperature at a glance. HWiNFO64 is more comprehensive and shows hotspot temperature, VRAM temperature, power draw, and fan RPM alongside the core reading — all simultaneously. For gaming, the in-game overlays in Nvidia and AMD's software show you live readings while you play without needing a second screen.

What Temperature Readings Actually Mean

Pay attention to three different temperature readings if your monitoring tool exposes them:

Core temperature is the main reading most people look at. For modern GPUs, the safe operating range under load is typically between 65°C and 85°C. Temperatures above 90°C under sustained load are a clear warning sign, and most GPUs will begin throttling their clock speeds — reducing performance to protect themselves — somewhere in the 83–95°C range depending on the card and its thermal limits. Above 100°C, you have a serious problem that needs immediate attention.

Hotspot temperature (also called junction temperature on some cards) measures the single hottest point across the GPU die rather than an average. This reading runs roughly 10–20°C higher than the core temperature under load and is normal. However, a hotspot consistently above 95°C warrants attention even if the core temperature looks acceptable.

VRAM temperature tracks the memory chips, which are a separate thermal concern from the GPU die itself. GDDR6X memory in particular is known to run hot — some cards have VRAM that regularly reaches 100–110°C, which is within its specified operating range but something to be aware of. Temperatures above 110°C for VRAM on any modern card should be investigated.

Check both idle temperatures (while sitting at the desktop doing nothing) and load temperatures (running a game or a stress tool like Furmark or Unigine Superposition for 10–15 minutes). Idle temperatures above 50°C suggest a cooling issue even before any gaming begins. Load temperatures above 85°C consistently point to one or more of the problems described below.

Cause 1: Dust Buildup on the Heatsink and Fans

Dust is the single most common cause of GPU overheating, and the one most people underestimate. GPU coolers work by drawing air through a heatsink — a dense array of thin metal fins — to dissipate heat. Over time, dust accumulates on those fins and on the fan blades, restricting airflow and reducing cooling efficiency dramatically. A GPU that has not been cleaned in a year or two can easily run 10–20°C hotter than it did when new, purely from dust accumulation.

Remove the side panel of your case and visually inspect the GPU cooler. Look at the intake side of the fans and the heatsink fins visible from the exhaust side. If you can see a visible layer of grey fuzz or compacted dust on the fins or fan blades, that is your problem.

How to Clean It Safely

Use compressed air in short bursts to dislodge dust from the heatsink fins and fans. Hold the can upright and use short bursts rather than a continuous stream to avoid moisture. Hold the fan blades still with a finger or a pencil before blowing — allowing fans to spin freely from compressed air can damage the bearings. Work in short bursts from multiple angles to push dust out through the exhaust side of the heatsink rather than deeper into it.

For heavier buildup, an electric air blower (available for around $20–30 at most electronics retailers) is more powerful and more cost-effective than canned air over time. A soft-bristled paintbrush can remove stubborn compacted dust from between fins when compressed air alone is not shifting it.

For the most thorough cleaning, remove the GPU from the case entirely and take it outside or to a well-ventilated area before blowing. This keeps the dust out of the rest of your case and lets you work from all angles. Clean the dust filters on your case's intake fans at the same time — they are a major accumulation point that restricts fresh air reaching the GPU.

Cause 2: Poor Case Airflow

Even a perfectly clean GPU cooler cannot work well if the airflow inside your case is inadequate. The GPU draws in cool air from the surrounding case environment and exhausts hot air out through the rear or top of the heatsink. If the case itself has poor airflow — not enough intake air, blocked exhaust paths, or cables interfering with the airflow path — the GPU ends up recirculating warm air rather than drawing in fresh cool air.

Evaluate your case airflow setup against the standard that works: more air coming in at the front and bottom (positive pressure intake), with air exiting at the rear and top (negative pressure exhaust). A case with two or three front intake fans and one or two rear/top exhaust fans will maintain a steady flow of cool air across the entire system including the GPU.

Common Airflow Problems to Fix

If your case has intake fan mounts at the front but no fans installed in them, that is the first thing to address. Adding one or two 120mm or 140mm case fans to the front intake position for $10–20 each can drop GPU temperatures by 5–10°C in a restricted airflow scenario. The physical size of the fans matters less than ensuring there are enough of them moving sufficient air volume.

Cable management is the other major factor. A mass of cables hanging in front of the front intake fans or directly in the GPU's intake path can reduce effective airflow significantly. Route cables along the back panel of the case, use cable ties to bundle them out of the way, and ensure there is clear space in front of all intake fans and below the GPU if it uses bottom-intake fans.

Check the physical placement of your PC as well. A desktop tower positioned with its intake side against a wall, or sitting inside a closed cabinet, cannot pull in adequate fresh air regardless of how many fans it has. Position the case with at least 15–20 cm of clearance on all sides that have fans or vents.

Cause 3: Dried Thermal Paste Between GPU Die and Heatsink

Thermal paste fills the microscopic gaps between the flat metal surface of the GPU die and the base plate of the heatsink, ensuring that heat transfers efficiently between them. Over time — typically three to five years of regular use, though it varies by brand and application — thermal paste dries out, cracks, and loses its ability to fill those gaps effectively. When that happens, heat transfer degrades and temperatures rise even though nothing else about the cooling setup has changed.

This is most relevant for GPUs three or more years old that have been in regular use. If you have a relatively new card, dried thermal paste is unlikely to be the cause. But for a card purchased in 2020 or earlier that has been running hot despite clean fans and good airflow, replacing the thermal paste is often the most impactful single fix available and can drop temperatures by 10–20°C on older cards with particularly degraded paste.

How to Replace GPU Thermal Paste

This procedure requires removing the GPU's cooler from the PCB, which requires some care but is well within the ability of anyone comfortable opening a PC. The specific steps vary by card model — search for your exact GPU model and "disassembly" or "thermal paste replacement" to find a guide with the specific screw placements and any connectors you need to unplug on your card.

The general process is: remove the GPU from the case, remove the shroud and heatsink (typically held by screws on the back of the PCB), clean the old paste from both the GPU die and the heatsink base plate using isopropyl alcohol at 90% concentration or higher on a lint-free cloth, apply a fresh thin layer of new thermal paste (a pea-sized dot centered on the die is sufficient — it will spread under pressure), and reassemble.

Good thermal paste options in the $6–15 range include Thermal Grizzly Kryonaut, Arctic MX-6, and Noctua NT-H2. Avoid budget no-name pastes for this application — quality matters and the price difference is trivial compared to the value of the GPU you are servicing.

The key discipline is using a thin layer. Too much paste is actually counterproductive — it acts as an insulator rather than a conductor when applied thickly. The goal is a thin, gap-filling layer, not a thick coating.

Cause 4: Degraded Thermal Pads on VRAM and VRMs

The memory chips and voltage regulators (VRMs) on a GPU PCB are cooled by thermal pads — flat, slightly compressible pads that sit between the component surface and the heatsink plate. Unlike thermal paste, these are not used on the main GPU die, but they are essential for keeping VRAM and VRM temperatures under control. Thermal pads also degrade over time, losing their compliance and thermal conductivity as they age and harden.

If your monitoring tool shows VRAM temperatures climbing to concerning levels (consistently above 100°C on GDDR6 or above 105°C on GDDR6X) or if specific areas of the PCB near the memory chips are notably hotter than they should be, degraded thermal pads are a likely contributor.

Replacing thermal pads is a more involved procedure than replacing paste because you need to source pads of the correct thickness for your specific GPU model — using pads that are too thick or too thin will produce worse contact than the originals. Measure the original pads with calipers or look up the documented thickness for your card before ordering replacements. Thermal pad thickness typically ranges from 0.5mm to 3mm depending on the component and card. Quality replacement options include thermal pads from Thermal Grizzly, Fujipoly, or Gelid.

Cause 5: Suboptimal Fan Curve Settings

Modern GPU coolers are designed to be quiet at low and medium loads, which means the default fan curve — the relationship between GPU temperature and fan speed — is tuned to prioritize silence over aggressive cooling. On some cards and in some environments, that default curve does not respond to rising temperatures quickly enough, allowing temperatures to climb further than they need to before the fans spin up to counteract them.

A custom fan curve can lower your peak temperatures by 5–10°C on cards with overly conservative default profiles, with no hardware changes required at all. This is always worth doing before opening your GPU for thermal paste replacement.

How to Set a Custom Fan Curve

MSI Afterburner is the most widely used tool for this and works with Nvidia, AMD, and Intel Arc GPUs. In Afterburner, click the fan settings button (the small icon near the fan speed percentage), enable the custom fan curve option, and you will see a graph with temperature on the X axis and fan speed percentage on the Y axis.

A more aggressive but still reasonable fan curve might look like this: 0% at 30°C (fans off at idle), 30% at 50°C, 50% at 65°C, 70% at 75°C, 85% at 82°C, 100% at 88°C. The specific values that work best depend on your GPU cooler and your tolerance for fan noise. The key principle is to bring the fans up to meaningful speeds earlier in the temperature curve so they are already working before temperatures reach problematic levels rather than only ramping up reactively.

Nvidia's own app and AMD's Adrenalin software both include fan curve editors as well if you prefer to stay within the vendor's tools rather than using a third-party application.

Additional Fix: Undervolting Your GPU

Undervolting is often overlooked by casual users but can be one of the most effective ways to reduce GPU temperatures, reduce fan noise, and improve stability all at once. Modern GPUs receive more voltage than they strictly need at any given clock speed — it is a factory safety margin. Reducing that voltage while maintaining the same clock speed means the GPU generates less heat, which translates directly to lower temperatures under load.

The process is done in MSI Afterburner using the voltage-frequency curve editor. Press Ctrl+F to open the curve, identify the point corresponding to your GPU's typical boost clock speed (check GPU-Z under load to find this), and set a lower voltage for that frequency point while keeping the frequency target the same. Common results are a reduction of 15–30°C on cards that accept aggressive undervolts well, with no reduction in performance since the clock speed is unchanged.

Undervolting does require some trial and error — start conservatively (reduce voltage by 50–75mV below the default) and run a stability test (Furmark or a demanding game for 30 minutes) before going further. If the system crashes, the undervolt is too aggressive and needs to be raised slightly. Most modern Nvidia and AMD GPUs respond well to undervolting in the -50 to -150mV range.

When the Problem Might Be the Fans Themselves

If you have cleaned the fans, improved case airflow, refreshed thermal paste, and tuned the fan curve — but one or more GPU fans still do not spin up during load, or make grinding, clicking, or scraping noises when they do — the fans themselves may have developed a mechanical fault. Fan bearings wear out, particularly on cards that have been running continuously for several years or that have operated at high temperatures for extended periods.

A single stuck or failed fan can cause significant temperature spikes because the heatsink depends on both fans drawing air through simultaneously. Monitor fan RPM alongside temperature in HWiNFO64 — if one fan shows significantly lower RPM than the other, or shows zero RPM while the other is running, that is the issue.

GPU fans can often be replaced individually. Some manufacturers sell replacement fans for popular models, and third-party replacements are available for most major GPU heatsink designs on sites like AliExpress. The replacement process is typically straightforward — remove a few screws, unplug the fan connector from the PCB, and swap in the new unit. This is considerably cheaper than replacing the GPU.

The Systematic Fix Order: Work Through These Steps

If you are not sure where to start, work through the causes in this order — from easiest and cheapest to more involved:

Check temperatures using GPU-Z or HWiNFO64 at idle and under load. Confirm you actually have a temperature problem and note which readings are elevated.
Clean the GPU cooler and case dust filters with compressed air. This alone resolves the majority of cases where temperatures have climbed gradually over time.
Improve case airflow by adding intake fans, tidying cables, and ensuring adequate clearance around the case.
Adjust the fan curve in MSI Afterburner to a more aggressive profile. Check whether this brings temperatures into range before going further.
Try undervolting in MSI Afterburner if the fan curve adjustment alone is not sufficient. This can be done without opening the GPU.
Replace thermal paste if the GPU is three or more years old and the above steps have not resolved the issue.
Replace thermal pads if VRAM or VRM temperatures remain elevated after replacing thermal paste.
Check and replace fans if individual fans are not spinning correctly even after cleaning.

The vast majority of overheating GPUs are resolved by the first three to four steps without needing to open the card at all. A thorough clean combined with a smarter fan curve and improved case airflow handles the majority of situations. Thermal paste and pads are the next layer when those steps are insufficient — and they make a dramatic difference on older cards that have never had them replaced.

A GPU that runs at 75°C instead of 92°C does not just have lower temperatures — it boosts more consistently, crashes less, fans run quieter, and the card itself will last considerably longer. These are worthwhile results from a few hours of work that costs little to nothing in parts.

Want more hardware how-to guides, GPU performance tips, and PC maintenance advice? Browse our other posts for more practical coverage.