Your CPU temperatures look fine at idle, then climb into the 90s under load even though your cooler is rated well above your chip’s output. You repaste, temps drop for a few weeks, then slowly creep back up. On many modern platforms, especially Intel’s LGA1700 and LGA1851 sockets and any system where the cooler applies significant horizontal stress to the IHS, this is the classic thermal paste pump out issue: the paste migrates away from the center of the heat spreader over repeated heat cycles, leaving metal-on-metal contact in the hottest zone.
It is urgent but rarely catastrophic. Sustained temperatures above 95 °C trigger thermal throttling, which costs you performance and shortens paste and thermal interface lifespan, but modern CPUs protect themselves by shutting down around 100–105 °C. Still, don’t ignore it — every thermal cycle accelerates the pump-out effect, so a system left in this state degrades faster. Budget an hour for diagnosis and another hour for the fix itself.
As an Amazon Associate we earn from qualifying purchases at no extra cost to you.
Our Top Picks
PTM7950 Thermal Pad Phase Change Heatsink 80x80x0.25mm 8.5W/mK3.15 x 3.15 x 0.01 inches0 poundsCheck on Amazon
Likely Causes
| Cause | How to confirm | Fix difficulty |
|---|---|---|
| Thermal paste pump-out from heat cycling | Remove cooler after weeks of use and find paste pushed to one edge of the IHS with a dry center | Easy |
| Insufficient mounting pressure | Fingers can wiggle the cooler noticeably; screws spin freely at rated torque | Easy |
| Direct-die paste used where liquid metal or a phase-change pad is appropriate | Paste on an exposed die (AMD Ryzen 7000/9000 or Intel GPU dies) shows cracking and gaps after a few dozen cycles | Moderate |
| Incorrect paste application pattern for a large IHS | Peel the cooler and inspect: corners of the IHS are dry, paste blob is off-center | Easy |
| Clogged or stalled cooler fans or a failed pump (AIO) | Check fan RPM in BIOS hardware monitor; feel for AIO pump vibration; radiator is cold while CPU is at 95 °C | Moderate |
| Ambient or case airflow problems amplifying the cycle | Intake air above 30 °C, GPU exhaust dumping onto the CPU cooler, or dust mats on radiator fins | Easy |
Fix 1: Confirm the Problem with a Load Test and Temperature Log
What causes it
Before opening anything, you need to verify the symptoms are actually thermal and not a sensor or software reporting issue. Overheating symptoms get misattributed constantly.
How to check
Open HWiNFO64, expand your CPU, and watch “Core (Tctl/Tdie)” or “Core Temperature” while running Cinebench 2024 (10-minute loop) or OCCT’s CPU test with the “Data Load” thread count set to match your physical core count. Note the peak temperature and whether “CPU Package Throttling” flags appear in HWiNFO’s sensor panel.
What to do
Log temps for one session. If peaks exceed 95 °C with throttling flags, or if you see a 10–15 °C climb over the first 20 minutes as the TIM degrades in real time, pump out is likely. Write the numbers down — you’ll compare after repasting.
How to undo it
Nothing to undo. Close HWiNFO and Cinebench when done.
Fix 2: Improve Case Airflow and Fan Curves
What causes it
High ambient and stagnant case air raise the delta across the cooler, driving hotter heat cycles that accelerate pump out. They also mask whether the paste itself is the problem.
How to check
In your fan control software (BIOS hardware monitor, Fan Control by Rem0o, or the motherboard utility like Armoury Crate FanExpert), check current fan RPM against the maximum listed on each fan’s label. A 120 mm fan idling at 400 RPM under load is barely moving air.
What to do
Set a chassis fan curve keyed to CPU package temperature: 40 % PWM at 40 °C, ramping linearly to 100 % at 80 °C. In the BIOS under “Hardware Monitor” or “Q-Fan Control”/”Smart Fan 6,” set the CPU fan source to “CPU Package” rather than a socket thermistor — socket sensors lag by 10+ seconds. Confirm at least one front intake fan is moving more air than the exhausts push out (slight positive pressure).
How to undo it
Restore the default “Standard”/”Auto” fan curve in the same BIOS menu, or delete the custom profile in Fan Control.
Fix 3: Reseat the Cooler with Correct Mounting Pressure
What causes it
Insufficient or uneven mounting pressure lets the IHS and cold plate micro-separate during thermal expansion, mechanically squeezing paste outward. Over-tightening in sequence (tightening one screw fully first) does the same by tilting the cold plate.
How to check
After removing the cooler, examine the paste footprint on the IHS. Even coverage edge to edge means good contact; a thick blob on one side with a dry opposite corner means uneven pressure. Check whether your bracket’s springs compress fully — on LGA1700, the anti-bending frame must be torqued to roughly 0.4–0.5 N·m per screw in the sequence printed on the board.
What to do
Clean both surfaces with 99 % isopropyl alcohol and a lint-free cloth. Apply fresh paste (a 4–5 mm dot centered, or five small dots for large IHS units), then tighten screws in a diagonal cross pattern, two full turns each pass, until the springs bottom out. Never tighten one screw fully before starting the next.
How to undo it
Unscrew in reverse diagonal order, clean, and reapply paste if you need to remove the cooler again — never reuse disturbed paste.
Fix 4: Repaste with a Pump-Out-Resistant Compound
What causes it
Thin, high-flow pastes (many are rated 3–8 W/mK with low viscosity) migrate easily under thermal cycling. Thicker, higher-viscosity compounds resist pump out even if peak conductivity is slightly lower.
How to check
Look at the compound you’re using. If it’s a runny paste or one marketed primarily on peak W/mK, it’s more prone to migration. If you can see the old paste has beaded at the IHS edges rather than spread, it flowed too easily.
What to do
Clean with 99 % isopropyl alcohol. Apply a paste specifically formulated against pump out — thicker compounds or phase-change material pads (which soften at operating temperature but stay solid at idle). For a dot application, aim for a 4 mm diameter dot; for phase-change pads, cut to IHS size and press the cooler down squarely in one motion.
How to undo it
Remove the cooler, clean both surfaces with isopropyl alcohol (let phase-change residue soften with gentle warmth from 20 minutes of low-load use first), and apply a different compound.
Fix 5: Adjust Power Limits to Reduce Heat Cycles
What causes it
Aggressive power limits cause sharp temperature swings — chips hitting 100 °C then dropping to 50 °C hundreds of times per hour. Each cycle pushes paste a fraction of a millimeter. Smoothing the swing slows the mechanism even if you don’t repaste.
How to check
In HWiNFO64, watch “CPU Package Power” during your load test. If it spikes to the PL2 limit (often 253 W on Intel K-series, or “PPT” of 230 W on AMD Ryzen 9) for the first seconds of every load, you’re on the aggressive end.
What to do
Enter BIOS, go to the overclocking/performance menu (often “Advanced \ CPU Configuration” or “Tweaker”), and find “Long Duration Power Limit” (PL1) and “Short Duration Power Limit” (PL2). Set PL1 to a value your cooler sustains at or below 85 °C — try starting at 125–181 W for a modern Intel desktop chip, or set “PPT” to 142–200 W on AMD. Set “CPU Loadline Calibration” to a middle setting (level 3–4 on ASUS, “Standard” to “High” on MSI) to reduce transient voltage spikes.
How to undo it
In the same BIOS menu, set power limits back to “Auto” or select the “Remove All Limits”/”Unlimited” option, then choose “Discard Changes and Exit” if you want to bail entirely.
Fix 6: Switch to a Phase-Change Pad or PTM7950 on High-Cycle Systems
What causes it
Some hardware — consoles repurposed as HTPCs, small-form-factor builds, laptops repurposed as desktops — runs aggressive thermal cycling that destroys any paste within months. A phase-change material that stays structurally stable across thousands of cycles is the durable answer.
How to check
If you’ve repasted two or more times in a year and temps degrade each time within 4–8 weeks, you’re on the pump-out treadmill. Check your cooler’s mounting orientation too: a tower cooler hanging horizontally (radiator vertical, fans sideways) puts shear stress on the paste every cycle.
What to do
Clean both surfaces thoroughly. Cut a Honeywell PTM7950 sheet (or similar 0.25 mm phase-change pad) to match the IHS dimensions exactly — it should cover the die area plus 2–3 mm on each side. Peel one backing layer, press it onto the IHS, peel the top layer, then mount the cooler. Let it warm up to operating temperature once before judging performance; PTM needs a heat cycle to fully conform.
How to undo it
Remove the cooler, gently peel the pad off (a plastic spudger helps), clean residue with isopropyl alcohol, and repaste normally.
Fix 7: Re-Orient or Upgrade the Cooler Mounting
What causes it
Tower-style air coolers impose significant sideways shear on the IHS/paste interface — heavier coolers amplify this. Combined with a board’s stock bending, paste literally gets squeezed out sideways every cycle.
How to check
Measure how much your cooler weighs and check whether your motherboard has a backplate reinforcement. If the board visibly bows around the socket when the cooler is mounted (view from the side with the case on its side), you have flex contributing to pump out.
What to do
Install a contact frame (for LGA1700, this replaces the stock ILM) or an aftermarket anti-bending backplate. Torque it evenly, then remount the cooler with fresh paste. Alternatively, rotate a tower cooler 90 degrees so fins point front-to-back, reducing cantilever load on one side of the socket.
How to undo it
Remove the contact frame, reinstall the stock independent loading mechanism in its original orientation, and torque the four socket screws evenly in a cross pattern.
Fix 8: Address the Environment Around the Machine
What causes it
Ambient heat raises the whole cooling stack’s baseline temperature, so every thermal cycle starts hotter and lasts longer. This makes pump out worse and makes everything else you’ve done less effective.
How to check
Place a cheap thermometer near the intake fans. If it reads above 28 °C (82 °F), your room or desk enclosure is part of the problem. Check whether the PC sits inside a closed cabinet or next to a radiator/heat source.
What to do
Move the case out of enclosed furniture, ensure 15 cm minimum clearance behind the rear exhaust, and keep it away from direct sunlight. In a warm room, running the AC or a desk fan toward the intake side helps more than most software tweaks.
How to undo it
Move the PC back to its original location. Nothing is technically reversed.
If Nothing Works
If you’ve repasted with a pump-out-resistant compound, corrected mounting pressure, set sensible power limits, and temps still climb back within weeks, suspect a hardware fault rather than paste behavior. Check three things. First, verify the cooler itself: an AIO pump running below its rated RPM (compare against the pump’s spec sheet) or a heat pipe that no longer feels hot at the fin end has failed internally. Second, check for IHS delamination or a warped IHS — place a straightedge across the heat spreader; visible gaps at corners mean the package itself is deformed and the chip may need an RMA. Third, confirm your board’s BIOS is current: Intel microcode updates released through 2026 addressed several elevated-voltage behaviors that indirectly aggravated thermal cycling on 13th/14th-generation chips. If all three check out and degradation persists, the CPU package’s internal TIM may have degraded, which is a warranty matter — contact the manufacturer with your HWiNFO logs and repaste history.
FAQ
How often does thermal paste pump out actually happen?
On desktop chips with a stock IHS and a properly mounted cooler, meaningful pump out typically takes 12–24 months of heavy thermal cycling. On laptops, direct-die builds, and systems with high mounting shear, it can show up in a few weeks. Aggressive power limit swings shorten the timeline considerably.
Is liquid metal a good fix for pump out?
On a sealed IHS-to-cooler interface, gallium-based liquid metal resists pump out well and conducts better than paste, but it’s electrically conductive, aluminum-corrosive, and risky to apply. It should never touch an exposed AMD Ryzen CPU die with nearby capacitors unless you’ve conformed the board properly. For most people, a phase-change pad is safer and nearly as durable.
Can I just apply more paste to compensate?
No. Excess paste increases the bond line thickness, which raises thermal resistance — you lose performance at idle and only marginally delay the migration. Pump out is a migration problem, not a quantity problem. The right amount of the right compound beats a lot of the wrong compound.
Will pump out damage my CPU permanently?
Not directly. The chip throttles and shuts down well before silicon damage occurs. But sustained 95 °C+ operation ages the package’s internal TIM and can stress the silicon over months. Fix the paste situation promptly, and your CPU should outlive the rest of the build with no lasting harm.
