How to Fix CPU/GPU Temp Difference After Laptop Repaste

You just finished a repaste expecting lower temps but the CPU still runs 15 °C hotter than the GPU under the same load. That gap is not normal and it usually points to a mechanical issue rather than the paste itself. How to Fix CPU And GPU Temperature Difference After Repaste on a Gaming Laptop starts with measuring the actual delta and then checking the mounting pressure stack.

Most fixes take one careful reassembly once you know which variable is off.

Honeywell PTM7950 phase-changes at 45 °C and needs sustained heat above 50 °C for two minutes to wet both dies evenly. A shared vapor chamber that tilts 0.02 mm at one corner creates exactly the split you are seeing. Let’s walk through the diagnostic workflow that isolates the root cause.

How to Fix CPU And GPU Temperature Difference After Repaste on a Gaming Laptop

Measure the package delta with Cinebench and Furmark running together for fifteen minutes. If the gap exceeds 5 °C inspect the cold plate flatness and pad stack heights. Reapply paste using a 1 mm line across each die long axis.

Torque spring screws in a diagonal pattern to 0.35 Nm. Reset the EC and retest.

Why Your CPU Runs 15 °C Hotter Than the GPU After Repaste

A shared heatsink relies on even contact across both dies. The CPU die on Intel Core i9 HX or AMD Ryzen 9 HX is smaller than the GPU die on an RTX 4080 mobile. Uneven paste spread leaves a micro gap on the larger die.

Spring loaded screws that bottom out at different heights tilt the cold plate. Thermal pads that are 0.2 mm too thick on one side act like a wedge. Any of these factors shifts mounting pressure toward one die.

The result is a 10 to 25 °C split that looks like a paste failure but is actually a pressure failure.

The Cooling Stack Variables That Actually Decide the Delta

Die size mismatch is the first variable. Intel HX dies measure roughly 10 by 15 mm while NVIDIA mobile dies measure 15 by 20 mm. Cold plate flatness tolerance from the factory is often 0.03 mm which is enough to break contact on the larger die.

Pad stack height must match the compressed Z height of each component. VRAM pads on a 2 mm free height compress to 1.2 mm at 40 percent. PCH pads on 1 mm free height compress to 0.6 mm.

A single 0.1 mm error across four pads tilts the assembly. Spring screw torque spec varies by chassis. Clevo designs spec 0.35 Nm.

Tongfang designs spec 0.4 Nm. Alienware designs spec 0.5 Nm. Exceeding spec bows the plate.

Undershooting leaves gaps.

Variable Typical Spec Delta Impact
CPU die size 10 × 15 mm Smaller contact area
GPU die size 15 × 20 mm Larger contact area
Cold plate flatness ≤ 0.03 mm Gap on larger die
Pad compression 30–40 % Tilt if mismatched
Screw torque 0.35–0.5 Nm Bow or gap

Diagnostic Workflow: Baseline → Disassembly → Reassembly → Validation

Start with a dual load baseline. Run Cinebench 2024 multi core and Furmark 1080p stress simultaneously for fifteen minutes. Log CPU package GPU package CPU hotspot GPU hotspot VRAM junction and VRM temps in HWiNFO.

Note the delta at steady state. Power off and disconnect the battery. Photograph every cable and screw location.

Remove the heatsink assembly. Clean old TIM with 99 percent IPA and lint free swabs. Inspect both dies under 10x magnification for scratches or oxidation.

Check cold plate flatness with a 0.02 mm feeler gauge or razor blade light test. Measure every thermal pad with calipers compressed and free. Record thickness per location.

This data set tells you exactly which variable is out of spec.

TIM Selection Decision Tree: PTM7950, Kryonaut, or Liquid Metal?

If both dies hit 70 °C under load choose PTM7950. It phase changes at 45 °C and resists pump out for three to five years. If the GPU rarely exceeds 50 °C choose Kryonaut or TG900.

They wet instantly without a phase change window. If you accept conductive risk and can mask every nearby component choose liquid metal. Conductonaut runs 73 W/mK but migrates without a conformal barrier.

PTM7950 runs 8.5 W/mK but is non conductive and non capacitive. Kryonaut runs 12.5 W/mK but pumps out in twelve to eighteen months on bare die. Match the TIM to the thermal profile not the marketing number.

Torque Sequence, Pad Stack, and Spread Patterns That Cause Imbalance

Apply a 1 mm line of paste along the long axis of each die. Do not use a pea or cross hatch. Place fresh pads cut to OEM shape at the measured thickness.

Lower the heatsink straight down. Engage all four corner screws one turn each. Follow a diagonal sequence.

Torque in quarter turn increments to the chassis spec. Stop at 0.35 Nm for Clevo 0.4 Nm for Tongfang 0.5 Nm for Alienware. Verify the plate sits level with a straight edge.

Reconnect battery. Reset EC by holding power button thirty seconds with battery disconnected. Boot to BIOS load defaults save and exit.

Run the same dual load test. Delta should now sit under 5 °C. If it does not recheck pad heights and cold plate contact pattern with UV dye or pressure film.

Pro Validation: UV Dye, Pressure Film, and EC Reset Tricks

UV dye is the fastest way to see actual contact. Mix a drop of fluorescent tracer into a non-conductive paste, reinstall, run ten minutes of load, then pull the heatsink. Under a UV torch, bare patches on the larger die show up instantly.

That pattern tells you tilt, not paste quality, so it's worth doing whenever temps got worse instead of better after a redo.

Pressure-sensitive film works when you'd rather not disturb fresh paste. Trim a strip to die size, lay it on the cold plate, torque to spec, then peel. Low-grade Prescale film reads roughly 0.05 to 0.2 MPa.

Darker patches mean higher pressure, so compare left to right across the plate.

What does an EC reset actually change?

The embedded controller stores fan tables and thermal hysteresis. After hardware work, those tables can hold stale values and ramp late. Disconnect the battery, hold power for thirty seconds, then boot to BIOS and load defaults.

If the dye or film shows conductive material creeping past the die edge, stop and clean it before powering on. Our walkthrough on metal that has dried out or shifted covers the cleanup. For general handling, review safe liquid metal practice first.

Log ambient temperature with every retest. A room 4 °C warmer skews every delta you recorded.

Frequently Asked Questions

Is a 10 °C gap between CPU and GPU actually a problem?

Often, yes. Shared vapor chamber designs should land within about 5 °C under matched load. A wider gap means uneven mounting pressure rather than a bad paste.

Check pad heights and cold plate flatness before buying more TIM.

Should I use liquid metal on both dies or only the CPU?

Only the GPU die if you can mask it properly. Both dies are bare on most HX chassis, so both accept metal, but the GPU sits closer to VRAM and power stages. If masking feels rushed, keep PTM7950 on the GPU and metal on the CPU.

How long does PTM7950 need to settle?

Two to three full heat cycles. Honeywell's published material data puts the phase change at 45 °C, with full wetting above 50 °C. Run a stress test, cool fully, then repeat.

Temps usually drop 2 to 4 °C by the third cycle.

Why did thicker pads make things worse?

Thicker pads over-compress and act like a wedge under the cold plate. Measure compressed height with calipers, then target 30 to 40 percent compression. Anything stiffer steals pressure from the dies.

Do I need fresh pads on every repaste?

Reuse them only if they still spring back to original thickness. Compressed or torn pads should go. Fresh kits cost little, and if your fans turned louder after the last teardown, a bad pad stack is a common culprit.