Thermal Paste in Laptop Cooling: Managing 35 W/cm2 on a Short Thermal Path

A notebook heatsink is the most unforgiving thermal assembly in consumer electronics. A 28 W processor with a 100 mm2 die dissipates 28 W/cm2, and a 45 W discrete GPU on a 150 mm2 die is not far behind. Unlike a desktop tower, there is no room to buy back temperature with a larger heatsink: the path is a 1.2 mm vapour chamber, two heat pipes, a fin stack and one blower, and the whole assembly is held down by four spring screws at a few hundred newtons. In that architecture the result is decided in a 20 to 50 micron layer that nobody sees after assembly, which is exactly where thermal paste earns or loses its place.

Why the interface is 30 percent of the budget

Thermal resistance in a notebook is distributed over a very short chain, so the interface is a large fraction of a small total. If the die-to-heatsink path is roughly 0.9 K per watt from junction to air, an interface contributing 0.03 to 0.05 K per watt is more than a rounding error, and with 35 W/cm2 of local flux a 0.3 mm gap filler would contribute 17 K on its own. That figure explains why a pad is used on a chassis or an SSD but not on a bare die.

The flux is not uniform: where the paste actually has to work

In laptop thermal management there are four interface situations, and applying one product everywhere is how a good design becomes an average one. The die and the cold plate need the thinnest achievable thermal interface material joint, the VRM and the memory need a compliant layer over a larger area, and the heat pipe to chassis joint is a low-flux mechanical problem.

What actually fails in the field, and why laptop thermal management degrades

Pump-out on a thin die. The die expands more than the cold plate, and at 35 W/cm2 the paste is squeezed outward from the centre within a few hundred power cycles. A thixotropic non-curing compound with low slump resists it; nothing eliminates it, so the design has to tolerate a slow rise rather than assume a constant value.

Dry-out after two to three years. Oil leaves the filler network, the thermal interface material becomes a dry film with visibly higher resistance. This is the mechanism behind the familiar "the laptop got slower after two summers" complaint, and it is a material property that should be specified, not a user error.

Application variation between operators. Hand dispensed thermal paste dots on a 100 mm2 die vary by a factor of two in mass, which is a factor of two in bond line and up to 2 K in result. Volume dispensing or stencil printing is a process decision, not a cost decision.

Mounting pressure unevenness. Four spring screws on a thin cold plate produce a pressure map, not a single value. If one corner sees 10 psi and another 30 psi, the paste thickness follows, and the die hotspot lands where the pressure is lowest.

Migration into places it should not go. Low-molecular siloxane from a silicone-based thermal paste can fog a camera window or deposit on a display backlight within the same chassis over years of use. For thin notebooks this volatility check belongs in the specification.

Selecting a grade rather than a brand

The TSAD Series from TOUSEN is a solvent-free, non-curing, thixotropic paste in gray, spanning 1.2–6.5 W/m·K with measured impedance from 0.201 down to 0.03 °C·cm2/W at 30 psi, specific gravity 2.5–2.9 g/cm3, and a 12-month shelf life. In notebook assembly a thermal interface material is chosen by the flux of the specific location and by the bond line the dispensing process can repeat, not by the highest number on the datasheet.

How much paste, in milligrams

Convert the target bond line into mass and the guesswork disappears. At a specific gravity of 2.7 g/cm3, a 30 µm joint needs about 8 mg per cm2 of die area: a 100 mm2 die takes roughly 8 mg, a 150 mm2 GPU about 12 mg. Weighing a dispensed dot on a balance is a one-minute check that prevents both starvation and overflow, and it is far more repeatable than judging by eye.

Flatness and pressure, not just conductivity

Measure the cold plate flatness and the die warpage before choosing the grade of thermal paste. A cold plate that is concave by 0.08 mm cannot be rescued by a better paste; it needs a compliant thermal interface material, a thicker joint, or a redesigned stiffener. Likewise, a pressure map below 10 psi at the die corners means the effective joint is thicker than the drawing claims, whatever the datasheet says.

When a pad is the better component

On SSDs, on the underside of a mainboard, on the hinge and on chassis contact points, a silicone pad is the correct part: it absorbs tolerance, damps vibration and never pumps out. Forcing paste into those positions trades a robust joint for a marginal one.

Qualification that matches notebook usage

Test the assembled module, not a coupon, because laptop thermal management lives or dies on the assembled pressure map. Run 1,000 power cycles between idle and sustained boost while logging junction temperature, then repeat the impedance measurement on a sacrificed unit. Add 500 hours at 60 °C and 90 % RH for the humidity case, a drop and vibration sequence to the chassis standard, and a gravimetric oil separation test at 150 °C. Where a camera or display shares the enclosure, add an outgassing check. Shelf life is 12 months, so paste near the end of its window should be re-qualified rather than used on a premium SKU.

Assembly and service rules

• Clean the die and the cold plate with isopropyl alcohol; a fingerprint under a 30 µm joint is a measurable hotspot.
• Dispense thermal paste by volume or print it by stencil, and verify the mass on the first unit of each shift.
• Tighten the four screws in a diagonal sequence in two passes so the pressure map stays symmetric across the die, which is the whole point of laptop thermal management in assembly.
• Never top up an existing joint during service: remove the old film completely, since two layers read worse than one fresh layer.
• Store the cartridge closed at room temperature and away from open flame; the product is non-hazardous but not indefinitely stable.

Conclusion

Laptop thermal management is a chain of very short resistances, and the interface is the one link the assembly process can still get wrong after the heatsink is designed; the paste chosen for a 35 W/cm2 die and the mass dispensed on it set the result. Specify thermal paste by the flux at each location and by the bond line the process can repeat, size the dispense in milligrams rather than in "a pea-sized dot", and verify the result on an assembled unit rather than on a coupon. Between a 30 µm joint and a 100 µm one at 35 W/cm2, there are more than two kelvin, which is the difference between a fan that stays quiet and one that does not.

Impedance data, conductivity grades and packaging options for the thermal paste described in this article are listed on the TSAD Series thermal paste product page, where samples can be requested for dispensing trials on your own cold plate.