Thermal Throttling Mechanics and Liquid Metal vs PTM7950 Phase-Change Thermal Pads: Fundamental Mechanics and Deep Mathematical Modeling

[Nathan Peterson]

I’ve been digging into why liquid metal consistently beats PTM7950 in thermal resistance measurements, and I want to understand the physics beyond just “it has higher W/mK.” Liquid metal sits around 73-80 W/mK while PTM7950 clocks in at roughly 12 W/mK once phase-changed, yet real-world temperature deltas on the same die are often smaller than that ratio implies. Can someone walk me through how interface thermal resistance and the gap-fill behavior of a phase-change pad actually compensate for the raw conductivity gap? At what point does the lower conductivity of PTM7950 stop mattering against the interface bottleneck?

My bigger concern is the pump-out effect over long competitive sessions. Everyone warns about liquid metal migrating and shorting traces, but how does the phase-change mechanism in PTM7950 actually resist pump-out mechanically, and is there any quantitative modeling on viscosity and re-solidification cycles that predicts pad longevity? Where’s the thermal gradient threshold where a hotspot on the IFSI starts throttling before the average die temp does, and can we model that with something practical rather than just trusting the vendor curves?

"[CPU: i9-14900K] • [GPU: RX 7900 XTX] • [RAM: 16GB DDR5 5600MHz]"

[Ethan Rivera]

Here’s the thing most people get backwards on throttling: throttling isn’t triggered by ambient or case temperature, it’s driven by the junction-to-interface gradient. The core equation is Q = ΔT / Rth_total, where Q is the power being dumped (watts), ΔT is the temperature differential across the thermal stack, and Rth_total is the sum of every resistance in series — junction-to-die-spreader, spreader-to-package, package-to-TIM, TIM

"London, UK • Ryzen 9 7950X3D & RTX 4090 24GB"

[Author]

Posts