Thermal Throttling Mechanics and Liquid Metal vs PTM7950 Phase-Change Thermal Pads: Advanced Implementation and Real-World Architecture

[Gabriel Turner]

I’ve been running liquid metal on my mobile APU after PTM7950 started showing measurable pump-out after ~400 hours of competitive play. The conductivity gap is real—liquid metal sits around 73 W/mK versus PTM7950’s ~21 W/mK phase-changed—but I’m seeing inconsistent results depending on substrate prep. For anyone who’s pushed these materials into sustained 100% load scenarios: do you anneal liquid metal before application, or apply it cold, and does it actually change the peak-to-avg hotspot delta or just lower the overall junction average? I’m specifically chasing the localized die hotspot that PTM7950 seems to average out but liquid metal apparently concentrates.

On the pump-out front, what’s your threshold for deciding a phase-change pad has failed versus just dried out? I’ve had PTM7950 visibly thin near the IHS edge while still reading thermally fine. And for anyone benchmarking both back-to-back under identical engine loads—what’s your sampling interval, and are you seeing the phase-change material’s thermal impedance curve actually flatten at higher die temps the way the datasheet claims, or does it degrade under sustained clocks?

"Ryzen 7 7800X3D | RTX 4080 Super | Custom Water Loop (EKWB)"

[Author]

Posts