Solution:
Root Cause: GPU Hotspot / GDDR6X VRAM Thermal Protection Cutoff
Graphics cards contain multiple internal temperature sensors. While standard 'GPU Core' temperature may show an acceptable 70°C, the
GPU Hotspot (highest single sensor point on the die) or
GPU Memory Junction (VRAM chips) can reach 105°C–110°C+. Excess thermal delta (>25°C between core and hotspot) indicates pump-out of thermal paste, degraded thermal pads, or warped cooler mounting pressure, causing the GPU driver or hardware thermal monitor to trip safety shutoffs.
Diagnostic Verification:
1Open HWiNFO64 and expand the GPU sensor section.
2Observe GPU Core Temperature, GPU Memory Junction Temperature, and GPU Hotspot Temperature under load.
3If GPU Hotspot exceeds 105°C or GPU Memory Junction hits 110°C, the card is triggering emergency thermal protection.
Step-by-Step Fix:
1Replace GPU Thermal Paste (Repaste):
Disassemble the graphics card shroud and heatsink assembly.Clean degraded dry thermal paste off the GPU die.Apply high-viscosity, non-pump-out thermal paste or phase-change thermal pads (e.g., Honeywell PTM7950) to prevent paste migration over time.2Replace Degraded VRAM Thermal Pads:
Replace torn or dried thermal pads over VRAM chips and VRM MOSFETs using exact manufacturer-specified pad thicknesses (e.g., 1.0mm, 1.5mm).3Increase GPU Fan Curve:
Use MSI Afterburner to set a more aggressive custom fan curve (e.g., 80% fan speed at 75°C).Prevention & Long-Term Monitoring:
Use phase-change thermal pads (PTM7950) on bare-die GPUs to prevent thermal paste pump-out under cyclic heating.