RTX 5070 Ti caught throttling at 107°C — Nvidia hid the hotspot temp sensor

Nvidia quietly removed public access to GPU hotspot temperature monitoring on RTX 50-series cards — the sensor reading that tells you if your die is cooking under a bad cooler mount. Tom's Hardware reports that only Nvidia's internal MODS diagnostic tool can still see it. Testing with that tool caught an RTX 5070 Ti throttling at 107°C because of a botched thermal interface material (TIM) application during assembly.

The hotspot sensor was visible in tools like HWiNFO on RTX 40-series and older cards. On Blackwell, it's gone from third-party software entirely. The card in question had uneven TIM spread — a known manufacturing defect that creates thermal runaway on one part of the die while the average GPU temp reads normal. Without hotspot visibility, you can't diagnose that until performance tanks.

This is a step backward for DIY troubleshooting. If your RTX 5070 Ti or 5080 runs into thermal issues, you won't know the die is 15-20°C hotter than the average package temp unless you RMA and hope Nvidia's internal checks catch it. The lack of transparency makes it harder to tell a bad mount from inadequate case airflow.

If you're running a Blackwell card and see unexplained performance drops or stuttering under load, check your case temps and fan curves on BetterFPS's optimization hub — it's the best you can do without the sensor data Nvidia used to let you see.