NVIDIA quietly removed the ability for regular users to read the "hotspot" temperature on its latest RTX 50 series graphics cards, built on the company's Blackwell architecture. That sensor measures the hottest single point on a GPU, rather than the averaged reading most monitoring apps show.
According to Tom's Hardware, the sensor was never actually deleted — it remains accessible through NVIDIA's own internal diagnostic tool, called MODS. VideoCardz.com reports the same: the hotspot data is still readable if you can get into that tool. Enthusiasts have used mods to pull the numbers out.
What that data shows is unflattering. Tom's Hardware describes an RTX 5070 Ti that was caught throttling — deliberately slowing itself down to avoid damage — at 107°C, which it attributes to a poor application of thermal interface material (the paste-like layer that moves heat from the chip to the cooler). Wccftech frames the findings more broadly, reporting "widespread thermal issues" across RTX 50 cards that can throttle gaming performance.
The implication drawn by Tom's Hardware is pointed: the fact that these GPUs can overheat and throttle so easily "could be why the sensor was kept hidden in the first place." In other words, hiding the reading may have obscured a manufacturing quality problem from buyers.
Why it matters: gamers and PC builders spend hundreds or thousands of dollars expecting full performance, and a card that silently throttles delivers less than advertised — so whether a company hides the data that would reveal that is a question of transparency, not just engineering.