Nomenclature is funky. What most refer to as a gpu, is actually a graphics/video card and the gpu is the graphics processing unit, same as a cpu. The only place on video cards back then to actually see a temp sensor was inside the core of a gpu. Temp sensors relating to the vrms/vram are nowhere to be found. With just the one fan running, if it's above the gpu, you'll see regular temps, but only passive radiation is affecting the rest of the card, including the vram/VRM's etc that aren't directly next to the gpu. So everything at the output end sees nothing. You could have temps beyond 90°C on the vram, and the gpu will still read 55°C.
Conversely, if the dead fan is the one above the gpu, and the vrm/vram fan is still running good, then only the gpu area is affected, which won't affect the gpu as much as it's surrounding chips those could hit 90+ and the gpu still read 55°C.
What's read from the gpu has little bearing on the rest of the card, except under normal operations where sufficient cooling for the gpu affects all the cooling across the board.