Surachai Locharoen wrote on Saturday, January 06, 2007 12:42 AM -0600: > polling_interval was set to 2, I changed this to 30 and observed that > sometimes the CPU temp spiked at 102, 105, 107 but for no more than 1 > second then immediately dropped back to sub-100. No instability, so > could be a glitch? > > Sometimes Linux will hit 100+ on 30 seconds and halt.
If the readings of around 100 degrees C are accurate, this is too hot for a processor die. While you can run power semiconductors that hot and still achieve acceptable reliability, this is not reasonable for a package like a processor. The primary failure mode is bond wires opening, and it is strongly temperature dependent. A typical power device has 3 bond wires, while a processor has several hundred, so the overall failure rate for this type of package gets too high at a much lower temperature. -- Seth Goodman
