Impact of data center cooling strategies on component reliability

In a data center environment, the temperature that a chip experiences is an important parameter determining its reliability. A high operating temperature can cause it to wearout prematurely. A low temperature results in additional cooling costs to operate the data center. In typical data center operation, there arises a distribution of component temperatures that are sensitive to the specific cooling strategies being used. This paper evaluates the impact of various data center cooling strategies on component reliability focusing on the gate oxide time dependent dielectric breakdown (TDDB) mechanism on a 32nm chipset component.