Overclock At Your Own Risk: Continued
How do components die? The simple answer is use, but the longevity of a semiconductor depends on a few factors. While there are a host of things that can make a semiconductor or other circuitry fail, there are three that affect most common components: electrical current, material purity and temperature. While the manufacturing process for semiconductors generally has stringent quality assurance, not all parts come out exactly the same. Flaws exist, and this is something we all have to understand and live with. While that is out of our control, the other two factors generally are not.
In an article by Thomas "Tom" Pabst, he mentions a phenomenon called electromigration, one of many processes that can lead to the degradation of components. Electromigration occurs as a result of metal atoms being moved via the momentum of electrons. Picture this as a sandblaster where the sand, at high velocity, is eroding the walls of the gun. In the case of electromigration, the electrons are moving the metal atoms away from one another. This can cause a circuit to fail by two means: either the atoms are moved apart breaking the circuit, or they are moved closer, so the circuit touches another causing a short. Either way, it is bad news for the components.
You might be saying, "I thought semiconductors were made of silicon?" Yes they are. The reason microchips are made from silicon is that it has the interesting property of being somewhere between a conductor and a resistor; it can allow electrical currents to pass through it or not, depending on what is done to the silicon during the fabrication process. The silicon can conduct electrical currents if impurities exist, so a process called "doping" is used, where the silicon is bombarded by impurities creating positively and negatively charged areas. This is how electrical gates are formed on the silicon chip.
The interesting fact is that semiconductors do not fail from electromigration, as they do not have enough charge carriers. However, when the silicon is doped above a 1% variation from pure, it can conduct electricity and the issue of electromigration can occur. Additionally, microchips have many layers with metal interconnects which naturally are susceptible.
I bring up the fact that conductive materials migrate inside components because there are two major factors that impact the rate at which this happens. The first has been explained - the amount of current flowing through the circuit - and the second is temperature.
In an article about electromigration, Dr. J.R. Lloyd states that "just how much current can be permitted and still maintain reliability as the temperature is changed will depend on whether you have nucleation or growth dominated failure and what the dominant diffusion mechanism is. If we have growth-dominated diffusion and we increase the temperature such that we double the diffusion coefficient (approximately 20 degrees for Al alloys and grain boundary diffusion), we must reduce the current density by half. Conversely, if we want to increase the current density by a factor of two, we must ensure that the temperature is at least 20 degrees cooler. If failure is nucleation dominated, an approximate 30% reduction in current is needed for a similar temperature increase to maintain equal reliability."
In laymen's terms, he is stating that for every 20 degrees Celsius above the ambient test temperature, the current flowing through must be halved. Of course, this is the opposite of what happens while overclocking: not only do we increase the amount of energy passing through the chip (adding to the breakdown effects of electromigration) but we also increase the temperature, since putting more electrical energy through the same circuit heats it. All wires and circuits have to deal with internal resistance. However, interestingly, as temperatures rise, semiconductors conduct better and metals resist more. Therefore, as the flow moves from silicon to metal, the force of the flow will cause the metal to separate, either breaking the circuit or short-circuiting as it comes in contact with another circuit. Due to these and many other factors, adequate cooling is mandatory for overclocking.