Overclocking is the operation of processors or other computer components (hardware components) at a higher clock speed above the normal manufacturer’s specification, with the aim of increasing computing power. The opposite of this is underclocking, which is usually used as an energy-saving measure or with the aim of a longer service life.
Occasionally, features that dynamically increase processor clock speed are also referred to as overclocking, such as AMD’s Turbo Core Technology and Intel’s Turbo Boost Technology. However, this is a clock increase intended by the manufacturers within a fixed framework. Increasing the clock rates of components, such as CPUs, graphics cards or RAM, results in an increase in the performance of the overall system.
This is made possible by the fact that manufacturers only design a few versions of a microchip due to the cost of development and production. Furthermore, the microchips produced on a wafer differ in quality in that the heat generation varies at the same clock speed. Small microchips are overclocked by replacing the clock generator (a quartz crystal).
---
In order to keep the system stable even at higher clock speeds, the operating voltage (in the case of CPUs, the core voltage) is often increased. However, this can lead to overheating and other problems. The clock speed of a component of a computer can usually be set in the BIOS and/or by special software during operation (app in case of Android). Older motherboards also often have jumpers for frequency adjustment. In CPUs, the clock is made up of a reference clock and a multiplier. For example, a reference clock of 200 MHz and a multiplier of 10 will result in a processor clock of 2000 MHz. The reference clock is classically the clock of the Front Side Bus (FSB); however, since modern CPUs no longer have an FSB, a different clock, for example the uncore clock, is used today. In many cases, an increase in the reference clock also leads to an increase in the clock speed of components other than the CPU (e.g., buses or memory). Many CPUs have a multiplier locked, so they can only be overclocked via the reference clock.
Hazards of Overclocking
Overclocked systems can become unstable and produce (computational) errors. These manifest themselves in a system crash, crashing programs or similar undesirable events. In an empirical study by Microsoft Research with one million computers and laptops examined in 2011, a significantly higher risk of crashing overclocked systems was observed. In particular, one-bit flip errors in the DRAM have been detected. Even systems that were only slightly overclocked had a higher rate of system crashes.
Overclocking a PC system is usually a step-by-step process. Stability tests are carried out after each increase in a clock. Stability after overclocking is often tested with programs that use the component to its full capacity, such as Prime95 for processors. If the overclocked system survives the several hours of so-called torture tests, it is usually considered stable.
Operating components outside of their specifications will void the manufacturer’s warranty in almost all cases. In addition, the life expectancy of overclocked components can sometimes be significantly reduced.
Due to a higher clock speed and, above all, the increased voltage, electromigration is favored, which became known as Sudden Northwood Death Syndrome, for example. The power dissipation increases linearly with the clock and quadratically with the voltage.
Another source of danger is the heat generated during overclocking. While overclocking or under-cooling or cooling failures in older CPU models led to overheating damage, there is now a technique called throttling that prevents the processor from getting too hot by simply skipping clocks. However, this can lead to lower performance and thus counterproductive overclocking.
