Clock speed, in relation to
processors, refers to the rate at which the "
clock" (just an oscillating
signal produced by a crystal) runs within the
processor. The oscillations of the clock and therefore overall clock speed is measured in
Hertz (cycles per second).
The CPU, in many modern systems, runs an ISA based off of the transmission of electrical signals through transistors. These electrical signals are synchronized withing the CPU by means of a consistent pulse (or the clock). In these synchronized CPU's, one cycle, refers to one pulse from the clock and usually means one action has been completed by the CPU. Asynchronous processors do exist, mostly in research only, and they would not require a clock to run, letting the different parts of the CPU signal the system when they are completed rather than synchronous CPU's where a part of the CPU must finish in one clock cycle. Clock speed would have no meaning for an asynchronous CPU.
Clock speed is frequently used as one type of performance measure since most processors aim to complete one instruction per clock cycle. However, there are many exceptions to this and the only true indication of a processors performance is how long it takes it to execute a program. One example of an exception to the one instruction per clock cycle guide is the Intel Itanium processor which has a theoretical potential for 6 instructions per cycle. This would make it much faster than a conventional microprocessor running at the same speed. However, it frequently doesn't achieve 6 instructions per cylce, and it frequently stalls due to data hazards and memory accesses. Also, many microprocessors vary in the number of instructions they can process per clock cycle and can frequently achieve much less than 1 instruction per cycle.
There has recently been a move away from clock speed as a measure of performance due to AMD's lacking in the clock speed realm, and Intel's sacrificing some performance in order to achieve greater and greater clock speeds.
RPGeek thinks it is worth mentioning that "clock speed does form a good performance metric between two otherwise identical processors, since with the same architecture performance should scale linearly with clock speed."