clock speed

The transfer of data within the microprocessor and between the microprocessor and memory must be synchronized to ensure that the data needed to execute each instruction is available when the flow of execution has reached an appro-priate point. This synchronization is accomplished by mov-ing data in intervals that correspond to the pulses of the system clock (a quartz crystal). This is done by sending control signals that tell the components of the processor and memory when to send or wait for data. Thus, if the microprocessor is the heart of the computer, the clock is the heart’s pacemaker. Because most devices cannot run at the same pace as the processor, circuits in various parts of the motherboard create secondary control signals that run at various ratios of the actual system clock speed.

The following table shows the speed of various system components in relation to the system clock rate. Although the example uses a 600-MHz clock, the ratios will generally hold for faster processors.

Device         speed Relationship   

  

Processor    600 System bus * 4.5   

System   

(Memory) Bus 133 (depends on multiplier)   

Level 2 Cache 300 Processor / 2   

AGP          66 System bus / 2   

PCI bus         33 System bus / 4  

Microprocessors are rated according to the frequency (that is, number of pulses per second) of their associated clock. For example, a 1.2-GHz Pentium IV processor has 1.2 billion (giga-) pulses per second. It follows that all other things being equal, the higher a processor’s clock frequency, the more instructions it can process per second. An alterna-tive way to rate processors is according to the number of a standard type of instruction that it can process per second, hence MIPS (millions of instructions per second).

The relationship between clock speed and processor performance is not as simple as the preceding might imply, however. Each processor is designed with circuits that can move data at a certain rate. In some cases a processor can be run at a higher clock rate than specified (this is called overclocking), but then reliability comes into question. Also, the actual processing power of a processor depends on many other factors. If a processor implements instruc-tions in its microcode that are more efficient for handling certain operations (such as floating point math or graphics rendering), applications that depend on these operations may run faster on one processor than on another, even if the two processors run at the same clock speed. The speed of the system bus (which connects the processor to the RAM memory) also affects the speed at which data can be fetched, processed, and stored. A processor with a clock speed of 733 MHz should perform better on a motherboard with a bus speed of 133 MHz than on one with a bus speed of only 100 MHz.

Speed is “sexy” in marketing terms, so the major chip manufacturers always tout their fastest chips. However, the difference in speed between, for example, a 2.2-GHz version of a processor and a 2.0-GHz version may be unnoticeable to the user of all but the most processor-intensive applica-tions (such as image processing). Indeed, if the system with the slower chip has a faster bus, faster memory (such as RDRAM), or a larger processor cache (see cache) it may well outperform the one with a faster chip.

Another reason for caution in interpreting clock speed is that many recent PCs have two or even four proces-sors (see multiprocessing). Performance in such systems is likely to depend at least as much on optimization of the operating system and applications as on any multiple of raw clock speed. This trend to multicore CPUs is also seen as an alternative to any substantial increase in processor speed, because higher speeds bring increasing concerns about heat and power usage.

In PCs the term “clock” can also refer to the battery-pow-ered “real-time” clock that provides a timing interval that can be accessed by the operating system and applications.