Increasing clock speed
All other things being equal, a processor with clock speed or frequency of 2ghz should execute the same machine code program twice as fast as a processor from the same family with clock speed 1ghz. This makes sense if we assume that the machine code instruction is executed in once clock cycle or tick in both processors. As the 2ghz processors clock cycle is half the duration of the 1 GHz processors clock cycle,, it will take half the time to execute an instruction.
Limits on clock speed
A limit has to be set on clock frequency, because the heat generated ion the chip by higher clock frequencies cannot be removed quickly enough. The problem is made worse when more transistors are packed into the same space. The reason for the deviation from Moores law is that packing more transistors into the same area increases the power consumption which results in more heat. A similar consequence occurs when the clock frequency is increased. If transistor density and clock frequency continued to increase it would eventually generate more heat per square centimetre than is generated on the surface of the sun.
Multicore processors
The strategy now is to use the increase in transistor density to put more than one processor into the microprocessor chip. These processors are called cores and operate at lower frequencies than single-core processors to overcome the heating problem caused by the clock frequency. With more than one processor per chip multiple tasks can be run at the same time or single task can be split across several processors. Multicore processors are ideal for multimedia work. Microsoft Windows Movie Maker 2.0 and Adobe Photoshop CS are examples of widely used applications that show significant performance improvement on multicore systems.
Increasing word length
The word length of its registers affects the processors speed performance. The registers that can do arithmetic operations- the general-purpose registers and accumulator- usually have a similar word length. Typical word lengths are 32 bits and 64 bits in modern processors. The bigger the word length, the bigger the operands and results they can accommodate. When operands and results exceed the word length of the registers, extra processing must be done to split operands and results across several registers. This reduces speed performance. Registers used to store the binary codes for main memory addresses affect the number of bytes or locations of main memory that can be used. A longer word means more binary codes and therefore more memory bytes that can be used. The main registers used for addressing main memory are the program counter and the memory address register.
Increasing bus width
The worst penalty occurs in the system bus between the processor and main memory. If the current instruction registers word length is now big enough to hold a complete instruction, the system bus must be used again to transfer one or more operand words into other registers, which slows down the performance of the processor. The system bus is the bottleneck. The larger the data word or instruction word that can pass along its length in one go, the fewer the number of times it needs to be used. In a von Neumann computer, a data or instruction word passes over the data bus. A wider data bus should improve a computers speed performance. Address words pass along the address bus. A wider address bus allows longer address words, so more memory can be addressed.