Lloyd Robert Borrett

  search
 
Follow Lloyd on Twitter Friend Lloyd on Facebook Connect with Lloyd on LinkedIn Follow Lloyd on Pinterest
Computing

RISC versus CISC

by Lloyd Borrett
Australian Personal Computer, June 1991

As the microprocessors used in personal computers become faster and more powerful they are increasingly being compared with the processors used in the low-end workstation and minicomputer market. Essentially the debate becomes a comparison of Reduced Instruction Set Computing (RISC) and Complex Instruction Set Computing (CISC).

History of RISC and CISC

RISC methodology was developed at IBM in the late 1960s before microprocessors were even invented. Yet despite that, until recent times, all of the major manufacturers of microprocessors used CISC based designs.

The reason for this is that microprocessors were introduced in the early 1970s to be used in simple electronic products such as calculators, stereos and video games, not personal computers. CISC technology was more suitable for this application.

Eventually these CISC based microprocessors found their way into personal computers. The PCs evolved with generation after generation of CISC based microprocessors being introduced to meet the ever increasing demands of PC users. The microprocessor manufacturers began to refocus their CISC microprocessor design efforts away from general purpose designs to a high-performance computing orientation.

RISC Comeback

RISC designs made a comeback in the mid 1980s via workstations. The first workstations were based on CISC designs, but the workstation manufacturers were keen to find a technology that broke the market control of Intel and Motorola.

RISC based designs were able to be produced and implemented faster than CISC based designs. More importantly the workstation manufacturers could be seen to be offering a very different product than that offered by personal computer manufacturers.

Performance Comparisons

An instruction is the lowest level command that can be executed by the microprocessor. The instruction set is simply the list of instructions that a microprocessor will understand. CISC microprocessors typically have 100-150 instructions, while RISC microprocessors tend to have 70-80 instructions.

Many of the extra instructions on CISC microprocessors are really supersets of the simpler instructions in a RISC instruction set. Thus CISC microprocessors are complex in two senses: they have more instructions and some of these instructions are complex.

The speed at which a computer can accomplish any task is a function of the number of instructions that the task has to be broken in to, the number of cycles it takes to complete one instruction, and the number of cycles that can be accomplished in one second.

Speed = (Instructions/Task) * (Cycle/Instruction) * (Time/Cycle)

CISC microprocessors aim to achieve speed by having such a rich instruction set that it takes fewer instructions to complete a task. RISC microprocessors require more instructions to complete a given task, but they focus on reducing the number of cycles it takes to complete an instruction. Clock cycles per second is largely independent of the microprocessor architecture.

The case for higher performance of RISC based microprocessors is based on the premise that superiority in instructions per clock cycle (the MIPS rating) more than offsets the handicap of requiring more instructions to complete the task. This is highly situation dependent and the factors involved include system hardware configuration, memory speed and capacity, as well as the application software.

The performance improvements offered in current RISC based workstations has very little to do with inherent differences in RISC/CISC architectures. The Intel 80486 and Motorola 68040 are the first generation of CISC microprocessors that specifically set speed as a performance objective. There performance compares very well with RISC microprocessors.

The factors that make the difference are the non RISC/CISC parts of the systems, that is, register capacity, cache memory, pipelining and floating point processing. Thus it would be easy to see CISC processors evolve and match the RISC performance, in part by adopting some of the features currently associated with RISC but that are in fact independent of it.

We believe that the future will see a convergence of microprocessor architectures as the CISC designers adopt some of the features of RISC microprocessors, and RISC microprocessors enlarge their instruction set so as to improve performance.

In fact we're already seeing this. The IBM RS/6000 microprocessor that powers IBM's RISC System/6000 systems is actually more complex than the CISC based design of the Intel 80386. The RS/6000 has 184 instructions and the 80386 has 150 instructions.

Summary

  • RISC and CISC are merely different design methodologies. Neither is inherently superior to the other.
     
  • CISC architectures are equally capable of evolving to the levels of performance being claimed by RISC.

The performance of microprocessor based computers is a function of more factors than just the microprocessor.

Last modified: Saturday, 15 October 2011

 
 


 

 
home | about | weird mob | computing | interests | insight
Copyright © 1995-2014 Lloyd Borrett. All rights reserved.  ::  www.borrett.id.au
mob 0418 170 044  ::  tel +61 3 5904 9005  ::  email lloyd@borrett.id.au  ::  skype lloyd_borrett
twitter @borrett  ::  facebook lloyd.borrett  ::  linkedin lloydborrett