tech TALK

I have found an easy question that gauges someone's familiarity with the computer world: What kind of computer do you have? When I ask a non-technical person the question, the response is a brand-name, like Compaq, Gateway 2000 or Apple. When I ask a "computer person" the question, the response is the name of the computer's central processing unit (CPU), such as 486 or Pentium.

The difference between the responses is a fundamental one; but it is frequently misunderstood outside of the computer world. The name on the front of the computer's case is simply the name of the company that put the computer together. A CPU is the single-chip brain of modern personal computers that performs the most basic computer operations, like addition or multiplication.

The CPU--and not the brand--defines the type of computer on your desk, although some brand-names are strongly linked to a microprocessor--for example, Macintosh is linked to both Motorola's 68000 and the PowerPC chip. When the computer store salesperson says that a computer conforms to the IBM PC standard or is Windows compatible, he or she is really referring to the computer's microprocessor.

Microprocessors can be divided into two groups: CISC chips and RISC chips. The two architectures are fundamentally different approaches to processor performance, which is the measurement of how long it takes for the chip to perform a task using three factors: cycles per instruction, time per cycle and instructions per task.

Cycles per instruction and time per cycle together determine processor speed; if a chip can execute one instruction per cycle and can operate at 100 million cycles per second or 100 MHz, the computer can execute 100 million instructions per second or 100 MIPS.

The CISC (an acronym for complex instruction set computer) approach is to have many instructions and few instructions per task. The RISC (an acronym for reduced instruction set computer) approach is to cut down on the number of instructions the microprocessor understands, but to execute the small number of instructions more quickly.

The RISC approach emerged from statistical analyses of existing programs. The analyses showed that most compilers (programs that translate higher level languages like C and BASIC into machine code) rarely used complex instructions. Rather, the computer spent most of its time only executing a small core of instructions.

In other words, a RISC processor reduces the number of cycles per instruction and reduces the time per cycle, but increases the number of instructions per task. A CISC processor, on the other hand, requires fewer instructions per task but typically processes fewer instructions per second. However, since compilers rarely generate complex instructions, RISC chips should be able to be pushed faster and faster, while CISC performance must plateau.

Each architecture represents a series of tough trade-offs. CISC's major problem is that a processor which allows many instructions is difficult to design and difficult to speed up; RISC chips are easier to develop and easier to speed up. The RISC architecture, however, requires greater memory and hard disk space than the CISC architecture, as most PowerMac owners will readily attest.

In theory, therefore, RISC chips are the wave of the future and CISC chips are a technological dead-end. Most microprocessor companies have voted with their feet by shifting design resources to RISC-based architectures. MIPS, Hewlett-Packard, Digital Sun Microsystems, IBM, Apple and Motorola are all converts.

In practice, CISC chips in the form of Intel's x86 family are ingrained in the marketplace and cannot be ignored. But Intel has claimed the P7, scheduled to hit the market in the year 2000, will be a combination CISC-RISC (CRISC?) chip.

Matt Howitt '97, Crimson Sports Editor, is an economics concentrator in Mather House. He founded Twisted Pair Consulting in 1989. He may be reached on-line at mhowitt@fas.harvard.edu. He is not a microprocessor.