Most consumers believe that a computer’s speed is determined by its' rating in megahertz (MHz). Megahertz is certainly part of the story, but is far from determinative.

Megahertz, or million cycles per second, describes the computer's internal processor’s clock rate. Think of this as like rpm (revolutions per minute) on your car’s engine as displayed on the tachometer. Most of us can picture what a revolution is inside a car engine, but what about a computer processor cycle?

A clock cycle is the lowest common denominator of a processor. A computer instruction, such as an "ADD" entails several clock cycles worth of work. First you have to fetch the first number from memory and store it in "Register A" (a register is a storage space inside the processor), then fetch the second number and put it in "Register B," a binary add then takes place with the result being stored in Register S (S for sum) which must finally be moved to a location in memory. So a simple ADD will take several clock cycles to accomplish. In this case, the faster the megahertz of the clock the faster the completion of the ADD instruction. But measuring the speed of a computer system means looking at more than how fast it can add.

Think of how your car engine's rpm relates to speed. Two cars traveling down the same road, both at 3000 rpm, are not likely to be moving at the same speed. Why not? Because it depends on a number of factors, including gear ratios, the car’s drag coefficient and how "powerful" the engine is. Your computer, however, does not ordinarily move (laptops excluded) and speed over ground is not a valid measure. The real measure of a computer's speed is in units of work performed over time.

Today, we have many benchmark tests designed to determine the speed of a computer system. Probably the best benchmark data comes from using real world tasks, similar to what you and I would use a computer for, not floating point or integer arithmetic. Many factors must be examined to understand why one computer system is faster than another performing these tasks.

A computer's processor must communicate with the other components of the computer system. This is done over the computer's "bus," or data pathways on the "mother," or main circuit, board. Bus speeds vary and are usually one of the following: 40, 66, 80, 100, 133 or 200 megahertz. 40, and more recently, 66 megahertz are more common in laptops while most desktops have either a 66 or 100 MHz system bus. The newer 133 and 200 MHz system bus specifications are used only very high end systems.

The mother board may have "slots" for adding cards to the computer. The slots introduce a new type of bus which has its' own specifications. Most PCs retain ISA slots for compatibility with older peripherals (Legacy) as well as PCI slots. ISA is slow, and PCI not much faster at 33 or 66 MHz. Further, the data path over the ISA bus is 16 bits wide and 32 bits wide for PCI. Think of lanes on a highway when considering data paths. At any given fixed speed, a highway 32 lanes wide can move twice as many cars over a period of time as a 16 lane road. You can begin to see that that 600 MHz super processor is not necessarily the weak link in the chain when performing real work that involves communicating over a 100 MHz system bus to a device on the other side of a card plugged into a 33 MHz PCI slot.

The system bus will also have a fixed width. The system bus on most PC's is 16 or 32 bits wide. Since memory (RAM) is on the mother board, communication between the processor and memory is limited by the system bus speed and width. And if the memory itself is not "up to speed," several "wait cycles" (latency) may have to be inserted before the requested data is ready to be sent to the processor, kind of like the stoplight meters on freeway entrance ramps letting only one car from each lane enter per cycle.

And all computers must have an operating system, the main program under whose supervision all your applications run. These operating systems have grown exponentially in their hunger for memory and hard disk space. So it goes without saying that the more memory your computer system has the faster it will run (caveat, Windows 95 does not work with more than 64 megabytes of RAM).

Hard disk speeds and controllers also effect the computer systems seed in real world tests. Hard disks spin at a fixed rate, usually 4500 rpm (laptops), 5400 rpm, 7200 rpm or 10000 rpm. The faster they spin the faster they can transfer data to the controller, which usually sits in a slot on the motherboard and is limited by the slot bus in the amount of data it can transfer in any period of time.

Finally, the processor's architecture has much to do with the overall speed of the computer as a system. Remember that a 1.4 liter four cylinder engine can turn at 3000 just as it's 5.4 liter V8 cousin. But whether your goal is 0 to 60 or towing a big trailer, the V8 is likely to prevail.

Whether a processor has one, two or four instruction registers, whether it is a CISC (Complex Instruction Set Computer) or a RISC (Reduced Instruction Set Computer), chip whether it is interrupt driven or not and the amount of on-chip and high speed cache memory available are some of the design elements that determine the difference between a puny four cylinder and a big V8, processor-wise.

So, when Apple Computer states that their 400 MHz G4 computer beats the pants off of an 800 MHz Pentium III running Adobe Photoshop, believe it.

Back

 © 2000, Jim Schoenberger