January 8, 1998

Same box, different results

d5admin Archived Articles, Buying Guides

Question: I have noticed that many of the Pentium systems have identical parts and equipment such as video cards, hard drives, cache, CPU speeds, etc. Why does performance vary between systems? What factors account for the performance difference? Why are some systems more reliable than other systems even though they have the same parts? For example, if I buy the identical parts and equipment for a Pentium system and assemble it precisely, will it be as reliable as one bought from Dell or IBM? Will the performance be different? —Stephen

Answer: I have twin brothers. They’re fraternal and not identical, so they have marginally different DNA code, but darn it if they don’t look alike. Mom even dressed them the same, right down to their booties. They even came off the same production line, 10 minutes apart.

You’d think they’d operate the same, but they don’t. There are subtleties in their makeup that make them different people. And so it is with computers. If two computer systems from the same manufacturer have identical parts, including processors, they will typically run almost identically. But change them to different motherboards, RAM, or even video cards or hard drives, and they may perform noticeably differently.

Compare two similar machines from two different manufacturers and you get results as different as me and Cindy Crawford: Same species, vastly different bathing suit fittings.

Pay particular attention to the chipset on the motherboard, advises David Peterson, vice-president, marketing, of Open Concept International, an Edmonton-based systems company.

A chipset is a collection of integrated circuits designed to be used together for some specific purpose. The latest greatest Pentium chipset from Intel is the TX chipset. If chipsets differ, so may performance.

Also, look at the amount of secondary cache. Most ship with 512 KB, but more or less will cause a difference in performance. Even a different video card and amount of video memory will affect a system’s zippiness.

And, of course, different amounts of memory (RAM), as well as type and speed rating, will greatly affect a system’s overall behavior. Hard drives also have several speed ratings. See an for an overview.

Faisal Premji, a systems technician at Voodoo Computers Inc. of Calgary, points to motherboard manufacturers. An ASUS motherboard can have a better performance index than a Tyan and so on, he suggests. Also, a hard drive controller on a Tyan may be better than one on an ASUS.

Even some manufacturers have proprietary memory-CPU-bus architectures, which enhance performance. Sometimes, though, the more they try to enhance performance, the more money you spend, and it may be for only a marginal improvement, said Peterson. As for reliability, there should be no difference if all components are the same. You typically pay higher prices from the name-brand companies for their warranties and service plans. Often, locally made clones are as good as, or better than, the name-brand machines, if premium parts are used in their assembly.

Reliability problems tend to develop with components that have moving parts. Hard drives have motors which spin usually around 5,400+ RPM. They also have read/write head servos that are constantly moving to access data. Monitors have guns that accelerate electrons within the Cathode Ray Tube (CRT).

These moving parts will tend to wear out over time, said Peterson. Most other parts—including motherboards, CPUs, memory, video cards—are printed-circuit boards and will wear since they expand and contract as they heat and cool.

Components generally have a Mean Time Between Failure (MTBF) rating. These are commonly in the neighbourhood of 100,000 to 500,000 hours of power-on use before the component is expected to fail, explained Peterson. This is the equivalent of 11 to 57 years. If failure is going to occur “BGS” — that’s “Before Garage Sale”—it will likely happen in the first two weeks of system use. That’s typically due to faulty manufacturing.

The only other time a computer component consistently fails is 14 to 30 days after the manufacturer’s warranty expires. Like the Roswell saucer incident, that phenomenon is still a mystery.

When shopping for a computer, I watch for the comparisons made by the U.S. computer magazines. Many of them have labs where they test the top models against each other for performance. Some even publish reliability ratings. Check PC Magazine, PC World, and PC Computing.

If you want to test your own machine against a friend’s, or the machine at work, you might want to download free benchmarking software called Fresh Diagnose.

I have added some places where you can purchase new computers online.