Don't simply judge a PC by its processor clock speed alone
Speed sells. An average user judges a PC by its processor clock speed, but a power user
knows that hard-drive access time, video accelerators, processor efficiency and memory
speed play equally important roles.
There's always one component that limits performance. You must loosen that bottleneck
to see any real performance improvement.
The component most likely to be the limiting factor changes as one part or another
leaps a generation ahead. Hard drives with built-in caches can really push a system these
days, and video is pretty darn quick. Processors can perform more than one operation per
clock cycle and even estimate which data will be needed next.
But dynamic RAM, the inexpensive memory used in average computers for everything but
video and cache memory, has only speeded up its access time from 100 nanoseconds to about
50 ns. in the last seven years.
If you have a system with other high-end components, the usual bottleneck today is the
PC vendors push several kinds of memory, some more suited to notebooks because of lower
power consumption. It's difficult to make a blanket endorsement of one type; other
components may have just as great an effect on overall system performance.
Doubling the memory cycle speed (halving the access time measured in nanoseconds) only
improves performance by about 25 percent or 35 percent, but only if there's no other
That can be a significant improvement, but performance never doubles just because
memory speed doubles. Factor that into your cost-effectiveness analyses of competing
Here are some current memory buzz words:
In general, DRAM is outmoded for today's machines, and fast-page mode DRAM is on the
way out. Expect to see SRAM mostly in Pentium Pro systems for now. It's faster than EDO
DRAM but more expensive.
In Pentium notebooks, EDO is popular for relatively low power consumption combined with
enough speed to eliminate the need for a separate Level-2 cache.
Insist on SRAM in 150-MHz and faster PCs and servers. It exacts a power penalty on
battery-operated systems and may not be available for them, so carefully check real-world
performance of models faster than about 133 MHz against the next-slower model. If both
have EDO or burst EDO, the extra cost of the faster clock-speed unit may not be justified
by actual performance.
EDO is a benefit for PCs faster than 60 MHz to 90 MHz. BEDO benefits those faster than
about 133 MHz.
Real-world testing is the only way to ensure that a particular implementation actually
has better performance. This column certainly isn't the last word on memory technology. By
the time you read this, there might be a new kind of memory in production.
John McCormick, a free-lance writer and computer consultant, has been working with
computers since the early 1960s. He welcomes mail from readers. Write to him care of
Government Computer News, 8601 Georgia Ave., Suite 300, Silver Spring, Md. 20910.