Benchmarking Performance Enhancements

There are plenty of registry tips and hacks floating around on the internet, and some of them are excellent performance boosters. However, the usefulness of most tips depends on your independent configuration and needs. For example, using a large [Server] cache may indeed speed things up under some circumstances. However, if you don't multitask heavily, and/or have limited RAM (under 64 MB), using a larger-than-default cache may actually cause a considerable slowdown.

With many tips, the speed increases (or decreases) are tiny enough as to possibly be unrecognizable, especially if you're not putting your system through the wringer trying to detect a difference.

Worse yet, if you try many new settings at once, without testing them one by one, the "good" hacks may be undermined by the "bad" ones, and you might not realize any increase in performance.

And some might possibly boost performance in the target area, but flog it in another area. Without testing, it might prove difficult to determine which hack had what effect on your system.

Therefore, it's generally recommended that you run a benchmark program to detect these performance measurements, and let the numbers speak for themselves.

It's not just registry hacks, either, that can increase or decrease performance, of course. New drivers, which can mean more features for certain peripherals and sometimes performance enhancements, can also prove detrimental to system speed. You should test your system after adding a new driver (or service), as well.

Benchmarks

Here are 5 benchmarks you can download now to measure your NT system performance. Then there are 2 others you can order (one is free save for shipping, and the other is about $500).

For a tiny, no-nonsense NT benchmark program, try nbench.zip: it's a teensy 67606 byte download, and it measures CPU, memory, and disk performance. You can grab it from shareware.com.

Nbench is a small program which comes with only two sets of results (in TXT file format). The following programs have more whiz-bang features, including 3-D charting and many sets of results which you can use to compare your systems. They are also bigger to download, and some cost money (and one, a lot of money):

Byte Magazine's Bytemark is a command line benchmark that is as tiny as Nbench, but measures more, but is more difficult to use. Be sure to read the online documentation.

U Software's Bench32 measures CPU, memory, disk, and video performance, and compares them to a baseline system. Their web site also provides a Performance Tips page you should check out. Shareware.