real-timespeed.(Product Information)

How real does it get?

An engineer's need for speed never seems to be satisfied. With today's high-speed, multi-channel, plug-in I/O cards being capable of transferring hundreds of thousands of samples each second, eliminating performance bottlenecks has become an essential element for ensuring measurement integrity. And as data transfer rates increase, the need for faster processing escalates.

Using multiple CPUs to boost the available computational power is one way to generate faster system speed--especially with the recent availability of $2000-range dual-processor desktop PCs. Yet, until Windows 2000 arrived, adding CPUs to a Windows-based platform barely improved its performance when running standard business software applications.

This latest incarnation of Windows actually degrades the performance of even the fastest Pentium III single-processor computers and requires the equivalent of a 200 MHz CPU upgrade to match the application throughput of NT. Dell Computer Corporation (www.dell.com) has determined that dual-processor machines enjoy surprising performance improvements with Windows 2000, even when running ordinary software.

When Dell tested Office 2000 with this new operating system, the addition of a second processor boosted system performance 21%--the equivalent of a 305 MHz increase in CPU clock speed. The same scenario under Windows NT demonstrated no measurable speed improvement. Simulating a more complex, data-centric computing environment resulted in a 39% performance increase. Therefore, a shorter product development cycle may be possible if the application can interleave elementary code activities ("threads") to distribute instructions among the available CPUs. In other words, multithreaded applications can cooperate for system resources instead of competing for them.

Who's on first?

Unfortunately, interleaving thread calls in a standard programming language, such as C or Visual Basic, is a difficult undertaking. Some applications, like ...