How We Tested: Windows-Based Terminals
When we conducted our tests, the products included in this review were in various stages of development--ranging from prebeta to limited-volume shipment. The version of Windows CE 2.1 that serves as the basis for these products was released to resellers in late August--yet another example of Microsoft's traditionally late OS shipments. Once a reseller receives this code, it must adapt the code to its hardware platform by adding device drivers and other desired system functionality. After that task is completed, tuning can take place. As such, caution must be exercised in evaluating performance of these devices prior to shipment. Because of this, we only weighted performance as 20 percent of the overall Report Card score; under normal circumstances, it would have been substantially higher for this type of product.

The primary benchmark utility we used in our testing was WinTach version 1.2, which simulates common tasks, including word processing, spreadsheet, CAD and paint applications, and measures performance relative to a standard test platform. The use of a 386 PC as a reference platform is indicative of the age of this utility, but we chose it nonetheless because it did an excellent job of stressing the video I/O functionality that lies at the heart of any remote-presentation client device. We validated these benchmark results with informal tests of common applications, including Microsoft Word, Excel and Internet Explorer, and are confident that the benchmark results provide an accurate reflection of user-perceptible performance differences.

The server used in all our testing was a Micron Millenia Pentium Pro 200 with 128 MB of RAM running Microsoft Windows NT Server version 4.0 Terminal Server Edition, build 419.3. We also loaded Citrix MetaFrame version 1.0 for our ICA testing. All WinTach tests involved both the Microsoft RDP protocol and Citrix's ICA protocol with compression enabled. All tests were conducted on an isolated LAN segment with the average of the three tests reported. To provide a performance comparison with a standalone desktop computing platform, we also ran WinTach locally on a Hewlett-Packard Co. 166 Pentium computer. The WinTach results shown above represent relative performance compared to that of the P-166 machine.

As expected of remote-presentation devices that virtualize display functions, the Windows-based terminals performed better on text-oriented applications than they did on graphics applications.