from an NC1. Meanwhile, companies that rely on Windows applications will fare better with a NetPC.
Cost of ownership might be a
nother motivation for buying an NC. But the hardware dollar savings associated with NCs might not be as high as people think. Costs simply shift toward servers. The centralized, server-dependent, thin-client model creates the need for additional server memory, backup, and administration. Still, value and price-per-seat considerations of NCs are not a moot point. When every dollar counts, these factors still impact purchasing decisions.
To help you decide, for this review we did an analysis of NetPCs and NC1s in three areas: performance, price, and features. We weighted performance at 60 percent, features at 30 percent, and price at 10 percent. Because all the units tested were early prototypes, no documentation was available. Therefore, usability, which focuses mainly on how easily a unit sets up in accordance with its written documentation, was not included as part of the scoring.
Performance
If your infrastructure can accommodate an NC, you'll want to look at how well the unit performs.
For our
performance tests
, we connected the NC1s, which run proprietary OSes, to a PC server running Citrix-compatible Windows NT 3.51. We connected the NetPCs, which ran NT 4.0 Workstation, to a PC server running NT 4.0 Server. However, since we were primarily testing the clients, we deemphasized their network functions and network performance in the final ratings.
In addition, because NC1s and NetPCs have fundamentally different architectures, we didn't test them in head-to-head competition. We assumed that real-world NC1s would be deployed in a corporate intranet and attached to a set of applications and data servers. When doing the NetPC tests, we assumed that a corporation would deploy them as a direct, low-administration replacement for standard desktop PCs.
We ran the standard NSTL InterMark graphics-intensive benchmark on both flavors of NCs. InterMark includes a video test of primitive GDI operations, as well as the display of entire pictures generated by common ap
plications, including CorelDraw, Excel, Freelance Graphics, Powerpoint, and Word. All tests are performed by drawing into memory, as well as drawing directly to the screen. The current procedure tests bit-mapped images in 160 by 120 and 320 by 240 pixels. InterMark also tests native-format, monochrome (4-, 8-, and 32-bit), and device-independent bit-mapped images. Other primitive operations tested include PatBLT, line draw, polygons, and ellipses.
Features
For the features score, we added up all extras that went beyond the initial specifications. We included such things as whether or not the Mac address could be both permanent and changeable.
In addition, systems with a smartcard interface or support, PC Card support, and MPEG-1 support were given top honors. Power buttons with multiple functions, hot-key reset buttons, and mouse and keyboard ports were also important features. Finally, we looked to see if the systems included audio, printer support, and a system warranty.
illustration_link (13 Kbytes)
These NCs' ability to deliver the graphics goods varied from excellent to awful.
Test Results
