What the GCN Lab looked for, and how it looked for it
- By Michael Cheek
- Mar 17, 1997
Thirteen vendors responded. Two were excluded from the final results due to component
failures: The MidWest Micro system's NIC could not communicate with the network, and the
hard drive in Austin Computers' system failed.
All systems were tested for more than a month on the GCN Lab's multiprotocol Microsoft
Windows NT 4.0 network. NIC speeds were found acceptable on all 11 clients. The full
GCNdex32 benchmark suite was run on every system at three video resolutions: 640 by
480 pixels at 256 colors, 800 by 600 pixels with 32-bit or 24-bit color, and 1,024 by 768
pixels with 16-bit color. All but the Compaq Deskpro 2000 reached those color depths.
The GCNdex32 baseline is a 66-MHz 486 PC with 1M video, 2X CD-ROM drive and 1G IDE hard
drive. A GCNdex score of 3.5 means a system performed 312 times better than the
Floating-point and integer math scores primarily gauge processor performance, though
all benchmark scores can be improved by faster processors.
Video scores rate video card performance. Small-file access measures performance at
reading and writing to the hard drive, taking into account the disk cache. Large-file
access measures the drive performance when files are too large to be cached. CD-ROM access
measures the speed of the CD drive.
All systems used Windows 95 as the operating system, though vendors had the option of
installing Windows NT 4.0 Workstation.
The GCNdex PowerScores took into account GCNdex32 benchmarks; configuration quality,
including processor, RAM, storage, video, bays, card slots, ports, sound and other
features; an architecture rating; and an overall rating by the reviewer.
To focus on client performance, the equation positively weighted scores in the
following areas: NIC, accessibility, reliability and expandability.
Any system with component failures was penalized, as was any vendor that did not send
the system described in its written documentation.