Testing was open to all 120- and 133-MHz Pentiums. We requested that vendors configure their systems with a minimum of 32 MB of RAM, at least a 1-GB hard drive subsystem, and an MPC 2-compliant CD-ROM drive. Performance is the most heavily weighted (60 percent) category in determining winners in both processor classes. However, we also did a head-to-head features analysis and a detailed usability assessment (weighted at 20 percent each).
PERFORMANCE
We determined the performance of each system with a suite of BYTE low-level DOS tests and
NSTL's application-based
Windows and DOS benchmarks, along with NSTL's low-level InterMark Windows tests.
NSTL's Windows application suite consists of Microsoft Excel 5.0, Microsoft Word 6.0, Novell WordPerfect 6.0, Microsoft FoxPro 2.6, Autodesk AutoCAD re
lease 12, and Adobe PhotoShop 2.5. The DOS application suite includes WordPerfect 6.0, Lotus 1-2-3 release 3.4, and FoxPro 2.5. The application-based tests portray real-world performance by running macros that execute common functions of each application. For example, the Word for Windows benchmark includes subtests that measure search-and-replace functions, changing fonts, scrolling by page and line, spelling checking, and printing to a file.
Our DOS low-level test isolates the floating-point and integer performance of each system's CPU relative to a 90-MHz Dell Pentium. NSTL's InterMark subsystem-level Windows tests exercise the Windows Graphical Device Interface (GDI), as well as low-level graphics, CPU/FPU, and hard drive performance. The GDI component determines how well a system executes basic calls within Windows.
We installed Windows 95 on a freshly formatted 500-MB primary partition; we put our test files on a 500-MB extended partition. We ran the Windows tests with 1024- by 768-pixel res
olutions and 256 colors using default video drivers (when possible).
Due to the late release of Windows 95, the application-based tests were not automated (but will be for our next system review). We ran each test three times, or until we achieved consistent results. A fresh install of Windows 95 was re-created prior to each test using an NSTL-designed utility that builds a mirror image of the primary partition. We derived the overall performance score from the weighted average of the indexes of individual tests.
Note that performance results of the test systems running under the new 32-bit Windows OS cannot be directly compared to systems in previous reviews running under Windows for WorkGroups 3.11, even though we use the same benchmarks. Windows 95 offers significant speed enhancements over its 16-bit predecessor, such as dynamically configured virtual-memory settings, optimized video-driver code, and a revised file system featuring newer 32-bit protect-mode mini drivers. We used only 16-bit ap
plications for testing. It is likely that 32-bit applications will provide increased performance under Windows 95 (see
"32-Bit Performance Advantages"
).
EASE OF USE
For usability, we focused on two areas: system design and documentation. In looking at system design, we rated how easy it is to remove and replace the cover and how easy it is to upgrade the system's RAM and add mass-storage devices. We took points off if I/O connections blocked an expansion slot.
We also looked for a reset button and clearly marked I/O ports. We went into each system's setup utility to see if the integrated hard drive controller and built-in I/O ports could be disabled (if applicable). This is important if you want to use the I/O port's interrupt settings or want to switch to an alternate hard drive subsystem (i.e., from IDE to SCSI). We gave top honors to systems with vendor-specific manuals that were comprehensive, had easy-to-read diagrams, and offered up-to-date technical informati
on.
FEATURES
We asked each vendor to complete a lengthy questionnaire about its system's features and support options. We then weighted each feature and calculated an overall features score.
Important features to consider when choosing one of these systems are those related to expansion (e.g., the maximum RAM capacity, the number of floppy-drive bays, and available PCI and ISA/EISA expansion slots).
Features related to performance include the size of the secondary memory cache, the presence of EDO RAM, and a SCSI Wide hard drive interface. Intel claims that use of its Triton chip set can significantly boost system performance, especially when used with EDO RAM (see
"Triton-Based Pentiums"
). Systems configured with SCSI Wide hard drives, which offer data transfer rates of up to 20 MBps, display the best performance in our disk-intensive benchmarks.
Warranty and support policies are what frequently separate major PC manufacturers from second- and third-tie
r vendors. The length of the standard warranty is our highest rated feature.
Contributors
Anthony J. Lennon, Project Manager/NSTL
Siva Kumar, Technical Analyst/NSTL
John McDonough, Technical Editor/NSTL
Susan Colwell, Technical Editor/BYTE
photo_link (20 Kbytes)
NSTL project manager Anthony Lennon tests Pentium performance, features, and ease of use.
How We Tested
