| |
 |
|
Date |
:
Feb 1st, 2001 |
| Category |
:
Editorials |
| Manufacturer |
:
N/A |
| Author |
:
Jin-Wei Tioh |
|
WinBench 99 measures a hard disk's performance from two aspects. The first are low-level measurements, such as CPU utilization
(Disk CPU Utilization), sequential transfer rate (Disk/Read Transfer
Rate), and random access time (Disk Access Time). Next are the high-level benchmarks which ZD have termed the
Business Disk WinMark 99 and the High-End Disk WinMark
99.
The Business Disk WinMark 99 is basically a subset of ZD's Business Winstone 99. It basically attempts to emulate the disk access patterns of typical business applications. How exactly does it do this? According to ZD, "Business Winstone 99 keeps multiple applications open within each suite, and switches task between those applications and Netscape Navigator". These applications include :
- Corel WordPerfect 8
- Lotus SmartSuite
97
- Microsoft Office 98
We do realize that there are many people who are currently using either Internet Explorer 5/5.5. However, ZD ran into technical road blocks in trying to include IE as part of their test suite. The Business Disk WinMark is just the isolated disk access pattern of Winstone 99, and only one aggregate score is reported. It takes into account many other factors such as the ATA controller's speed, drivers, OS cache settings, as well as the system's memory subsystem (ie. the processor's L1/L2 cache and the system RAM). Thus, on first instinct, it would be reasonable to conclude that the Business Disk WinMark 99 should scale upwardly as external attributes of the systems, eg. CPU, RAM, ATA controller, etc. increase in speed/efficiency. After doing a couple of testruns on both a Celeron 466 and PIII 700E, we did indeed notice this trend.
Well then, after reading the above, you might well expect
that the High-End Disk WinMark 99 has something to do with the disk access patterns of "high-end" applications. You would be absolutely correct. The High-End Disk WinMark 99 is a subset of ZD's High-End Winstone 99, and it includes the disk access routines of the following applications :
- AVS/Express 3.4
- FrontPage 98
- MicroStation 95
- Photoshop 4.0
- Premiere 4.2
- Sound Forge 4.0
- Visual C++ 5.0
Having used quite of few of these applications, I can definitely say that they are rather demanding in terms of hard drive usage. However, there is an important difference in the way the High-End Winstone 99 runs the applications. The applications are
run serially, and individual application scores are presented in addition to a weighted average. Like the Business Disk WinMark, the High-End Disk WinMark also scales proportionally to a system's external attributes.
Out of the low-level drive parameters that are measured, only the sequential transfer rate and random access time will be published. During a few
test runs, WinBench 99 gave rather inconsistent figures for CPU
utilization. This could well be due to running WinBench 99 on Windows 2000, but anyhow,
we would not recommend relying on this figure. Instead, look to the results reported by IOMeter as a more consistent, if not reliable measure.
The Disk/Read Transfer Rate measures the sequential transfer rate of a drive across a partition. By default, WinBench 99 only reports the sequential transfer rate at the beginning and end of the partition, which would give the biggest delta if the the partition spanned the entire drive. However,
we have set WinBench 99 to save a graphical representation of the STR across the entire partition, as it is not only the maximum STR that is important, but how far along the drive can it maintain its maximum STR. Certainly, it is not much good in having a drive with an impressive maximum STR that significantly decreases along a greater portion of it?
Here is a sample of the STR graph that WinBench 99 generates
:
The Disk Access Time measures the average random access time across a partition. A drive's average random access time consists of the manufacturer's stated seek time, command overhead time, settle time and any other overheads. WinBench 99 will take all these into account to report an aggregate score.
Prior to running WinBench 99, the drive to be tested is partitioned with a maximum sized extended partition, and a logical drive of maximum size created and finally formatted with NTFS. Thus, the created logical drive is now labeled "E:", all ready to be tested. The whole suite of tests outlined above are repeated 3 times. The drive is repartitioned,
reformatted and the system cold-booted between each iteration, which would clear the drive's buffer as well as the operating system's disk cache.
|
|
