I have personally always been a bit more comfortable with the Ziff-Davis series of benchmarks than others, simply because they seem to provide more information about what they are doing, and why they are doing it. While performing the tests that resulted in this article, I found some interesting issues that will be covered in more detail in an upcoming Benchmark Examiner.
I have, somewhat arbitrarily, broken up the disk access issues into three categories: Boot time, Application Load Time, and Application Operation Time. Depending upon the way you use your system any or all of these may be relevant (or irrelevant) to what you perceive as the performance of the system.
As an example, I typically boot my system only once per day, at most. At home, my system will typically stay running for many days before it is shut down. The reason for this is that I like to jump onto the system for short periods between other activities, and waiting for the system to boot would prevent me from doing much in the 5 or 10 minutes I can spare at that particular moment. In addition, my wife will also occasionally use my system (even though she has her own) for exactly the same reason. At work, I do shut down my system at the end of the day, but it stays up from the time I come in until I leave for the evening. Therefore, since I boot my system only once per day, and it isn’t a server that has others waiting on it to come up quickly, boot time just isn’t an issue for me.
Another example has to do with application load time. At home, I am in and out of various applications but I leave my most frequently used ones up all the time (email and browser). Only when I have to edit an article do I launch my word processor, HTML editor, FTP application, graphics editor, etc. At that time, application load time is important, but normally it is not. Similarly, at work I will launch all of my frequently used applications right after booting (email, 3270 emulator, browser, bookmanager for technical publications, etc.) and they will stay up all day long. Therefore, application load time is somewhat important for me, but not a huge issue.
Obviously, this leaves the application operation time as 80% or more of my ‘wait’ time. As it turns out, this is primarily what Winstone and Content Creation are actually measuring. For those who have a different system usage, these benchmarks will not accurately reflect what they experience. Obviously, there is no way for such a benchmark to measure system boot time, which may be very important for some people who shut down their systems several times a day. For those who don’t have applications that they will use virtually every session and leave active, application load time will be very important and Winstone and Content Creation benchmarks will not provide an accurate measure of the performance these users will see.
Therefore, the results and recommendations in this article must be used wisely, with the system usage taken into careful consideration. Specifically, this means that these results mostly apply to users of office type applications (no 3D games, though possibly some 2D games during work breaks) on a system that is rarely rebooted and where the most frequently used applications are active constantly between system startup and shutdown.
Be the first to discuss this article!